Systems and methods for a multi-user adaptive simulation with dynamic competency assessment

Abstract

Methods, systems, and apparatuses are described for conducting adaptive simulations with multiple concurrent users, where each user occupies a distinct character role within a shared simulation scenario. An Al orchestration engine synthesizes independently-assessed user inputs to generate unified narrative adaptations while maintaining per-user information asymmetry. The systems dynamically determine which competency dimensions are relevant to score for each user response, generating competency scores for dimensions exceeding a relevance threshold and aggregating scores in persistent learning profiles using contextually relevant data points. An Al coaching process operates separately from assessment and narrative generation processes, having read-only access to assessment output and maintaining persistent conversational memory across sessions via an external memory service. The systems deliver adaptive content through lightweight browser-based immersive environments comprising spherical projections of images with assessment-responsive interaction points that adapt based on user performance.

Description

ATTORNEY DOCKET NO.: 38562.0001P1SYSTEMS AND METHODS FOR A MULTI-USER ADAPTIVE SIMULATION WITH DYNAMIC COMPETENCY ASSESSMENT CROSS REFERENCE TO RELATED PATENT APPLICATION

[0001] This Application claims priority to U.S. Provisional Application No.63 / 788,037, filed April 13, 2025, and U.S. Provisional Application No. 64 / 021,854, filed March 30, 2026, which are herein incorporated by reference in their entirety.BACKGROUND

[0002] Adaptive learning systems have evolved to provide personalized educational experiences by adjusting content difficulty, sequencing, and feedback based on learner performance. These systems typically employ techniques such as item-response theory', Bayesian knowledge estimation, and branching-scenario designs to tailor instruction to individual users. However, such systems generally operate in single-user modes where a learner interacts with pre-authored content that adapts primarily through difficulty adjustment or content reordering.

[0003] Current adaptive learning platforms face limitations in their ability to assess higher-order cognitive competencies such as critical reasoning, perspective synthesis, and collaborative decision-making. Existing systems tend to evaluate whether a user arrives at a correct answer but do not capture how the user arrives at their reasoning, what specific competency dimensions a particular response reveals, or how reasoning quality manifests when users interact with one another in shared scenarios.

[0004] Assessment methodologies in existing adaptive systems typically apply uniform scoring across all competency dimensions for every’ user response, regardless of whether a given response actually provides meaningful evidence for a particular dimension. For instance, a user's decision to negotiate rather than confront an adversary in a simulation may provide rich evidence of conflict resolution and communication competencies but may provide little or no evidence of factual domain knowledge. Uniform scoring across all dimensions can produce competency profiles diluted by assessments that lack contextual relevance, potentially reducing the accuracy and utility of learner data.

[0005] Multiplayer educational environments present additional technical challenges. While multiplayer games and collaborative platforms exist, these systems generally lack mechanisms for Al-driven narrative adaptation that responds to multiple concurrent users' inputs while maintaining per-user competency assessment. Coordinating multipleATTORNEY DOCKET NO.: 38562.0001P1users within a shared adaptive simulation requires processing concurrent inputs, synthesizing potentially conflicting decisions, managing information asymmetry between different user roles, and evaluating collaborative reasoning patterns that emerge from user interactions.

[0006] Al-powered coaching and tutoring systems have been developed to provide guidance to learners during educational activities. However, integrating coaching assistance alongside competency assessment raises concerns about measurement validity. When a coaching system provides hints or guidance during an assessed activity, distinguishing between a user's unassisted reasoning ability and coached performance becomes difficult. Maintaining separation between assessment processes and coaching processes while still enabling coaching to respond to assessment data presents architectural challenges.

[0007] Delivering immersive educational experiences in resource-constrained settings, such as schools with limited bandwidth, older devices, or restricted computing resources, presents further technical obstacles. High-fidelity' three-dimensional environments with complex rendering requirements may be impractical in such settings, yet simpler presentation formats may fail to engage learners or provide the contextual richness that supports deeper learning.SUMMARY

[0008] It is to be understood that both the following general description and the following detailed description are exemplary' and explanatory only and are not restrictive.

[0009] Methods, apparatuses, and systems for conducting multi-user adaptive simulations that provide dynamic competency assessment across multiple concurrent participants are described. Multiple users may engage in a shared simulation session where each user is assigned a distinct character role with character-specific knowledge and available actions, creating information asymmetry between participants. User inputs from each concurrent session are independently assessed to generate structured assessment outputs comprising competency evaluation data with per-dimension relevance weights, enabling selective scoring that only updates competency dimensions for which a given user response provides meaningful evidence. An orchestration process synthesizes the multiple assessment outputs to resolve conflicts between user decisions,ATTORNEY DOCKET NO.: 38562.0001P1detect interaction effects, and evaluate collaborative reasoning quality, producing a unified simulation state update from which character-specific narrative content is generated for each user. A coaching process operates separately from the assessment and narrative generation processes with read-only access to assessment data, enabling responsive coaching support through a distinct communication interface without contaminating the measurement of unassisted user reasoning. The system further supports immersive presentation through Al-generated panoramic environments displayed in browser-based viewers with assessment-responsive interaction points that adapt based on demonstrated user competencies.

[0010] In an embodiment, are methods comprising receiving, by a computing device, via a user interface, user input associated with a simulation session, generating, by one or more artificial intelligence (Al) modules, based on the user input, a structured assessment output, generating, by the one or more Al modules, based on the structured assessment output, adaptive narrative content for the simulation session, causing the adaptive narrative content to output via the user interface, and causing, based on the output of the adaptive narrative content, the simulation session to update.

[0011] In an embodiment, are methods comprising generating, by a computing device, a shared simulation session comprising a plurality of concurrent user sessions, wherein each concurrent user session of the plurality of concurrent user sessions is associated with data indicative of a character role within the shared simulation session, receiving, from each concurrent user session of the plurality of concurrent user sessions, user input data, generating, based on the user input data received from each user session, an assessment output associated with each concurrent user session, updating, based on the plurality7of assessment outputs, the shared simulation session, and generating, based on updating the shared simulation session and based on each character role of the plurality of concurrent user sessions, character content for each concurrent user session.

[0012] In an embodiment, are methods comprising determining, by a computing device, a persistent learning profile comprising a plurality of competency dimensions, receiving a user response associated with a shared simulation session, generating, by an artificial intelligence (Al) module, based on the user response and data indicative of the shared simulation session, a structured assessment output comprising a relevance weight for each competency dimension of the plurality of competency dimensions, generating, based on the structured assessment output, a plurality of competency scores for theATTORNEY DOCKET NO.: 38562.0001P1plurality of competency dimensions, and causing, based on the plurality of competency scores, the persistent learning profile to update the plurality of competency dimensions.

[0013] In an embodiment, are methods comprising receiving, by a computing device, from a user device, one or more user inputs associated with a shared simulation session, wherein the shared simulation session is output to the user device via a first communication interface, generating by a first artificial intelligence (Al) module, based on the one or more user inputs, a structured assessment output, generating, by a second Al module, based on the structured assessment output, adaptive narrative content, generating, by a third Al module, based on the structured assessment output, one or more responses, outputting, to the user device, via the first communication interface, the adaptive narrative content within the shared simulation session, and outputting, to the user device, via a second communication interface, the one or more responses.

[0014] In an embodiment, are methods comprising generating, by a computing device, an image associated with a shared simulation session, displaying the image according to an interior surface of a virtual sphere within a browser-based three-dimensional (3D) viewer, wherein a virtual camera is positioned at a center of the virtual sphere to create a panoramic viewing experience, receiving an assessment output comprising competency evaluation data for a user, configuring, based on the assessment output, a plurality of interaction points within the shared simulation session, displaying the configured plurality of interaction points within the shared simulation session for user interaction.

[0015] This summary is not intended to identify critical or essential features of the disclosure, but merely to summarize certain features and variations thereof. Other details and features will be described in the sections that follow.BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to provide understanding techniques described, the figures provide nonlimiting examples in accordance with one or more implementations of the present disclosure, in which:

[0017] FIG. 1 shows an example system for conducting adaptive simulations with multiple concurrent users;

[0018] FIG. 2 shows an example adaptive simulation cycle overview;

[0019] FIG. 3 shows an example game play system architecture;

[0020] FIG. 4 shows example dynamic skills transcript generation system;ATTORNEY DOCKET NO.: 38562.0001P1

[0021] FIG. 5 shows an example framework- agnostic assessment configuration system;

[0022] FIG. 6 shows an example system architecture for conducting adaptive simulations with multiple concurrent users;

[0023] FIG. 7 shows a flowchart of an example process for implementing shared simulation sessions with multiple concurrent users;

[0024] FIG. 8 shows an example management system for implementing shared simulation sessions with multiple concurrent users;

[0025] FIG. 9 shows an example assessment framework data flow;

[0026] FIG. 10 shows an example process for determining competency relevance;

[0027] FIG. 11 shows an example process for generating an assessment output, a narrative, and coaching responses;

[0028] FIG. 12 shows an example interface for displaying interaction points;

[0029] FIG. 13 shows an example curriculum context integration process;

[0030] FIG. 14 shows an example teacher-mediated feedback release system;

[0031] FIG. 15 shows an example multi-modal assignment system architecture;

[0032] FIG. 16 shows an example multi -character perspective system

[0033] FIG. 17 shows an example Al-generated assessment framework creation system;

[0034] FIG. 18 shows an example struggle detection and coaching trigger logic system;

[0035] FIG. 19 shows an example external memory' service system architecture;

[0036] FIG. 20 shows an example learning accommodation integration and privacy¬ compliance system;

[0037] FIG. 21 shows an example Al-powered pedagogical insights generation system;

[0038] FIG. 22 shows an example conflict resolution process;

[0039] FIGS. 23A-23I show example user interfaces;

[0040] FIG. 24 shows a flowchart of an example method;

[0041] FIG. 25 shows a flowchart of an example method;

[0042] FIG. 26 shows a flowchart of an example method;

[0043] FIG. 27 shows a flowchart of an example method; and

[0044] FIG. 28 shows a flowchart of an example method.ATTORNEY DOCKET NO.: 38562.0001P1DETAILED DESCRIPTION

[0045] As used in the specification and the appended claims, the singular forms “a,” “an,’" and “the” include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from “about” one particular value, and / or to “about” another particular value. When such a range is expressed, another configuration includes from the one particular value and / or to the other particular value. When values are expressed as approximations, by use of the antecedent “about,” it will be understood that the particular value forms another configuration. It will be further understood that the endpoints of each of the ranges are significant both in relation to the other endpoint, and independently of the other endpoint.

[0046] It is understood that when combinations, subsets, interactions, groups, etc. of components are described that, while specific reference of each various individual and collective combinations and permutations of these may not be explicitly described, each is specifically contemplated and described herein. This applies to all parts of this application including, but not limited to, steps in described methods. Thus, if there are a variety of additional steps that may be performed it is understood that each of these additional steps may be performed with any specific configuration or combination of configurations of the described methods.

[0047] As will be appreciated by one skilled in the art, hardware, software, or a combination of software and hardware may be implemented. Furthermore, the methods and systems may take the form of a computer program product on a computer-readable storage medium (non-transitory) having processor-executable instructions (e.g., computer software) embodied in the storage medium. Any suitable computer-readable storage medium may be utilized including hard disks, CD-ROMs, optical storage devices, magnetic storage devices, memristors, Non-Volatile Random Access Memory (NVRAM), flash memory', or a combination thereof.

[0048] Throughout this application reference is made to block diagrams and flowcharts. It will be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, respectively, may be implemented by processor-executable instructions. These processor-executable instructions may be loaded onto a computer (e.g.. a special purpose computer), or other programmable data processing apparatus to produce a machine, such that the processorexecutable instructions which execute on the computer or other programmable dataATTORNEY DOCKET NO.: 38562.0001P1processing apparatus create a device for implementing the functions specified in the flowchart block or blocks.

[0049] This detailed description may refer to a given entity performing some action. It should be understood that this language may in some cases mean that a system (e.g., a computer) owned and / or controlled by the given entity is actually performing the action.

[0050] Blocks of the block diagrams and flowcharts support combinations of devices for performing the specified functions, combinations of steps for performing the specified functions and program instruction means for performing the specified functions. It will also be understood that each block of the block diagrams and flowcharts, and combinations of blocks in the block diagrams and flowcharts, may be implemented by special purpose hardware-based computer systems that perform the specified functions or steps, or combinations of special purpose hardware and computer instructions.

[0051] Systems and methods are described for conducting adaptive simulations with multiple concurrent users participating in a shared narrative scenario. Each user occupies a distinct character role with associated knowledge, perspectives, and available actions that define what information that user can access within the simulation. When users submit inputs, an Al assessment process independently evaluates each user's response to generate competency scores across multiple dimensions, but only for those dimensions that are actually relevant to the specific response provided. An orchestration process then synthesizes the individual assessment outputs to update the shared simulation state, resolving any conflicts between user decisions and detecting how one user's choices affect others. A narrative generation process produces customized content for each user that reflects both the unified simulation state and that user's character-specific information access. Separately, an Al coaching process monitors assessment outputs and can proactively offer guidance when users appear to be struggling, while maintaining conversational memory across sessions through an external memory service.

[0052] The disclosed systems address several technical problems inherent in existing adaptive learning platforms. Conventional systems uniformly apply all assessment dimensions to every user response regardless of whether a given response actually provides meaningful evidence for a particular competency, resulting in diluted competency profiles that reduce the accuracy of learner data. The disclosed selective scoring approach computes relevance weights for each competency dimension andATTORNEY DOCKET NO.: 38562.0001P1generates scores only for dimensions exceeding a threshold, producing higher-fidelity profiles built from contextually appropriate data points. The systems further solve the technical challenge of coordinating multiple concurrent users in a shared adaptive environment while maintaining per-user assessment integrity and information asymmetry , which existing single-user adaptive systems and multiplayer platforms without Al-driven assessment cannot accomplish. The separation of the coaching process from the assessment and narrative processes addresses the technical problem of maintaining measurement validity when coaching assistance is provided during assessed activities, ensuring that the assessment captures unassisted user reasoning while still enabling responsive coaching support.

[0053] The method steps recited throughout this disclosure may be combined, omitted, rearranged, or otherwise reorganized with any of the figures presented herein and are not intended to be limited to the four corners of each sheet presented.

[0054] FIG. 1 shows an example system 100 for conducting adaptive simulations, or shared simulation sessions, with a plurality7of concurrent users session associated with a plurality of users. The system 100 may include a computing device 101, one or more electronic devices 102, and / or one or more servers 106. The computing device 101 may be in communication with the one or more electronic devices 102 and / or the one or more servers 106 via a network (e.g., network 162). The computing device 101 may comprise a server computer, a workstation, a personal computer, or any other suitable computing platform capable of executing the simulation software and Al processing modules.

[0055] The computing device 101 may include a bus 110, one or more processors 120, a memory7140, an input / output interface 160, a display 170, and a communication interface 180. The bus 110 may comprise one or more of a system bus. a memory bus, a peripheral bus, or any combination thereof using various bus architectures including Industry Standard Architecture, Peripheral Component Interconnect, and the like. In certain examples, the computing device 101 may omit at least one of the aforementioned elements or may additionally include other elements. The computing device 101 may comprise one or more of a tablet computer, a mobile phone, a smartphone, a laptop computer, a desktop computer, and the like.

[0056] The bus 110 may comprise a circuit for connecting the bus 110, the one or more processors 120, the memory 140, the input / output interface 160. the display 170. and / or the communication interface 180 to each other and for delivering communication (e.g., aATTORNEY DOCKET NO.: 38562.0001P1control message and / or data) between the bus 110, the one or more processors 120, the memory 140, the input / output interface 160, the display 170, and / or the communication interface 180.

[0057] The one or more processors 120 may include one or more of a Central Processing Unit (CPU), an Application Processor (AP), or a Communication Processor (CP). The one or more processors 120 may control, for example, at least one of the bus 110, the memory 140, the input / output interface 160, the display 170, and / or the communication interface 180 of the computing device 101 and / or may execute an arithmetic operation or data processing for communication. The processing (or controlling) operation of the one or more processors 120 according to various embodiments is described in detail with reference to the following drawings.

[0058] The processor-executable instructions executed by the one or more processors 120 may be stored and / or maintained by the memory 140. The memory 140 may include a volatile and / or non-volatile memory. The memory 140 may include random-access memory (RAM), flash memory, solid state or inertial disks, or any combination thereof. As an example, the memory' 140 may include an Embedded MultiMedia Card (eMMC). The memory 140 may store, for example, a command or data related to at least one of the bus 110, the one or more processors 120, the memory 140, the input / output interface 160, the display 170, and / or the communication interface 180 of the computing device 101. According to various examples, the memory 140 may store data and / or software or may comprise firmware. For example, the memory 140 may store an operating system 151, simulation program 152, simulation data 153, one or more artificial intelligence (Al) programs / modules 154, learning profile data 155, and / or assessment data 156.

[0059] The operating system 151 may include a kernel, a middleware, one or more Application Programming Interfaces (APIs), and / or the like, configured for controlling one or more functions of the computing deyice 101 and / or an external device (e.g., the one or more electrical devices 102, and / or the one or more servers 106). The memory 140 may include a computer-readable recording medium (e.g., a non-transitory computer-readable medium) having a program recorded therein to perform the methods according to various embodiments by the one or more processors 120.

[0060] The kernel may control or manage, for example, system resources (e.g., the bus 110, the one or more processors 120, the memory 140, etc.) used to execute an operation or function implemented in other programs (e.g., the middleware, the API, the simulationATTORNEY DOCKET NO.: 38562.0001P1software 152, and / or the one or more Al programs / modules 154). Further, the kernel may provide an interface capable of controlling or managing the system resources by accessing individual elements of the computing device 101 in the middleware, the API, the simulation software 152, and / or the one or more Al programs / modules 154.

[0061] The middleware may perform, for example, a mediation role, so that the API, the simulation software 152, and / or the one or more Al programs / modules 154 can communicate with the kernel to exchange data. Further, the middleware may handle one or more task requests received from the simulation software 152, and / or the one or more Al programs / modules 154 according to a priority. For example, the middleware may assign a priority of using the system resources (e.g., the bus 110, the one or more processors 120, or the memory 140) of the computing device 101 to the simulation software 152, and / or the one or more Al programs / modules 154. For example, the middleware may process the one or more task requests according to the priority' assigned to at least one of the application programs, and thus, may perform scheduling or load balancing on the one or more task requests.

[0062] The one or more APIs may include one or more interfaces or functions (e.g., instructions), for example, for file control, window control, video processing, and / or character control, as an interface capable of controlling a function provided the simulation software 152, and / or the one or more Al programs / modules 154 in the kernel or the middleware. As an example, the simulation software 152, and / or the one or more Al programs / modules 154 may be independent of each other or integrally combined, in whole or in part, into a single processing program.

[0063] The simulation program 152 may include logic (e.g., hardware, software, firmware, etc.) that may be implemented to cause the computing device 101 to conduct adaptive simulations with multiple concurrent users (e.g., shared simulation sessions). For example, the simulation program 152 may be configured to implement an adaptive learning engine that dynamically adjusts simulation content based on real-time assessment of user performance. For example, the simulation program 152 may cause the computing device 101 to modify narrative difficulty, adjust the complexity of decision options presented to users, and alter the pacing of simulation progression based on competency scores generated by the Al programs / modules 154. The simulation program 152 may further implement branching narrative structures where user decisions influence subsequent storylines, character interactions, and available action paths, creatingATTORNEY DOCKET NO.: 38562.0001P1personalized learning experiences that respond to each user's demonstrated strengths and identified areas for improvement. As an example, the simulation program 152 may be configured to support historical simulation scenarios where users assume roles of historical figures or participants in significant events. For example, the simulation program 152 may enable users to experience scenarios such as the American Revolution, the Great Depression, or other historical periods, with Al-generated narrative content that adapts based on user decisions and assessed understanding of historical concepts. The simulation program 152 may integrate with curriculum standards and learning objectives, enabling educators to deploy simulations that align with specific educational requirements while providing engaging, interactive experiences that promote deeper learning through perspective-taking and decision-making within authentic historical contexts. As an example, the simulation program 152 may be configured to support diverse subject matter domains beyond historical simulations. For example, the simulation program 152 may enable deployment of simulations for science education where users engage with scientific phenomena through inquiry’ -based scenarios, for civics education where users participate in democratic processes and policy debates, for literature education where users interact with characters and themes from literary’ works, or for professional training contexts such as medical decision-making, legal reasoning, or business strategy. The simulation program 152 may be configured to receive domainspecific content packages that define the narrative context, character roles, competency frameworks, and assessment criteria appropriate for each subject area, enabling the same underlying adaptive simulation engine to support diverse educational applications.

[0064] In an example, the computing device 101 may receive user input associated with a simulation session. For example, user input data may be received from a user device (e.g., one of the electronic devices 102) participating in the simulation session. The user input may comprise data indicative of one or more of a user decision, a user action, or a textual response. As an example, the user input may comprise a user interaction with the simulation session. As an example, the user input may comprise free-form text responses, selection of predefined decision options, or combinations thereof depending on the simulation design and the specific decision point presented to the user. In an example, the user input may be transmitted from the user device to server-side processing components via network communication protocols such as WebSocket connections for real-time bidirectional communication or HTTP requests for discreteATTORNEY DOCKET NO.: 38562.0001P1interactions. The user input may be structured according to predefined input formats that facilitate downstream processing by assessment and narrative generation modules.

[0065] The simulation software 152 may implement the Al programs / modules 154 to generate a structured assessment output based on the user input. The structured assessment output may comprise evaluation data comprising one or more competency scores across one or more competency dimensions. The one or more competency dimensions may comprise knowledge dimensions, theme dimensions, and skill dimensions. In an example, the simulation software 152 may cause the computing device 101 to maintain a persistent learning profile associated with a user of the user input. The persistent learning profile may comprise cumulative competency data aggregated across one or more simulation sessions. The simulation software 152 may cause the computing device 101 to update the persistent learning profile based on the one or more competency scores. In an example, the simulation software 152 may cause the computing device 101 to determine one or more competency gaps associated with the user based on the persistent learning profile. The simulation software 152 may cause the computing device 101 to adjust one or more parameters of a subsequent simulation session based on the one or more competency gaps. As an example, adjusting the one or more parameters may comprise one or more of modifying a difficulty level, reordering a sequence of simulation modules, or inserting a supplementary simulation associated with one or more competency dimensions associated with the one or more competency gaps. As an example, the simulation software 152 may implement the Al programs / modules 154 to cause the computing device 101 to transmit assessment prompts to external Al module services via application programming interfaces, wherein the assessment prompts may comprise the user input data, simulation context, character role information, and a predefined output schema that constrains the Al programs / modules 154 to generate validated structured data conforming to specified typed fields. As an example, the predefined output schema may ensure that the Al service produces consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations.

[0066] In an example, the structured assessment output may further comprise a relevance weight for each competency dimension of the one or more competency dimensions. As an example, the relevance weights may indicate the degree to which the user response provides evidence for each competency dimension, identified reasoningATTORNEY DOCKET NO.: 38562.0001P1patterns including logical structure and evidence cited, narrative adjustment signals indicating how the user's decision should influence simulation progression, and user struggle indicators that identify when a user appears to be experiencing genuine confusion or misunderstanding. As an example, generating the structured assessment output may comprise determining the one or more relevance weights satisfy a threshold, and generating, based on the one or more relevance weights satisfying the threshold, the one or more competency scores for the one or more competency dimensions. In an example, competency scores may not be generated for competency dimensions associated with relevance weights that do not satisfy the threshold.

[0067] In an example, the simulation session may comprise a shared simulation session accessible by a plurality of concurrent user sessions. Each concurrent user session of the plurality of concurrent user sessions may be associated with data indicative of a character role within the shared simulation session. In an example, a second user input associated with a second concurrent user session associated with a second character role may be received. The simulation software 152 may implement an Al module of the Al programs / modules 154 to generate second adaptive narrative content for the second concurrent user session. The second adaptive narrative content may be associated with the second character role and an information state associated with the second character role.

[0068] The simulation software 152 may implement the Al programs / modules 154 to generate adaptive narrative content for the simulation session based on the structured assessment output. The adaptive narrative content may comprise one or more of dialogue, scene descriptions, or decision options that reflect the evaluation data. As an example, the adaptive narrative content may comprise one or more of text, audio, one or more images, video, or an interactive three-dimensional environment, and wherein an output modality of the adaptive narrative content is selected based on one or more of a device capability associated with the computing device or a deployment context of the simulation session.

[0069] In an example, the simulation software 152 may implement the Al programs / modules 154 to cause the computing device 101 to process narrative adjustment signals from the structured assessment output to determine how the simulation session’s narrative should progress based on the user’s decisions and reasoning quality. In an example, the adaptive narrative content may further compriseATTORNEY DOCKET NO.: 38562.0001P1character-specific dialogue and scene descriptions that reflect the user’s character role and information state, character-appropriate decision options for subsequent simulation turns, and feedback integrated into the narrative as in-character reactions from nonplayer characters that reflect the quality of the user's reasoning. In an example, the Al programs / modules 154 may also receive curriculum context data to incorporate references to prior learning and cross-unit conceptual connections within the generated narrative content, enabling curriculum-coherent educational experiences that build upon previously covered material, for example.

[0070] In one example, the Al programs / modules 154 may comprise a single Al module. In another example, the Al programs / modules 154 may comprise a first Al module and a second Al module. The computing device 101 may implement the first Al module to generate the structured assessment output based on the user input and the computing device may implement the second Al module to the adaptive narrative content for the simulation session based on the structured assessment output.

[0071] The simulation software 152 may cause the computing device 101 to cause the adaptive narrative content to output via the user interface. The adaptive narrative content may be rendered within a browser-based immersive environment comprising a projection of an image (e.g., a spherical projection of an equirectan gular image), enabling immersive experiences (e.g., immersive panoramic experiences) without requiring specialized hardware or high-performance graphics processing.

[0072] The simulation software 152 may cause the computing device 101 to cause the simulation session to update based on the output of the adaptive narrative content. For example, the adaptive narrative content may cause the simulation session to adapt based on the structured assessment output, creating a personalized learning experience that responds to the user's demonstrated competencies and identified areas for improvement. In an example, the simulation session update may comprise updating the current narrative position within the simulation scenario, recording cumulative effects of user decisions on environmental conditions and non-player character states, modifying resource or constraint conditions, and updating relationship states between characters. The simulation software 152 may cause the computing device 101 to further update a persistent learning profile associated with the user based on the competency scores generated in the structured assessment output, enabling longitudinal tracking of learning progression across multiple simulation sessions. As an example, the persistent learningATTORNEY DOCKET NO.: 38562.0001P1profile may be configured to implement selective aggregation, wherein competency¬ scores may be aggregated only from simulation turns where the corresponding dimension was determined to be relevant based on relevance weights exceeding a configurable threshold.

[0073] In an example, the simulation software 152 may implement a first Al module of the Al programs / modules 154 to maintain a dynamic skills transcript associated with a user of the user input. The dynamic skills transcript may comprise a persistent profile maintained outside of any individual simulation session. The simulation software 152 may implement the first Al module to output data from the dynamic skills transcript to a second Al module of the Al programs / modules 154. The simulation software 152 may implement the second Al module to update the adaptive narrative content based on the data from the dynamic skills transcript and based on a cumulative profile of the user derived from one or more prior simulation sessions.

[0074] The simulation software 152 may7be further configured to manage execution of shared simulation sessions. For example, the simulation software 152 may be configured to maintain the shared simulation sessions, wherein the shared simulation sessions maybe accessible by a plurality of concurrent user sessions. In addition, the simulation software 152 may be configured to process user inputs, coordinate the Al programs / modules 154, and output character-specific narrative content to one or more users (e.g., via one or more user devices such as the one or more electronic devices 102).The simulation software 152 may be configured to implement orchestration logic that synthesizes (e.g., generates) assessment outputs (e.g., per-user assessment outputs) to update the shared simulation sessions (e.g., generate unified simulation state updates) in order to resolve conflicts between user decisions and detect interaction effects where one user's decision creates consequences for other users.

[0075] The Al module programs / modules 154 may be configured to implement one or more of assessment processes, orchestration processes, narrative generation processes, or coaching processes. The Al module programs / modules 154 may include logic (e.g., hardware, software, firmware, etc.) that may be implemented to perform Al-driven evaluation of user inputs, generate structured assessment outputs with competency scores and relevance weights, generate adaptive narrative content, and provide coaching interactions (e.g., messages, responses, etc.) based on assessment data. In one example, the Al module programs / modules 154 may be configured to interface with external AlATTORNEY DOCKET NO.: 38562.0001P1module services (e.g., via one or more of the servers 106) via application programming interfaces to leverage cloud-based Al processing capabilities. In another example, the Al module programs / modules 154 may be configured to execute Al modules locally on the computing device 101.

[0076] The learning profile data 155 may store and maintain persistent learning profiles with competency scores aggregated across simulation sessions (e.g., associated with each user). The learning profile data 155 may store cumulative competency data for each user wherein scores for each competency dimension are aggregated from simulation turns where that dimension was determined to be relevant to the user's specific response. For example, the learning profile data 155 enables the system 100 to track longitudinal learning progression and to adapt curriculum elements based on aggregated performance patterns. In an example, the Al programs / modules 154 may be configured to analyze longitudinal data associated with throughput data originating from gameplay obtained from each user of a plurality of users. The longitudinal data may comprise trends in a user’s problem-solving speed, improvement or decline in certain competencies, or patterns in choices across different scenarios. These insights may be logged in the learning profile data 155. The Al programs / modules 154 may also be configured to transmit suggestions for improvement of the simulation based on the throughput data, including suggestions related to the structure of narrative and feedback provided to users. In some examples, the Al programs / modules 154 may learn from anonymized and aggregated data across many users to improve accuracy and effectiveness, detecting patterns in how effective certain adaptations are and recommending sequences for new users who fit similar profiles. For example, the Al programs / modules 154 may learn that users who struggle with a certain concept benefit if the narrative model provides a specific type of hint at a certain point, and may adjust the behavior of the narrative model proactively based on this knowledge. In another example, the Al programs / modules 154 may update the weighting of scoring criteria in the assessment process based on a determination that initial weighting did not correlate well with final learning outcomes for many users. The Al programs / modules 154 may also synthesize inputs from multiple sources including simulation data comprising results and feedback from each simulation module, cross-simulation analysis comprising comparisons of user performance across different contexts, teacher or instructor input comprising feedback or ratings provided by educators who review the learning profile data 155, and externalATTORNEY DOCKET NO.: 38562.0001P1assessments comprising results from outside assessments such as standardized tests or quizzes that may be fed into the system to refine the learner profile.

[0077] The assessment data 156 may store structured assessment output including competency evaluation data, relevance weights, identified reasoning patterns, struggle indicators, and narrative adjustment signals generated during simulation sessions. For example, the assessment data 156 provides the data foundation for the coaching process to monitor user performance across multiple turns and to identify intervention conditions that warrant proactive coaching interactions. In an example, the Al programs / modules 154 may be configured to evaluate an effort level of the user and use the effort level as a basis for reconfiguring a particular aspect of the simulation to better motivate the user to improve the effort level. The system may learn where users are not trying with certain responses and reconfigure that particular scene or scenario accordingly. For example, the Al programs / modules 154 may detect that a user is providing minimal-effort responses to a particular scene and may reconfigure that scene to present the content in a different format, adjust the difficulty, or introduce additional motivational elements to encourage more thoughtful engagement.

[0078] As an example, the simulation program 152 may cause the computing device 101 to generate a shared simulation session comprising a plurality of concurrent user sessions. Each concurrent user session of the plurality of concurrent user sessions may be associated with data indicative of a character role within the shared simulation session. In an example, the shared simulation session may be accessible by the plurality of concurrent user sessions. The data indicative of each character role may comprise data defining character-specific knowledge and one or more available actions, wherein the data indicative of each character role defines a subset of the shared simulation session according to a character of a corresponding character role of the plurality of character roles. The computing device 101 may receive user input data from a user device (e.g., a electronic device 102) associated with each concurrent user session of the plurality of concurrent user sessions. The simulation program 152 may cause the computing device 101 to generate an assessment output associated with each concurrent user session based on the user input data received from each concurrent user session. Each assessment output of the plurality of assessment outputs may comprise evaluation data and one or more narrative adjustment signals. The simulation program 152 may cause the computing device 101 to update the shared simulation session based on theATTORNEY DOCKET NO.: 38562.0001P1plurality of assessment outputs. For example, the simulation program 152 may cause the computing device 101 to update the shared simulation session based one or more of: resolving conflicts between mutually exclusive user decisions; detecting interaction effects of a user decision that creates consequences for other users; and evaluating collaborative reasoning quality. The simulation program 152 may cause the computing device 101 to generate character content (e.g., adaptive narrative content) for each concurrent user session based on updating the shared simulation session and based on each character role of the plurality of concurrent user sessions. The character content for each concurrent user session may comprise one or more of character-specific dialogue and scene descriptions, character-appropriate decision options, and feedback integrated into the shared simulation session as in-character reactions from non-player characters. The simulation program 152 may cause the computing device 101 to output the character content, in addition to the shared simulation session to the user devices (e.g., the electronic devices 102, wherein each user device may display the character content within the shared simulation session for that corresponding user.

[0079] In an example, the simulation program 152 may cause the computing device 101 to determine a plurality of relevance weights for a plurality of competency dimensions based on the user input data. The plurality of competency dimensions may comprise one or more of one or more knowledge dimensions, one or more theme dimensions, or one or more skill dimensions. The simulation program 152 may cause the computing device 101 to generate one or more competency scores associated with one or more competency dimensions, of the plurality of competency dimensions, associated with the one or more relevance weights based on one or more relevance weights of the plurality of relevance weights satisfying a threshold.

[0080] In an example, the simulation program 152 may cause the computing device 101 to generate competency data for a plurality of users associated with the plurality of concurrent user sessions based on updating the shared simulation session. The computing device 101 may store (e.g., in the memory 140) the competency data for each user of the plurality of users in a corresponding persistent learning profile (e.g., in learning profile data 155 stored in the memory 140).

[0081] As an example, the simulation program 152 may cause the computing device 101 to determine a persistent learning profile comprising a plurality of competency dimensions. For example, the computing device 101 may retrieve the persistent learningATTORNEY DOCKET NO.: 38562.0001P1profile from the learning profiled data 155 stored in the memory 140. In an example, persistent learning profile may be configured to maintain, for each competency dimension of the plurality of competency dimensions, a cumulative score derived from user turns associated with the shared simulation session, wherein the corresponding competency dimension was determined to be relevant. The plurality of competency dimensions may comprise one or more of one or more knowledge dimensions, one or more theme dimensions, or one or more skill dimensions. The computing device 101 may receive a user response (e g., from a user device such as one of the electronic devices 102) associated with a shared simulation session. The simulation program 152 may cause the computing device 101 to implement one of the Al programs / modules 154 to generate a structured assessment output comprising a relevance weight for each competency dimension of the plurality’ of competency dimensions based on the user response and data indicative of the shared simulation session. For example, the simulation program 152 may cause the computing device 101 to implement the Al program / module 154 to generate the structured assessment output further based on the curriculum context data. The structured assessment output may comprise one or more of a relevance weight for each competency dimension of the plurality of competency dimensions, an overall numerical score, per-dimension scores, category -level relevance weights, per-dimension applicability flags, justification text, or a commentary field. The simulation program 152 may cause the computing device 101 to generate a plurality of competency scores for the plurality of competency dimensions based on the structured assessment output. The simulation program 152 may cause the computing device 101 to update the persistent learning profile by updating the plurality of competency dimensions based on the plurality of competency scores.

[0082] In an example, the computing device 101 may receive curriculum context data from a broader course associated with the shared simulation session. The curriculum context data may comprise prior unit content, previously covered concepts and skills, and course-wide learning objectives. The simulation data 153 may include pre-set or pre-defined high-level plot overviews that establish guardrails within which users may make changes to the narrative including an outcome of the narrative. At the end of a simulation, the user may be prompted to compare results of the user to an actual outcome so that the user may analyze why the results may differ from the actual outcome.ATTORNEY DOCKET NO.: 38562.0001P1

[0083] In an example, the simulation program 152 may cause the computing device 101 to determine one or more additional competency scores associated with one or more additional shared simulation sessions based on the persistent learning profile. The simulation program 152 may cause the computing device 101 to aggregate the one or more additional competency scores. The simulation program 152 may cause the computing device 101 to adjust one or more curriculum elements associated with the shared simulation session based on the aggregated one or more competency scores. For example, the simulation program 152 may cause the computing device 101 to adjust one or more curriculum elements associated with the shared simulation session by one or more of reordering a sequence of simulation modules to address identified competency gaps, modifying emphasis within instructional materials based on identified weaknesses, inserting supplementary simulation episodes targeting specific competency dimensions, or generating customized assignments targeting competency dimensions identified as weak points in the persistent learning profile.

[0084] As an example, the simulation program 152 may cause the computing device 101 to output one or more coaching messages (e.g., one or more responses) based on implementing the coaching process. The computing device 101 may receive one or more user inputs associated with a shared simulation session from a user device (e.g., one of the electronic devices 102). As an example, the computing device 101 may output the shared simulation session to the user device, wherein the user device may display the shared simulation session via a first communication interface, or channel. The simulation program 152 may cause the computing device 101 to implement a first Al program / module 154 of the Al programs / modules 154 to generate a structured assessment output based on the one or more user inputs. The simulation program 152 may cause the computing device 101 to implement a second Al program / module 154 of the Al programs / modules 154 to generate adaptive narrative content based on the structured assessment output. The simulation program 152 may cause the computing device 101 to implement a third Al program / module 154 of the Al programs / modules 154 to generate one or more responses (e.g., one or more coaching messages) based on the structured assessment output. For example, the third Al program / module 154 may generate the one or more responses across a plurality of user turns associated with the shared simulation session. In an example, the third Al program / module 154 may comprise an independent Al module configured to receive the structured assessmentATTORNEY DOCKET NO.: 38562.0001P1output via read-only access. In an example, the third Al program / module 154 may receive anonymized user personally identifiable information, wherein the user’s personally identifiable information may be anonymized via a hashing function. In an example, the third Al program / module 154 may generate the one or more responses based on the structured assessment output satisfying one or more conditions. The one or more conditions may comprise one or more of a struggle indicator being set to true, consecutive assessment scores falling below a defined threshold, identified misconceptions recurring across a plurality of user turns, or engagement pattern changes. In an example, the simulation program 152 may cause the computing device 101 to generate the one or more responses according to one or more of tone, vocabulary complexify, or support level based on user learning accommodations. The computing device 101 may output the adaptive narrative content within the shared simulation session and the one or more responses to the user device, wherein the user device may display the adaptive narrative content within the shared simulation session via the first communication interface and display the one or more responses via a second communication interface. In an example, the structured assessment output may comprise one or more of one or more competency scores, one or more relevance weights, or one or more user struggle indicators.

[0085] In an example, the system 100 may include one or more derivative versions of the Al programs / modules 154. Each of the one or more derivative versions of the Al programs / modules 154 may be individually configured for use by individual teachers and / or organizations such as learning institutions. Each derivative version of the Al programs / modules 154 may be configured to receive input from teachers including ratings of feedback provided to users and work samples of feedback the teachers provided directly to users that may differ from feedback originating from the system 100. The work samples of feedback may serve as a basis for the system 100 to create an adjusted version of the Al programs / modules 154 configured to provide feedback and narratives that correspond to the teacher’s style and preferences. This enables one school’s simulation experience to differ from another’s, aligned with their respective curriculum or teaching style. For example, a teacher may rate the Al's feedback on student responses and provide examples of the teacher's own feedback on the same student responses. Over time, the derivative version of the Al programs / modules 154 may learn the teacher's style and modify the narrative and assessment to match it. TheATTORNEY DOCKET NO.: 38562.0001P1derivative version may also receive input from organizations such as schools, enabling organization-level customization where the simulation experience is aligned with institutional curriculum or pedagogical style while still using the same core engine.

[0086] In an example, the simulation program 152 may cause the computing device 101 to determine one or more competencies associated with a user and dynamically generate one or more lesson plans corresponding to the determined competencies. As an example, the computing device 101 may rely upon pre-set or pre-defined lesson plans with textbased summaries, recommended videos, cumulative assessments, and / or suggested classroom activities as a basis for generating varying aspects of the simulation. The dynamic generation of lesson plans may adjust based on feedback, intermediary assessment data, and / or overall competency data stored in the learning profile data 155.For example, the lesson plans may comprise text-based summaries of key concepts, recommended educational videos, cumulative assessments that evaluate learning across multiple simulation episodes, and suggested classroom activities that complement the simulation experience. The Al programs / modules 154 may also dynamically generate or adjust educational content based on identified competencies, including formulating personalized practice exercises or hints aligned with the user's needs as determined from the simulation.

[0087] In an example, the simulation program 152 may be configured to integrate with physical learning tools such as engineering-based instruments, surgical instruments, and the like. The simulation may provide instructions prompting the user to interact with physical learning tools to advance through the simulation. Results of physical tasks may be captured by sensors or entered by the user and fed back as additional data points to the assessment process, enabling the simulation to blend experiential learning with the adaptive narrative. For example, the one or more instructions may comprise methods for performing a surgical procedure using surgical instruments, a step of a process in building a widget or performing any other engineering-based action using engineeringbased instruments, or any other actions related to any other physical learning tools. The result of a physical task, such as whether the task was completed successfully or how many attempts were required, may be fed back as input to the assessment process, and the narrative may then continue based on that result. For example, the narrative content may include references to what the user did with the physical tool, such as ‘‘Because you successfully assembled the circuit, the device in the story powers on.”ATTORNEY DOCKET NO.: 38562.0001P1

[0088] In an example, the system 100 may be applicable to domains beyond educational settings. The system 100 may be deployed for interactive storytelling where a simulation's plot adapts to a player's choices and play style, for professional training simulations such as medical or military training that adjust scenarios based on a trainee's performance, or for entertainment experiences such as role-playing games that react to the player in unique ways. In a professional training context, the system 100 may simulate emergencies or strategic scenarios and adapt to the trainee's decisions, logging competencies and maintaining a profile of the trainee over multiple training modules to track improvement and readiness. In an entertainment context, the Al programs / modules 154 may observe a player's style, such as whether the player prefers diplomatic solutions or direct confrontation, whether the player explores every area or sticks to the main plot, and whether the player is adept at solving riddles or gives up quickly, and may use these observations as data points in the learning profile data 155 to personalize subsequent simulation experiences.

[0089] In an example, the system 100 may be configured as a library of modular simulations, each offering a distinct interactive experience with specific learning objectives or story lines. Each simulation module may be considered a self-contained unit. As an example, the modular design allows flexibility and scalability’ where new modules may be added to the library over time, and users may navigate through modules in a sequence suited to their learning path or interests. All user interactions may' feed into the learning profile data 155 that resides outside any single simulation, enabling future simulations to automatically tailor their narrative and difficulty based on the user's past performance across modules. Each simulation module may have its own dedicated narrative model instance with domain-specific knowledge or constraints tailored to that module's theme and objectives. For example, if a user completes a first simulation module, the learning profile data 155 may contain the outcomes and assessments from that module. When the user launches a second simulation module, the narrative and challenges in the second module may be tailored from the outset based on what the user learned or struggled with in the first module. This may' manifest as the storyline in the second module acknowledging the user's prior achievements, or the difficulty’ in the second module being adjusted such as skipping topics the user has mastered or revisiting topics the user found difficult. The Al programs / modules 154 may function as a director Al that selects which simulation module should come next in a curriculum, altersATTORNEY DOCKET NO.: 38562.0001P1overarching story arcs, or suggests review modules for areas where the user struggled. The Al programs / modules 154 may also cause the computing device 101 to load or preload specific information, scenarios, or non-player characters based on user patterns detected in the learning profile data 155.

[0090] In an example, a skill level of the user may be selected at any point during interaction with the simulation or determined through interaction with the simulation. The skill level may be manually adjusted by the user at any point, causing the simulation to adjust a difficulty level accordingly. Additionally, the Al programs / modules 154 may be configured to autonomously determine the skill level of the user and automatically adjust the difficulty level based on the determination. For example, the Al programs / modules 154 may utilize one or more means of machine learning or neural network functionality to autonomously determine the skill level of the user based on the user's interaction patterns. The real-time adjustments made to the simulation may be performed regardless of the content and / or length contained within any type of user input, including both multiple choice responses and free-form text responses.

[0091] The simulation program 152 may cause the computing device 101 to generate an image associated with a shared simulation session. The image may comprise an equirectangular image, a cubemap projection, a cylindrical panorama, or other immersive display formats suitable for panoramic viewing experiences. The simulation program 152 may cause the computing device 101 to display the image according to an interior surface of a virtual sphere within a browser-based three-dimensional (3D) viewer. In one example, the electronic devices 102 may be configured to generate and display the image according to the interior surface of the virtual sphere within the browser-based 3D viewer. In another example, the image may be generated by one of the Al programs / modules 154 based on simulation context parameters comprising one or more of historical period, geographic location, time of day, weather conditions, narrative state, or emotional tone. The computing device 101 may receive an assessment output comprising competency evaluation data for a user. In one example, the computing device may receive the assessment output from a user device (e.g., one of the electronic devices 102). In another example, the user device may generate the assessment output. The simulation program 152 may cause the computing device 101 to generate a plurality of interaction points within the shared simulation session based on the assessmentATTORNEY DOCKET NO.: 38562.0001P1output. For example, the simulation program 152 may cause the computing device 101 to configure the plurality of interaction points within the shared simulation session by determining content and availability of each interaction point of the plurality of interaction points based on competency gaps or demonstrated mastery areas associated with the user. For example, the simulation program 152 may cause the computing device 101 to configure the plurality of interaction points by emphasizing interaction points targeting competency dimensions identified as weak in the assessment output and surfacing more complex decision opportunities when mastery is demonstrated. The plurality of interaction points may be organized as templates defining interaction mechanics. Each interaction point of the plurality of interaction points may be associated with a character role associated with the user, wherein each character role may be configured to cause the corresponding interaction point to display characterspecific visual elements and information overlays to the corresponding user (e.g., via a user device such as one of the electronic devices 102). In an example, the simulation program 152 may cause the computing device 101 to implement an Al program / module 154 to determine specific content inserted into each interaction point based on narrative state information and user assessment performance. The computing device 101 may display the configured plurality of interaction points within the shared simulation session for user interaction. In an example, the electronic devices 102 may be configured to display the configured plurality of interaction points within the shared simulation session for user interaction. In an example, one or more of the computing device 101 or the electronic devices 102 may display character visual assets as stylized two-dimensional (2D) billboard elements within the shared simulation session, wherein the billboard elements may comprise an orientation facing the virtual camera.

[0092] The input / output interface 160 may include an interface for delivering an instruction or data input from a user (e.g., an operator of the computing device 101) or from a different external device (e.g., the electronic devices 102) to the different elements of the computing device 101. Further, the input / output interface 160 may output an instruction or data received from one or more elements of the computing device 101 to one or more external devices (e.g., the electronic device 102 and / or the servers 106). For example, the input / output interface 160 may comprise one or more of a touch screen interface, a keyboard, a mouse, a microphone, and the like for receiving the user input. The input / output interface 160 may further comprise eye trackingATTORNEY DOCKET NO.: 38562.0001P1capabilities using cameras and mouse pointer movement tracking to detect user interactive input, which may bring additional weight to the assessment of user responses. For example, the eye tracking capabilities may detect user gaze patterns on screen, which may bring additional weight to the assessment of user responses by providing metadata about user engagement and attention. As an example, eye tracking data may bring more weight to a seemingly null multiple choice response by indicating whether the user read and considered the available options before selecting. In an example, the input / output interface 160 may be configured to accept diverse input modalities including typed text, spoken language, selections within a game, gestures, or other actions. For example, the input / output interface 160 may accept audio input including spoken natural language, wherein the Al programs / modules 154 may be configured to interpret spoken input directly using speech recognition and natural language understanding, without requiring the input to be converted into text by a separate intermediate step. For example, this may enable natural user interactions such as talking to characters in a simulation and having the Al understand the intent and tone of the speech. In some examples, an audio analysis model may directly feed the assessment process with text transcripts or emotion analysis from the user's speech.

[0093] The communication interface 180 may establish, for example, communication between the computing device 101 and one or more external devices (e.g., the electronic device 102, and / or the server 106). In an example, the communication interface 180 may communicate with one or more of the external devices (e.g., the electronic devices 102, and / or the servers 106) by being connected to a network 162 through wireless communication or wired communication. The network 162 may include, for example, at least one of a telecommunications network, a computer network (e.g., LAN or WAN), the Internet, and / or a telephone network. The network 162 may facilitate communication between the computing device 101, the electronic devices 102, and the servers 106. For example, the network 162 may enable geographically distributed users to participate in shared simulation sessions and may provide connectivity to cloud-based Al services and external memory services utilized by the coaching process.

[0094] The communication interface 180 may be configured to communicate with one or more of the external devices (e.g., the electronic devices 102, and / or the servers 106) via a wireless communication interface. In an example, as a cellular communication protocol, the wireless communication interface may use at least one of Long-TermATTORNEY DOCKET NO.: 38562.0001P1Evolution (LTE), LTE Advance (LTE-A), Code Division Multiple Access (CDMA), Wideband CDMA (WCDMA). Universal Mobile Telecommunications System (UMTS). Wireless Broadband (WiBro), Global System for Mobile Communications (GSM), and the like. In an example, the communication interface 180 may include or be communicably coupled to a transmitter, receiver and / or transceiver for communication with one or more of the external devices (e.g., the electronic devices 102. and / or the servers 106).

[0095] The display 170 may comprise one or more of a television, an audio / video monitor, a streaming device, and the like. The display 170 may include various types of displays, for example, a Liquid Crystal Display (LCD) display, a Light Emitting Diode (LED) display, an Organic Light-Emitting Diode (OLED) display, a MicroElectroMechanical Systems (MEMS) display, or an electronic paper display. In an example, the display 170 may be configured as a part of the computing device 101 or as a separate external device in communication with the computing device 101 via the communication interface 180. The display 170 may be configured to display a user interface. For example, the display 170 may be configured to display the spherical projections of images (e.g., equirectangular images, cubemap projections, cylindrical panoramas, or other immersive display formats suitable for panoramic viewing experiences) with configurable interaction points. In one example, the display 170 may be utilized for local user sessions. In another example, shared simulation session content may be provided to remote users via the network 162. In an example, the display device 170 may be in communication with audio output devices (e.g.. earphones, speakers) for outputting audio signals.

[0096] Each of the electronic devices 102 may comprise a user device such as a mobile device, a smart phone, a tablet computer, a desktop computer, and the like. Each of the electronic devices 102 may be configured to execute a client-side application (e.g., implemented as a web browser-based application) that renders the shared simulation sessions (e.g., simulation environment) and provides the user interface through which users submit inputs and receive narrative content. For example, each of the electronic devices 102 may be configured to display the character content for each concurrent user session within the shared simulation session via a browser-based immersive environment comprising a spherical projection of an image. The browser-based immersive environment may comprise a plurality of interaction points comprising content andATTORNEY DOCKET NO.: 38562.0001P1availability based on the assessment output. The plurality of interaction points may be configured to cause the shared simulation session to adapt to the plurality of character roles and a plurality of competency levels associated with a plurality of users associated with the plurality of concurrent user sessions.

[0097] The one or more servers 106 may include one or more groups of servers. The servers 106 may be configured to provide remote computing resources for Al processing, including assessment services, narrative generation, and external memory services for the coaching process. The servers 106 may comprise one or more cloud-based server instances, dedicated application servers, or distributed computing resources that augment the processing capabilities of the computing device 101. For example, all or some of the operations executed by the computing device 101 may be executed in a different one or a plurality of electronic devices (e.g., one or more of the electronic devices 102 and / or one or more of the servers 106). As an example, one or more of the servers 106 may be configured to host external Al model services that receive assessment prompts from the computing device 101 and return structured assessment output conforming to predefined output schemas. For example, if the computing device 101 needs to perform a certain function or service either automatically or based on a request, the computing device 101 may request at least some parts of functions related thereto alternatively or additionally to a different electronic device (e.g., one or more of the electronic devices 102 and / or one or more of the servers 106) instead of executing the function or the service autonomously. The different electronic devices (e.g., the electronic devices 102 and / or the server 106) may execute the requested function or additional function, and may¬ deliver a result thereof to the computing device 101. As an example, one or more of the servers 106 may store the Al programs / modules 154. The computing device 101 may send simulation data 153 to one or more of the servers 106, wherein one or more of the Al programs / modules 154 stored on one or more of the servers 106 may process the simulation data 153 and send the processed data to the computing device 101 and / or one or more of the electronic devices 102 for further processing. In an example, one or more of the servers 106 may also host an external memory^ service that is configured to maintain persistent conversational state for the coaching process across shared simulation sessions, storing conversation history and derived insights about user learning patterns.ATTORNEY DOCKET NO.: 38562.0001P1

[0098] In one example, the system 100 may be implemented in a distributed configuration where the computing device 101 serves as an application server coordinating multiple server 106 instances that provide specialized Al processing services. In such configurations, assessment operations, narrative generation, and coaching processes may execute on separate server instances that communicate via the network 162, enabling horizontal scaling to accommodate large numbers of concurrent users. In another example, the system 100 may be implemented in a centralized configuration where all processing occurs on the computing device 101 without reliance on external server resources, which may be appropriate for deployments with limited network connectivity or enhanced privacy requirements.

[0099] In an example, the system 100 may be implemented with hybrid processing configurations where lightweight Al operations such as coaching response generation are performed locally on the electronic devices 102 while computationally intensive operations such as multi-user orchestration synthesis (e.g.. generation) are performed on the computing device 101 or one or more of the servers 106. As an example, this hybrid approach enables responsive user experiences while leveraging centralized resources for complex processing tasks.

[0100] The components of the system 100 may be implemented using specific technical structures and processes. The Al module programs / modules 154 may be configured to employ structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing without error-prone text parsing. The simulation software 152 may be configured to implement conflict resolution algorithms that evaluate assessment scores and narrative logic using defined computational procedures to determine appropriate outcomes when user decisions are mutually exclusive. The learning profile data 155 may be configured to implement weighted aggregation algorithms that combine new scores with prior scores using relevance-weighted averaging, maintaining running statistics for each competency dimension including count, sum, and average relevance. As an example, the system 100 provides measurable technical benefits including reduced assessment data dilution through selective scoring of relevant competency dimensions, improved measurement validity through architectural separation of coaching and assessment processes, and enhanced scalability through parallel independent assessment of multiple concurrent users. The system 100 enables real-time processing of multiple user inputs withinATTORNEY DOCKET NO.: 38562.0001P1defined collection windows, synthesis of potentially conflicting decisions into coherent narrative outcomes, and maintenance of distinct information states for each user without cross-contamination.

[0101] The system 100 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing platforms. Prior systems uniformly apply all assessment dimensions to every user response regardless of contextual relevance, resulting in diluted competency profiles with reduced accuracy. Prior systems also lack mechanisms for coordinating multiple concurrent users in a shared adaptive environment while maintaining per-user assessment integrity and information asymmetry. Additionally, prior systems that provide coaching assistance during assessed activities cannot distinguish between unassisted user reasoning and coached performance, compromising measurement validity. The arrangement and operation of the components in the system 100 address these technical problems through the Al module programs / modules 154 that implement selective scoring based on per-response relevance determination, the simulation software 152 that maintains shared simulation states with per-user information asymmetry, and the architectural separation between assessment and coaching processes that preserves measurement validity while enabling responsive coaching support. This structural arrangement enables multi-user adaptive simulation with dynamic competency assessment that was not achievable with prior system architectures.

[0102] FIG. 2 shows an example adaptive simulation cycle overview 200 illustrating a four-stage circular process for conducting adaptive simulations with multiple concurrent users. The adaptive simulation cycle 200 may be configured to provide a continuous feedback loop architecture that enables real-time adaptation of simulation content based on user performance assessment, educator coaching input, and artificial intelligence processing. The adaptive simulation cycle 200 may represent the operational workflow through which the adaptive simulation platform processes user interactions, evaluates competencies, incorporates coaching guidance, and generates adapted content for subsequent user engagement.

[0103] The adaptive simulation cycle 200 may comprise a multi-user gameplay stage 202 that represents the phase where multiple concurrent users participate in a shared simulation session through their respective user devices. The multi-user gameplay stage 202 may be configured to receive user inputs from a plurality of concurrent userATTORNEY DOCKET NO.: 38562.0001P1sessions, wherein each user session may be associated with a distinct character role within the shared simulation session. The multi-user gameplay stage 202 may implement client-side application instances executing within web browsers on user devices, with each instance maintaining a persistent connection to server-side processing components via network communication protocols. The multi-user gameplay stage 202 may be configured to render character-specific narrative content, present decision options appropriate to each user's character role, and capture user responses including free-form text inputs, selection of predefined options, and other interaction modalities supported by the simulation design.

[0104] The adaptive simulation cycle 200 may comprise a data assessment stage 204 that receives user inputs from the multi-user gameplay stage 202 for evaluation against competency frameworks associated with the shared simulation session. The data assessment stage 204 may be configured to process user responses and generate structured assessment outputs comprising competency scores across multiple dimensions including knowledge, theme, and skill categories. The data assessment stage 204 may implement artificial intelligence processing to analyze the quality, reasoning patterns, and evidence utilization demonstrated in user responses. The data assessment stage 204 may generate per-dimension relevance weights indicating the degree to which each user response provides meaningful evidence for particular competency dimensions, enabling selective scoring that updates only those dimensions for which the response provides contextually appropriate assessment data. The data assessment stage 204 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including user response data, simulation context, character role information, and predefined output schemas that constrain Al service responses to validated structured data formats.

[0105] The adaptive simulation cycle 200 may comprise a teacher coaching stage 206 that represents the phase where educators and AT coaching modules provide guidance and feedback to users based on assessment data generated by the data assessment stage 204. The teacher coaching stage 206 may be configured to operate with read-only access to assessment output, ensuring that coaching guidance does not contaminate the assessment's measurement of unassisted user reasoning. The teacher coaching stage 206 may implement proactive coaching interventions when assessment output satisfies predefined intervention conditions, such as when a struggle indicator is set to true, whenATTORNEY DOCKET NO.: 38562.0001P1consecutive assessment scores fall below a defined threshold, or when identified misconceptions recur across multiple user turns. The teacher coaching stage 206 may interface with an external memory service that maintains persistent conversational state for the coaching process across simulation sessions, storing conversation history and derived insights about user learning patterns and communication preferences. The teacher coaching stage 206 may deliver coaching messages through a communication interface that is visually and functionally distinct from the simulation narrative output, enabling users to distinguish between assessed simulation content and auxiliary coaching guidance.

[0106] The adaptive simulation cycle 200 may comprise an Al adaptation stage 208 that receives assessment data from the data assessment stage 204 and processes this data to adapt the simulation narrative, difficulty, and content for subsequent user interactions within the multi-user gameplay stage 202. The Al adaptation stage 208 may be configured to implement orchestration logic that synthesizes per-user assessment outputs to generate unified simulation state updates, resolving conflicts between mutually exclusive user decisions and detecting interaction effects where one user's decision creates consequences for other users. The Al adaptation stage 208 may generate character-specific narrative content for each user session based on the unified simulation state and the information state associated with each user's character role, maintaining information asymmetry between users occupying different character roles. The Al adaptation stage 208 may configure interaction points within the simulation environment based on competency gaps or demonstrated mastery areas identified in the assessment output, emphasizing interaction points targeting competency dimensions identified as weak and surfacing more complex decision opportunities when mastery is demonstrated.

[0107] The circular arrangement of the four stages within the adaptive simulation cycle 200 may indicate that the stages operate as a continuous feedback loop, with each stage informing and enabling the next stage in the adaptive simulation process. The data flow within the system 200 may proceed through a defined sequence of operations wherein user inputs captured during the multi-user gameplay stage 202 are transmitted to the data assessment stage 204 for competency evaluation, assessment outputs are made available to the teacher coaching stage 206 for coaching intervention determination and to the Al adaptation stage 208 for narrative adaptation processing, and adapted content generated by the Al adaptation stage 208 is delivered back to the multi-user gameplay stage 202 forATTORNEY DOCKET NO.: 38562.0001P1presentation to users in subsequent simulation turns. The continuous cycling through these stages may enable progressive refinement of the simulation experience based on accumulated user performance data and coaching interactions.

[0108] In one example, the adaptive simulation cycle 200 may be implemented in a distributed configuration where the data assessment stage 204 and the Al adaptation stage 208 execute on cloud-based server infrastructure while the multi-user gameplay stage 202 renders on user devices connected via network communication. In another example, the adaptive simulation cycle 200 may be implemented in a centralized configuration where all stages execute on a single server system with shared memory¬ access. The teacher coaching stage 206 may be implemented as a separate Al module that operates independently from the assessment and narrative generation processes, with the architectural separation ensuring that coaching assistance does not compromise measurement validity of user competency assessment.

[0109] In another example, the adaptive simulation cycle 200 may be implemented with configurable stage parameters that enable educators to customize the behavior of each stage based on pedagogical objectives and learner characteristics. The data assessment stage 204 may support configuration of relevance thresholds that determine which competency dimensions receive scores for each user response. The teacher coaching stage 206 may support configuration of intervention conditions that determine when proactive coaching is initiated. The Al adaptation stage 208 may support configuration of adaptation parameters that determine how aggressively the simulation content adjusts based on assessment performance.

[0110] The components of the adaptive simulation cycle 200 may be implemented using specific technical structures and processes. The multi-user gameplay stage 202 may implement WebSocket connections for real-time bidirectional communication between user devices and server-side processing components. The data assessment stage 204 may employ structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing without error-prone text parsing. The teacher coaching stage 206 may implement intervention trigger logic that evaluates assessment output against configurable threshold conditions to determine when proactive coaching is appropriate. The Al adaptation stage 208 may implement conflict resolution algorithms that evaluate the relative strength ofATTORNEY DOCKET NO.: 38562.0001P1user arguments based on assessment scores and narrative logic to determine appropriate outcomes when user decisions are mutually exclusive.

[0111] The adaptive simulation cycle 200 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing platforms. Prior systems typically implement linear processing workflows where assessment, coaching, and adaptation occur as discrete sequential operations, resulting in delayed feedback and reduced responsiveness to user performance changes. Prior systems also lack mechanisms for maintaining architectural separation between coaching and assessment processes while still enabling coaching to respond to assessment data, compromising either measurement validity or coaching effectiveness. Additionally, prior systems that support multiple concurrent users do not implement continuous feedback loops that enable real-time adaptation based on collective user performance patterns. The arrangement and operation of the stages in the adaptive simulation cycle 200 address these technical problems through the circular feedback loop architecture that enables continuous processing, the read-only access constraint that ensures the teacher coaching stage 206 cannot modify assessment data while still receiving assessment output for coaching decisions, and the Al adaptation stage 208 that synthesizes multi-user assessment data into coherent narrative adaptations. This structural arrangement enables adaptive simulation with continuous feedback processing that was not achievable with prior system architectures implementing linear processing workflows.

[0112] FIG. 3 shows an example game play system architecture 300 illustrating the relationship between in-game and out-of-game processing components within the adaptive simulation system. The system architecture 300 may be configured to provide a hierarchical processing framework that separates longitudinal user data management from real-time simulation processing, enabling persistent learning profile maintenance while supporting dynamic in-session assessment and narrative adaptation. The system architecture 300 may represent the structural organization through which the adaptive simulation platform coordinates user profile data across multiple simulation sessions while managing the real-time interactions between assessment, narrative generation, and user modeling during active gameplay.

[0113] The system architecture 300 may comprise an outside of game region 302 that represents processing components operating independently from active simulation sessions. The outside of game region 302 may be configured to maintain persistent dataATTORNEY DOCKET NO.: 38562.0001P1structures and perform longitudinal analysis operations that span multiple simulation engagements. The outside of game region 302 may execute on server infrastructure that remains active regardless of whether users are currently engaged in simulation sessions, enabling continuous availability of user profile data and supporting asynchronous processing operations such as competency aggregation, curriculum adaptation, and crosssession insight derivation. The outside of game region 302 may implement data persistence mechanisms that ensure user learning data survives individual session terminations and system restarts.

[0114] The system architecture 300 may comprise a master model 306 positioned within the outside of game region 302. The master model 306 may be configured to serve as the primary coordination component for longitudinal user data management and cross-session learning profile maintenance. The master model 306 may implement aggregation algorithms that combine competency scores from multiple simulation sessions into unified competency profiles, applying relevance-weighted averaging to ensure that aggregated scores reflect contextually appropriate assessment data. The master model 306 may receive user playthrough data from the game play region 304 following simulation session completion, processing this data to update persistent learning records and derive insights about user learning patterns and competency development trajectories. The master model 306 may also provide user overall data back to the game play region 304 when users initiate new simulation sessions, enabling the simulation to adapt its initial configuration based on accumulated performance history'.

[0115] The system architecture 300 may comprise a user profile 308 positioned within the outside of game region 302. The user profile 308 may be configured to store persistent learning profile data including cumulative competency scores, historical performance records, derived learning insights, and user preference information. The user profile 308 may maintain per-dimension competency statistics including score counts, score sums, and average relevance weights across all contributing assessments. The user profile 308 may receive updates from the master model 306 following processing of playthrough data, with the master model 306 implementing selective aggregation that updates competency dimensions only when new assessment data provides contextually relevant evidence for those dimensions. The user profile 308 may be implemented using database systems that support efficient read and write operationsATTORNEY DOCKET NO.: 38562.0001P1for storing and retrieving competency data across potentially large numbers of simulation turns and sessions.

[0116] The system architecture 300 may comprise a game play region 304 that represents processing components operating during active simulation sessions. The game play region 304 may be configured to manage real-time interactions between users and the simulation environment, coordinating assessment, narrative generation, and user modeling operations within the temporal constraints of interactive gameplay. The game play region 304 may execute on server infrastructure or user devices depending on deployment configuration, with processing responsibilities distributed based on computational requirements and network latency considerations. The game play region 304 may maintain session-specific state information that tracks the current narrative position, user decisions made during the session, and assessment data accumulated across simulation turns.

[0117] The system architecture 300 may comprise an assessment model 316 positioned within the game play region 304. The assessment model 316 may be configured to evaluate user inputs during active simulation sessions and generate structured assessment outputs comprising competency scores across applicable competency dimensions. The assessment model 316 may implement artificial intelligence processing to analyze the quality, reasoning patterns, and evidence utilization demonstrated in user responses. The assessment model 316 may generate per-dimension relevance weights indicating the degree to which each user response provides meaningful evidence for particular competency dimensions, enabling selective scoring that updates only those dimensions for which the response provides contextually appropriate assessment data. The assessment model 316 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including user response data, simulation context, character role information, and predefined output schemas that constrain Al service responses to validated structured data formats.

[0118] The system architecture 300 may comprise a narrative model 312 positioned within the game play region 304. The narrative model 312 may be configured to receive user feedback data from the assessment model 316 and generate adaptive narrative content based on the assessment output. The narrative model 312 may produce customized dialogue, scene descriptions, decision options, and feedback content that reflects the user's character role, information state, and assessed competency level withinATTORNEY DOCKET NO.: 38562.0001P1the shared simulation session. The narrative model 312 may process narrative adjustment signals from the assessment model 316 to determine how the simulation narrative should progress based on the user's decisions and reasoning quality. The narrative model 312 may output content to the user model 314 for presentation to users through the simulation interface.

[0119] The system architecture 300 may comprise a user model 314 positioned within the game play region 304. The user model 314 may be configured to manage the userfacing aspects of the simulation session, receiving narrative content from the narrative model 312 and capturing user inputs and responses for transmission to the assessment model 316. The user model 314 may implement user interface rendering that presents narrative content, decision options, and feedback within the simulation environment. The user model 314 may capture diverse input modalities including typed text, spoken language, selections within the simulation, and other interaction types supported by the simulation design. The user model 314 may maintain a feedback loop connection to the master model 306, enabling user playthrough data to be transmitted to the outside of game region 302 for longitudinal processing and profile updating.

[0120] In one example, the system architecture 300 may be implemented in a distributed configuration where the outside of game region 302 executes on cloud-based server infrastructure while the game play region 304 executes on application servers or user devices connected via network communication. The master model 306 and user profile 308 may be implemented as persistent services that remain available across user sessions, while the assessment model 316, narrative model 312, and user model 314 may be instantiated for each active simulation session. In another example, the system architecture 300 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access, enabling reduced latency for data exchange between in-game and out-of-game processing components.

[0121] In another example, the system architecture 300 may be implemented with configurable data synchronization parameters that determine when and how playthrough data is transmitted from the game play region 304 to the outside of game region 302. Real-time synchronization may transmit assessment data to the master model 306 after each simulation turn, enabling immediate profile updates and supporting cross-session coaching continuity. Batch synchronization may accumulate playthrough data during theATTORNEY DOCKET NO.: 38562.0001P1session and transmit aggregated data upon session completion, reducing network overhead while still enabling longitudinal profile maintenance.

[0122] The components of the system architecture 300 may be implemented using specific technical structures and processes. The master model 306 may implement weighted aggregation algorithms that combine new scores with prior scores using relevance-weighted averaging, maintaining running statistics for each competency dimension including count, sum, and average relevance. The user profile 308 may implement efficient storage schemas that support rapid insertion of new assessment records and retrieval of historical performance data for adaptive system behavior. The assessment model 316 may employ structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing without error-prone text parsing. The narrative model 312 may implement template-based narrative generation with character-specific parameter substitution, or may utilize Al-powered narrative generation that produces contextually appropriate prose based on assessment outcomes and character perspectives. The user model 314 may implement WebSocket connections for real-time bidirectional communication between user devices and server-side processing components.

[0123] The system architecture 300 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing platforms. Prior systems typically conflate session-specific processing with longitudinal data management, resulting in either delayed profile updates that prevent cross-session adaptation or excessive processing overhead that degrades real-time simulation performance. Prior systems also lack mechanisms for separating the assessment model from the narrative model while maintaining coordinated data flow, resulting in tightly coupled architectures that prevent independent optimization of assessment accuracy and narrative quality. Additionally, prior systems that maintain user profiles do not implement selective aggregation based on per-response relevance determination, causing learning profiles to be diluted with assessment data points that do not reflect genuine demonstrations of competency in those dimensions. The arrangement and operation of the components in the system architecture 300 address these technical problems through the separation of the outside of game region 302 from the game play region 304, the bidirectional data flow that enables both real-time adaptation and longitudinal profile maintenance, the master model 306 that implements relevance-weighted aggregation forATTORNEY DOCKET NO.: 38562.0001P1selective profile updating, and the coordinated data flow between the assessment model 316, the narrative model 312, and the user model 314 that enables independent optimization while maintaining system coherence. This structural arrangement enables adaptive simulation with hierarchical processing that separates longitudinal learning management from real-time session processing, which was not achievable with prior system architectures that conflated these processing responsibilities.

[0124] FIG. 4 shows an example dynamic skills transcript generation system 400 for producing competency transcripts from user assessment data within a shared simulation session. The system 400 may be configured to generate both intermediary’ and overall competency transcripts that capture user performance across simulation interactions, enabling longitudinal tracking of learning progression and informing adaptive simulation behavior based on accumulated performance data.

[0125] The system 400 may receive user input 402 from a user during interaction with a shared simulation session. The user input 402 may comprise decisions, actions, textual responses, or selections submitted by a user through a user interface during simulation engagement. The user input 402 may be transmitted from a user device to assessment processing components via network communication protocols, with the input data structured according to predefined formats that facilitate downstream evaluation and transcript generation.

[0126] The system 400 may comprise an assessment model 404 that receives the user input 402 for evaluation against competency frameworks associated with the shared simulation session. The assessment model 404 may be configured to process user responses and generate competency evaluations across multiple dimensions including knowledge, theme, and skill categories. The assessment model 404 may implement artificial intelligence processing to analyze the quality, reasoning patterns, and evidence utilization demonstrated in user responses. The assessment model 404 may also receive input from a feedback 406 block that represents teacher or third-party7evaluation data, enabling human oversight and calibration of automated assessment processes. The integration of the feedback 406 with the assessment model 404 may enable refinement of assessment accuracy through incorporation of expert human judgment alongside AI-driven evaluation.

[0127] The system 400 may comprise an intermediary7skills transcript 408 that receives output from the assessment model 404. The intermediary7skills transcript 408 may beATTORNEY DOCKET NO.: 38562.0001P1configured to maintain per-section competency scores that capture user performance on individual simulation segments or decision points. The intermediary skills transcript 408 may comprise data structures for storing most recent prompt review information, all preceding prompts review data, competency -based scoring values, and competencybased evaluation and feedback content. The intermediary skills transcript 408 may provide granular performance tracking that enables identification of specific areas where users demonstrate strength or require additional support within the current simulation session.

[0128] The system 400 may comprise an overall skills transcript 410 that aggregates data from the intermediary skills transcript 408 over time. The overall skills transcript 410 may be configured to maintain a holistic competency depiction that synthesizes performance across multiple simulation interactions and sessions. The overall skills transcript 410 may comprise entire unit performance data that captures cumulative learning progression, along with points and feedback collection that aggregates assessment outcomes and instructional guidance provided throughout the user's engagement with the simulation platform. A data flow relationship between the intermediary skills transcript 408 and the overall skills transcript 410 may enable progressive aggregation of per-turn assessment data into longitudinal competency profiles.

[0129] The system 400 may comprise a display 412 that receives data from both the intermediary skills transcript 408 and the overall skills transcript 410 for presentation through a graphical user interface. The display 412 may be configured to render competency visualizations, progress indicators, and feedback content to users, instructors, or administrators. The display 412 may present real-time assessment feedback during simulation sessions as well as cumulative performance summaries accessible through dashboard interfaces.

[0130] The system 400 may comprise a master model 414 that receives data from the overall skills transcript 410 to inform future simulation playthroughs. The master model 414 may be configured to incorporate longitudinal performance data into adaptive simulation behavior, enabling the system to adjust difficulty levels, content sequencing, and available decision options based on demonstrated user competencies. The master model 414 may maintain persistent state information that enables continuity of adaptiveATTORNEY DOCKET NO.: 38562.0001P1behavior across multiple simulation sessions, ensuring that users receive appropriately calibrated experiences based on their accumulated performance history.

[0131] In one example, the system 400 may be implemented in a distributed configuration where the assessment model 404 executes on cloud-based server infrastructure while the display 412 renders on user devices connected via network communication. In another example, the system 400 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The intermediary' skills transcript 408 and the overall skills transcript 410 may be implemented using database systems that support efficient read and write operations for storing and retrieving competency data across potentially large numbers of simulation turns and sessions. The master model 414 may be implemented as a software module that queries the overall skills transcript 410 and applies algorithmic rules to determine adaptive simulation parameters.

[0132] The components of the system 400 may be implemented using specific technical structures and processes. The assessment model 404 may employ structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing and transcript generation. The intermediary’ skills transcript 408 may implement data structures optimized for rapid insertion of perturn assessment records and efficient retrieval of recent performance history. The overall skills transcript 410 may implement weighted aggregation algorithms that combine new scores with prior scores using relevance-weighted averaging, maintaining running statistics for each competency dimension. The master model 414 may implement adaptive parameter computation that evaluates transcript data against configurable thresholds to determine appropriate difficulty adjustments and content modifications.

[0133] The system 400 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing transcript generation platforms. Prior systems typically maintain only aggregate performance records yvithout preserving granular per-turn assessment data, preventing detailed analysis of learning progression patterns. Prior systems also lack mechanisms for integrating human feedback with automated assessment to calibrate Al-driven evaluation accuracy. Additionally, prior systems that generate competency transcripts do not provide bidirectional data flow to master models that inform adaptive simulation behavior, resulting in static experiences that cannot respond to accumulated userATTORNEY DOCKET NO.: 38562.0001P1performance. The arrangement and operation of the components in the system 400 address these technical problems through an intermediary skills transcript that preserves per-section competency data, feedback integration that enables human calibration of an assessment model, progressive aggregation from an intermediary skills transcript to an overall skills transcript, and data flow to a master model that enables adaptive simulation behavior based on longitudinal performance patterns. This structural arrangement enables dynamic skills transcript generation with adaptive simulation integration that was not achievable with prior system architectures.

[0134] FIG. 5 shows an example framework-agnostic assessment configuration system 500 for enabling selection and application of multiple competency frameworks to assessment processes within a shared simulation session. The system 500 may be configured to provide technical infrastructure through which one or more users (e.g., teachers, educators, etc.) may select, switch, and apply different competency standards to user assessment data, enabling the same simulation content to be evaluated against multiple assessment frameworks without requiring modification of the underlying simulation or re-collection of user responses.

[0135] The system 500 may comprise a plurality of competency framework blocks representing different assessment standards that may be applied to user performance evaluation. An AP framework 502 may represent Advanced Placement assessment standards with competency dimensions corresponding to historical periods, historical themes, and historical reasoning skills as defined by standardized educational curricula. A Common Core framework 504 may represent Common Core State Standards with competency dimensions aligned to cross-curricular literacy and analytical skill requirements. A third-party' competency framework 506 may represent competencybased assessment standards focused on deeper learning outcomes and transferable skills. A custom framework 508 may represent user-defined competency frameworks that educators may' configure to match specific instructional objectives, institutional requirements, or specialized assessment needs. The custom framework 508 may be configurable, enabling users to define custom competency dimensions, scoring criteria, and relevance weighting parameters tailored to particular educational contexts.

[0136] The system 500 may comprise a framework selector 510 that receives connections from each of the competency framework blocks. The framework selector 510 may be configured to maintain references to all available competency frameworksATTORNEY DOCKET NO.: 38562.0001P1and to designate one framework as the active assessment standard at any given time. The framework selector 510 may implement selection logic that enables switching between frameworks without requiring reprocessing of previously collected user data. The framework selector 510 may be configured for retroactive re-assessment, wherein the framework selector 510 may apply a newly selected framework to previously stored user responses to generate assessment data against the new competency standard.

[0137] A user 522 (e.g., teacher, educator, administrator, etc.) may provide user input to the framework selector 510. The user 522 may comprise an educator user who interacts with the system 500 through an administrative interface to configure assessment parameters. The user 522 may select which competency framework should be applied to current and future user assessments, may switch between frameworks during a course or unit, and may trigger retroactive re-assessment of stored user data against different competency standards. The connection between the user 522 and the framework selector 510 may enable real-time framework switching without interrupting ongoing simulation sessions.

[0138] The system 500 may comprise an assessment model 512 that receives the selected competency framework from the framework selector 510. The assessment model 512 may be configured to evaluate user responses against the competency dimensions defined in the currently active framework. The assessment model 512 may implement artificial intelligence processing to analyze user inputs and generate competency scores according to the scoring criteria and dimension definitions provided by the selected framework. The assessment model 512 may also receive input from user data points 514 that comprise user responses collected during simulation interactions. The user data points 514 may include multiple choice responses and free response inputs, representing different modalities of user interaction with the simulation.

[0139] The assessment model 512 may contain weighted scoring criteria 516 that define differential w eighting between response types. The w eighted scoring criteria 516 may indicate that free response inputs are assigned higher score weight than multiple choice responses, reflecting the greater depth of reasoning and competency evidence that freeform textual responses may provide compared to selection-based inputs. The weighted scoring criteria 516 may be configurable per framew ork, enabling different competency standards to apply different weighting schemes based on their assessment philosophies and competency dimension requirements.ATTORNEY DOCKET NO.: 38562.0001P1

[0140] The assessment model 512 may output to a competency evaluation 518 that represents the structured assessment output generated for each user response. The competency evaluation 518 may comprise per-dimension scores, relevance weights, justification text, and feedback content formatted according to the requirements of the active competency framework. The competency evaluation 518 may feed into a dynamic skills transcript 520 that maintains the cumulative competency record for each user. The dynamic skills transcript 520 may aggregate competency evaluation data over time, maintaining running statistics for each competency dimension and enabling longitudinal tracking of learning progression against the selected assessment standard.

[0141] The retroactive re-assessment capability of implemented by the framework selector 510 may enable the system 500 to apply a newly selected framework to previously stored user responses. For example, when the user 522 switches the active framew ork, the framework selector 510 may trigger re-evaluation of stored user data points 514 through the assessment model 512 using the new framework configuration. This retroactive capability may generate new competency evaluation 518 data against the alternative framework, enabling comparison of user performance across different assessment standards without requiring users to repeat simulation interactions.

[0142] In one example, the system 500 may be implemented in a distributed configuration where the framework selector 510 and the assessment model 512 execute on cloud-based server infrastructure while the teacher 522 interacts through a web-based administrative interface. In another example, the system 500 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The competency framework blocks may be implemented as data structures stored in a framework repository that the framework selector 510 queries to retrieve dimension definitions, scoring criteria, and weighting parameters. The dynamic skills transcript 520 may be implemented using database systems that support efficient storage and retrieval of competency data across multiple frameworks, enabling the system to maintain separate competency records for each frame wo rk applied to a given user's performance history.

[0143] In another example, the system 500 may be implemented with framework versioning capabilities that enable tracking of changes to competency framew ork definitions over time. The framework selector 510 may maintain version history for each framework, enabling the assessment model 512 to apply specific framework versions toATTORNEY DOCKET NO.: 38562.0001P1historical user data for consistent longitudinal analysis. The custom framework 508 may be implemented with a framework authoring interface that enables educators to define new competency dimensions, specify scoring rubrics, and configure relevance weighting parameters through a graphical configuration tool.

[0144] The components of the system 500 may be implemented using specific technical structures and processes. The framework selector 510 may implement framework switching logic that validates framework compatibility, manages state transitions between active frameworks, and coordinates retroactive re-assessment operations. The assessment model 512 may employ structured output schemas that constrain Al model responses to validated data formats conforming to the requirements of the active framework, enabling reliable downstream processing regardless of which framework is selected. The weighted scoring criteria 516 may implement configurable weighting algorithms that apply differential score contributions based on response type, response length, and evidence quality indicators. The dynamic skills transcript 520 may implement multi-framework storage schemas that maintain separate competency records for each framework applied, enabling cross-framework performance comparison and analysis.

[0145] The system 500 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing assessment configuration platforms. Prior systems typically implement assessment logic that is tightly coupled to a single competency framework, requiring substantial system modification or complete reimplementation to support alternative assessment standards. Prior systems also lack mechanisms for retroactive re-assessment, forcing educators to choose assessment framew orks before data collection and preventing comparison of user performance across different competency standards. Additionally, prior systems that support multiple response types typically apply uniform weighting regardless of the differential evidence value that different response modalities provide. The arrangement and operation of the components in the system 500 address these technical problems through a framework selector that enables dynamic framework switching with retroactive re-assessment capability, an assessment model that adapts evaluation logic based on the selected framew ork configuration, a w eighted scoring criteria that apply differential weighting based on response type, and a dynamic skills transcript that maintains multi-framework competency records. This structural arrangement enablesATTORNEY DOCKET NO.: 38562.0001P1framework-agnostic assessment configuration with retroactive re-assessment capability that was not achievable with prior system architectures.

[0146] FIG. 6 shows an example shared simulation session architecture 600 comprising structural arrangement and data flow relationships between multiple concurrent user sessions and the various artificial intelligence (Al) processing modules that enable multiuser adaptive simulation with dynamic competency assessment. The shared simulation session architecture 600 may be configured to provide the technical infrastructure through which a plurality of users may simultaneously participate in a shared simulation session (e.g., a shared narrative scenario) while receiving independently assessed, character-specific experiences. For example, the shared simulation session may comprise a current narrative position within the shared simulation session (e.g., shared simulation scenario), cumulative effects of all user decisions made during the shared simulation session, environmental conditions and parameters affecting available actions, non-player character states and dispositions, and resource or constraint conditions that affect the shared simulation simulation’s progression.

[0147] The shared simulation session architecture 600 may comprise a plurality of concurrent user sessions, such as user session A 611. user session B 612, and user session C 613. Each concurrent user session may comprise a distinct connection through which a user interacts with the shared simulation session. In an example, each concurrent user session may be associated with data indicative of a character role within the shared simulation session, wherein the character role may define character-specific knowledge, perspectives, and available actions that determine what information and decision options are accessible to that particular user. For example, the character role may define: the character’s identity, background, and motivations within the shared simulation session; the character’s information state, which specifies what information, kno ledge, and perspectives the character has access to; the character’s available actions and decision points, which may differ from those available to other characters; and the character’s relationships to other characters and to elements of the simulation scenario. For example, in a historical simulation of the American Revolution, one user may play a colonial merchant with knowledge of trade economics, another may play a Patriot lawyer with knowledge of legal and philosophical arguments, and a third may play a British loyalist with knowledge of Crown policies. Each character’s information state may define what that corresponding character knows, what that corresponding character canATTORNEY DOCKET NO.: 38562.0001P1see, and what decisions are available to that corresponding character. The concurrent user sessions may be implemented as client-side application instances executing within web browsers on user devices (e.g., electronic devices 102), with each instance maintaining a persistent connection to server-side processing components via network communication protocols.

[0148] The shared simulation session architecture 600 may comprise an Al assessment module 621 (e.g., one of the Al modules 154) that receives user input data from each of the plurality of concurrent user sessions. The Al assessment module 621 may be configured to independently evaluate user inputs during the interactive simulation and generate a structured assessment output for each concurrent user session. For example, the Al assessment module 621 may evaluate the users inputs against a competency framework associated with the shared simulation session to generate structured assessment output. For example, the Al assessment module 621 may evaluate each user's response in a context of that corresponding user’s character role and available information. As an example, a merchant’s economic argument may be assessed against knowledge of trade and commerce, while a lawyer's legal argument may be assessed against know ledge of political philosophy and rights. The Al assessment module 621 may evaluate different competency dimensions for different users on a same turn based on a nature of that corresponding user's character’s role and actions available to that corresponding user’s character. The structured assessment output may comprise competency evaluation data (e.g., competency scores) across a plurality7of competency dimensions, relevance weights indicating the degree to which each user response provides evidence relevant to particular competency dimensions, and user struggle indicators that identify when a user appears to be experiencing genuine confusion or misunderstanding. In an example, the structured assessment output may further comprise: identified reasoning patterns, including a logical structure of the user’s argument, evidence cited, and assumptions made: narrative adjustment signals indicating how the user’s decision should influence the shared simulation session’s progression; and interaction indicators specifying how the user’s decision relates to or conflicts with actions available to other characters. The Al assessment module 621 may process inputs from user session A 611, user session B 612, and user session C 613 through parallel assessment pathways, ensuring that each user’s response is evaluated in the context of their assigned character role and available information state. In an example, the AlATTORNEY DOCKET NO.: 38562.0001P1assessment module 621 may transmit assessment prompts to external Al model services (e.g.. at one or more of the servers 106) via application programming interfaces, with the assessment prompts, including the user response, simulation context, and a predefined output schema that constrains the Al assessment module 621 to generate a validated structured assessment output (e.g., validated structured data) conforming to specified typed fields.

[0149] The shared simulation session architecture 600 may comprise an Al orchestration module 623 (e.g., one of the Al modules 154) that receives the per-user structured assessment outputs from the Al assessment module 621. The Al orchestration module 623 may be configured to perform synthesis operations to generate unified simulation state updates based on the plurality of assessment outputs. The synthesis operations performed by the Al orchestration module 623 may comprise resolving conflicts between mutually exclusive user decisions, detecting interaction effects where one user’s decision creates consequences for other users, and evaluating collaborative reasoning quality that measures how users respond to each other’s reasoning within the shared simulation context. The Al orchestration module 623 may be configured to maintain bidirectional communication with persistent learning profiles 651, enabling the storage of competency scores aggregated across shared simulation sessions and the retrieval of historical performance data to inform assessment and narrative adaptation decisions.

[0150] As an example, for conflict resolution, when two or more users’ decisions are mutually exclusive or contradictory, the Al orchestration module 623 may resolve the conflict based on the simulation’s rules, the relative strength of the users’ arguments as determined by their assessment scores, and the narrative logic of the simulation scenario. For example, if a user associated with a merchant character role decides to support a trade embargo while a user associated with loyalist character role decides to expand trade with a country such as Britain, the Al orchestration module 623 determine the narrative consequences of this conflict and generate appropriate outcomes for each participant.

[0151] As an example, for interaction effect detection, the Al orchestration module 623 may identify instances where one user’s decision creates consequences, opportunities, or constraints for other users. For example, for a user associated with a lawyer character role, the user’s decision to publish a pamphlet advocating independence may create newATTORNEY DOCKET NO.: 38562.0001P1dialogue options for a merchant character role user and new threats for a loyalist character role user. These interaction effects are propagated through the shared simulation session and reflected in each user’s subsequent narrative content.

[0152] As an example, for collaborative competency evaluation, beyond evaluating each user’s individual competency, the Al orchestration module 623 may evaluate interuser reasoning quality. As an example, this includes: an ability to generate (e.g., synthesize) another participant’s argument into one’s own reasoning; an ability' to identify and respond constructively to logical weaknesses in another participant’s position; and an ability to formulate positions that account for multiple stakeholder perspectives. The collaborative competency data may be stored in each user’s persistent learning profile (e.g., persistent learning profile 651) alongside their individual competency data.

[0153] The shared simulation session architecture 600 may comprise a narrative generation module 641 that receives the unified simulation state update from the Al orchestration module 623. The narrative generation module 641 may be configured to generate character-specific narrative content for each user session based on the unified simulation state and the information state associated with each user’s character role. The narrative generation module 641 may produce customized dialogue, scene descriptions, decision options, and feedback content that reflects what each character would know and experience within the shared simulation session (e.g., shared simulation scenario), thereby maintaining information asymmetry between users occupying different character roles. The narrative generation module 641 may receive curriculum context data from a curriculum context data 661 store, enabling the generated narrative content to incorporate references to prior learning and cross-unit conceptual connections that support curriculum-coherent educational experiences.

[0154] The shared simulation session architecture 600 may include an Al coach module 622 that may be configured to operate as a separate process from the Al assessment module 621 and the narrative generation module 641. For example, the Al coach module 622 may only have read access to the assessment output generated by the Al assessment module 621, as indicated by the dashed connection between these components, as shown in FIG. 6. For example, the Al coach module 622 may not write to or modify the assessment output. This architectural separation ensures that coaching guidance does not contaminate the assessment's measurement of unassisted user reasoning. The Al coachATTORNEY DOCKET NO.: 38562.0001P1module 622 may be configured to monitor the structured assessment output across a plurality of user turns associated with the shared simulation session and to proactively initiate coaching interactions when the assessment output satisfies predefined intervention conditions, such as when a struggle indicator is set to true, when consecutive assessment scores fall below a defined threshold, or when identified misconceptions recur across multiple turns. In addition, the Al coach module 622 may also receive data from the persistent learning profiles 651 to inform coaching interactions based on longitudinal performance patterns, historical session insights, and documented user learning accommodations. For example, this enables the Al coach module 622 to tailor coaching tone, vocabulary complexity, and support level based on accumulated learner data and to reference prior competency development when providing guidance during current simulation sessions.

[0155] The shared simulation session architecture 600 may comprise an external memory service 631 that interfaces with the Al coach module 622 to maintain persistent conversational state for the coaching process across simulation sessions. The external memory service 631 may store conversation history' between a coaching assistant (e.g., the Al coach module 622, or a coaching Al module of the Al modules 154) and each user, derived insights about user learning patterns and communication preferences, and contextual information about user simulation states and assessment performance. In example, when a user returns to the platform for a new simulation session, the Al coach module 622 may retrieve stored context from the external memory' service 631, enabling the coaching process to reference prior interactions, recall user-specific learning patterns, and maintain continuity in its coaching relationship with the user. In an example, the external memory service 631 may be configured to operate with observed and unobserved participant classifications, where user messages contribute to insight derivation while coaching messages are stored but do not recursively influence the insight derivation process.

[0156] In an example, the shared simulation session architecture 600 may be configured to implement a feedback review queue where Al-generated assessment feedback is held for teacher user review before being released to student users. The shared simulation session architecture 600 may be configured to implement configurable release modes: an automatic mode where Al feedback is delivered directly to students without teacher intervention; a teacher-hold mode where all Al feedback requires teacher approvalATTORNEY DOCKET NO.: 38562.0001P1before release; and a threshold-based mode where feedback is automatically released when the Al's confidence in its assessment exceeds a defined threshold and held for teacher review when confidence falls below the threshold. Teachers may view, approve, modify, or reject Al-generated feedback within the review queue.

[0157] In an example, the shared simulation session architecture 600 may be configured to include a full assessment system operating alongside the simulation, comprising: essay assignments with Al rubric evaluation using category-specific scoring criteria; short response questions with criteria-based scoring; and multiple choice questions with question variant generation for retake integrity. Assignments may be configured as presimulation (to establish baseline knowledge), post-simulation (to evaluate learning), standalone, or end-of-unit assessments. The Al assessment module 621 may evaluate assignments using a competency framework and a relevance weighting system, producing unified competency profiles stored in the persistent learning profiles 652 that combine simulation performance with traditional assignment performance in a single persistent learning record.

[0158] In an example, the shared simulation session architecture 600 may be configured to generate Al-driven pedagogical analysis at multiple levels: individual student analysis identifying strengths, learning gaps, misconceptions, engagement patterns, and differentiation suggestions; class-level analysis identifying common misconceptions, skill mastery distributions, and suggested student groupings for differentiated instruction; and assignment-level analysis identifying performance patterns, error analysis, and assessment validity indicators. These insights may be derived from aggregated simulation and assignment assessment data stored in the persistent learning profiles 652 and may be updated as new data becomes available from the Al assessment module 621.

[0159] In an example, the shared simulation session architecture 600 may be configured to support a plurality of playable characters per simulation episode for each user, each with distinct roles, information access, gameplay descriptions, and background contexts. The same simulation scenario experienced through different character viewpoints may produce different narrative content generated by the narrative generation module 641, different available decisions, and different competency assessment opportunities evaluated by the Al assessment module 621. Cross-character performance may be tracked in the persistent learning profiles 652 to evaluate the user’s ability to synthesizeATTORNEY DOCKET NO.: 38562.0001P1perspectives across multiple playthroughs of the same scenario. Non-player characters (NPCs) within each episode may have configurable personalities and response patterns that interact differently with different playable characters.

[0160] In an example, the shared simulation session architecture 600 may be configured to provide a content management system with a curriculum editor role that enables rapid course creation through Al assistance. For example, features may include: Al autogeneration of character images from character descriptions; Al auto-generation of episode background images from scene descriptions; Al auto-generation of lesson plans and unit plans from course descriptions and source materials; and Al generation of assessment methods from course descriptions and subject area specifications.Hierarchical content organization (e.g., course, unit, episode, character) may be implemented with scoped inventory management and materials attachment at each level.

[0161] In an example, the shared simulation session architecture 600 may further implement a calibration methodology that enhances the accuracy of the Al assessment module 621 through a human-consensus-driven prompt optimization process. For example, one or more users (e.g., expert educators), via one or more user devices (e.g., one or more of the electronic devices 101) may provide one or more inputs for independently scoring simulation transcripts against a competency framework during norming workshops to establish human consensus scores as ground truth calibration targets (e.g., human ground truth scores). The consensus scores may then be provided to an automated prompt optimization process in which an Al coding agent (e.g., one of the Al modules 154) may iteratively refine the assessment prompt used by the Al assessment module 621 until the Al assessment module’s 621 scoring output converges with the ground truth calibration targets (e.g., human ground truth scores). The optimized prompt may be subsequently validated against held-out transcripts that were not used during the optimization process, and upon successful validation, the refined prompt may be deployed to the Al assessment module 621 for production use. As an example, this calibration approach provides a technical advantage in that the human consensus scores serve as a temporary' calibration input that permanently improves the assessment prompt rather than creating an ongoing dependency at inference time. The production prompt deployed to the Al assessment module 621 does not embed few-shot examples or scored transcripts. Instead, the prompt itself has been structurally refined through the optimization process to produce assessments that align with expert human judgment.ATTORNEY DOCKET NO.: 38562.0001P1This creates a repeatable calibration pipeline, wherein norming workshops generate human consensus scores, the automated optimizer iterates using those scores as convergence targets, validation confirms alignment on held-out data, and the improved prompt is deployed to enhance the assessment accuracy of the Al assessment module 621 across all user sessions (e.g., user session A 611, user session B 612, and user session A 613)

[0162] In an example, the Al assessment module 621 and the Al coach module 622 may receive curriculum context data comprising materials from a broader course surrounding the active simulation, including prior unit content, previously covered concepts and skills, course-wide learning objectives, and upcoming instructional targets. The Al assessment module 621 may utilize the curriculum context data to evaluate whether the user is building on knowledge from prior units, making cross-unit conceptual connections, and progressing along the intended learning arc, and not merely evaluating the response against the current simulation’s content in isolation. For example, if a student in a colonial economics simulation references concepts from a prior unit on mercantilism, the Al assessment module 621 may recognize this cross-unit connection and credit it in the competency evaluation. The Al coach module 622 similarly may utilize the curriculum context data to reference prior learning, connect current challenges to previously covered material, and provide guidance that is coherent with the broader instructional sequence. For example, the Al coach module 622 may generate a response such as “Remember when we explored the Stamp Act in the last unit? The same principles apply here.” The curriculum context data may be provided to the Al assessment module 621 and the Al coach module 622 alongside the simulation-specific context, enabling curriculum-coherent evaluation and guidance without requiring the Al services to maintain persistent knowledge of the full course structure.

[0163] In an example, the shared simulation session architecture 600 may be configured to dynamically adjust a broader curriculum surrounding the shared simulation session based on aggregated user performance data stored in the persistent learning profiles 652. For example, dynamic curriculum adaptation may operate at a longer time horizon, modifying an overall learning pathway across multiple simulation sessions and instructional activities. The adjustments may comprise: reordering the sequence of simulation modules to address identified competency gaps before progressing to more advanced content; modifying the emphasis, content, and recommended activities withinATTORNEY DOCKET NO.: 38562.0001P1surrounding instructional materials such as lesson plans, discussion prompts, and curated resource recommendations; inserting supplementary simulation episodes targeting specific competencies where class-wide or individual weaknesses are detected through aggregated assessment data from the Al assessment module 621; adjusting the scope and depth of upcoming units based on the pace of competency development observed across prior simulations; and generating customized pre-simulation and post-simulation assignments that target specific knowledge, theme, or skill dimensions identified as weak points in the persistent learning profiles 652. The dynamic curriculum adaptation may be applied at the individual learner level (e.g., personalizing the pathway for a single user session such as user session A 611, user session B 612, or user session C 613), at the class level (e.g.. adjusting the curriculum for an entire group based on common performance patterns identified through Al-powered pedagogical insights), or at the course level (e.g., modifying a default curriculum structure based on longitudinal data from multiple cohorts). Curriculum adjustments may be recommended to the teacher through a teacher insights interface In an example, adjustments may be applied automatically based on predefined performance thresholds.

[0164] In one example, the shared simulation session architecture 600 may be implemented in a distributed configuration where the Al assessment module 621, the Al orchestration module 623, the narrative generation module 641, and the Al coach module 622 execute on separate server instances that communicate via network protocols. In another example the Al assessment module 621, the Al orchestration module 623, the narrative generation module 641, and the Al coach module 622 may be implemented as software processes executing on a single server system with shared memory access (e.g., the computing device 101, the electronic devices 102, the servers 103, one or more combinations thereof, and the like). In an example, the external memory service 631 may be implemented as a cloud-based database service, a dedicated memory management server, or a local storage system depending on deployment requirements and privacy considerations. The user sessions may connect to the server-side components via WebSocket connections for real-time bidirectional communication, HTTP requests for discrete interactions, or other suitable network protocols.

[0165] The shared simulation session architecture 600 may alternatively be implemented with varying degrees of on-device processing. In one example, lightweight Al operations such as coaching response generation may be performed locally on userATTORNEY DOCKET NO.: 38562.0001P1devices (e.g., the electronic devices 102) while computationally intensive operations such as multi-user orchestration synthesis are performed on server infrastructure (e.g.. the servers 106). In another example, all Al processing may be performed on server infrastructure (e.g., the servers 106) with user devices (e.g., the electronic devices 102) serving primarily as display and input terminals. The shared simulation session architecture 600 may support hybrid configurations where processing responsibilities are dynamically allocated based on network conditions, device capabilities, and privacy requirements.

[0166] The components of the shared simulation session architecture 600 are implemented using specific technical structures and processes rather than generic computing abstractions. For example, the Al assessment module 621 may employ structured output schemas that constrain Al model responses to validated data formats, enabling reliable downstream processing without error-prone text parsing. The Al orchestration module 623 may implement conflict resolution algorithms that evaluate the relative strength of user arguments based on assessment scores and narrative logic to determine appropriate outcomes when user decisions are mutually exclusive. The narrative generation module 641 may maintain per-user information state records that define the boundary between character knowledge and full simulation state, using this asymmetry data structure to generate targeted narrative content. The Al coach module 622 may implement intervention trigger logic that evaluates assessment output against configurable threshold conditions to determine when proactive coaching is appropriate.

[0167] The shared simulation session architecture 600 provides measurable technical benefits including reduced assessment data dilution through selective scoring of only relevant competency dimensions, improved measurement validity7through architectural separation of coaching and assessment processes, and enhanced scalability through parallel independent assessment of multiple concurrent users. The architecture enables real-time processing of multiple user inputs within defined collection windows, synthesis of potentially conflicting decisions into coherent narrative outcomes, and maintenance of distinct information states for each user without cross-contamination.[0016S] The shared simulation session architecture 600 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing platforms. Prior systems uniformly apply all assessment dimensions to every user response regardless of contextual relevance, resulting in dilutedATTORNEY DOCKET NO.: 38562.0001P1competency profiles with reduced accuracy. Prior systems also lack mechanisms for coordinating multiple concurrent users in a shared adaptive environment while maintaining per-user assessment integrity and information asymmetry. Additionally, prior systems that provide coaching assistance during assessed activities cannot distinguish between unassisted user reasoning and coached performance, compromising measurement validity. The arrangement and operation of the components in the shared simulation session architecture 600 address these technical problems through the parallel independent assessment pathways of the Al assessment module 621, the synthesis and conflict resolution operations of the Al orchestration module 623, the informationasymmetry-aware content generation of the narrative generation module 641, and the architecturally separated coaching process of the Al coach module 622 with its read-only access to assessment data and persistent memory through the external memory service 631. This structural arrangement enables multi-user adaptive simulation with dynamic competency assessment that was not achievable with prior system architectures.

[0169] FIG. 7 shows a flowchart of an example process 700 for implementing shared simulation sessions with multiple concurrent users, such as by processing multi-user inputs through parallel assessment, synthesis, and per-user narrative generation within a multi-user adaptive simulation system (e.g.. shared simulation session). The process 700 may be implemented by one or more of the computing device 101, the electronic devices 102, the servers 106, one or more combinations thereof, and the like.

[0170] At step 710, user inputs may be received from a plurality of concurrent sessions. For example, user input data may be collected from each active user session participating in the shared simulation session. The user input data may be associated with, or correspond to, decisions, responses, or actions submitted by users via the users’ respective client interfaces. In an example, a collection window mechanism may be implemented that aggregates user inputs received within a defined time period before proceeding to assessment operations, enabling a plurality' of concurrent inputs to be processed as a coordinated batch. The collection window may be configured with variable duration based on simulation parameters, user activity' patterns, or network latency considerations. In an example, the user inputs may be received via WebSocket connections for real-time communication, HTTP requests for discrete interactions, or other suitable network protocols depending on deployment configuration.ATTORNEY DOCKET NO.: 38562.0001P1

[0171] At step 720, the user inputs may be assessed. For example, parallel assessment operations may be initiated that evaluate each user’s input independently. For example, as shown in FIG. 7, at step 721, User A user inputs may be assessed, at step 722, User B may be assessed, and at step 723, User C may be assessed. Each of steps 721, 722, and 723 may be implemented independently and concurrently, enabling the system to evaluate multiple user inputs without sequential bottlenecks. For example, each user’s response may be assessed in the context of the corresponding user’s assigned character role and available information state, without cross-contamination from other users’ user inputs or assessment results.

[0172] As an example, at step 721, User A assessment operations may be performed by evaluating User A’s input against the competency framework. User A’s character role, the information state associated with that character, and the specific decision context presented to User A may be consider, at the current shared simulation session turn, for example. A structured assessment output may be generated. The structured assessment output may comprise competency scores across applicable competency dimensions, relevance weights indicating the degree to which User A’s response provides evidence for each competency dimension, identified reasoning patterns including logical structure and evidence cited, and narrative adjustment signals indicating how User A’s decision should influence simulation progression. In an example, an assessment prompt may be output to an Al module (e g., one of the Al modules 154) via an application programming interface, with the prompt including User A’s response, simulation context, character role information, and a predefined output schema that constrains the Al module to generate validated structured data.

[0173] At step 722, User B assessment operations may be performed using the same assessment methodology applied at step 721, but configured for User B’s character role and information state. For example, different competency dimensions may be evaluated for User B based on the nature of User B's character role and the actions available to that character. For example, if User A plays a merchant character and User B plays a lawyer character in a historical simulation, economic knowledge competency dimensions may¬ be emphasized for User A while legal and philosophical knowledge competency dimensions may be emphasized for User B. A structured assessment output specific to User B’s response and character context may be generated.ATTORNEY DOCKET NO.: 38562.0001P1

[0174] At step 723, User C may be performed assessment operations following the same parallel assessment pattern as for User A and User B. For example, different competency dimensions may be evaluated for User C based on the nature of User C’s character role and the actions available to that character. Steps 721, 722, and 723 may be executed concurrently, enabling assessment operations to be completed for all users within a single processing cycle rather than requiring sequential assessment that would increase latency proportionally with user count.

[0175] At step 730, assessment outputs may be generated for each user. Each assessment output may comprise competency evaluation data including per-dimension scores and relevance weights, narrative adjustment signals specifying how each user’s decision should affect simulation state, and interaction indicators specifying how each user’s decision relates to or conflicts with actions available to other characters. In an example, validation operations may be implemented to verify a structural integrity of each assessment output against the predefined output schema before proceeding to synthesis.

[0176] At step 740, the assessment outputs from all users may be collected according to an orchestration process. The plurality of per-user assessment outputs may be aggregated and prepared for synthesis operations. For example, a complete set of assessment outputs may be received and relationships between user decisions may be identified, including potential conflicts, interaction effects, and collaborative reasoning patterns. In an example, queuing mechanisms may be implemented to ensure all expected assessment outputs are received before initiating synthesis, with timeout handling for cases where individual assessment operations experience delays.

[0177] At step 750, synthesis may be performed. For example, the plurality of independent assessment outputs may be transformed into a unified simulation state update. For example, conflicts between mutually exclusive user decisions may be resolved, interaction effects may be detected where one user’s decision creates consequences for other users, and collaborative reasoning quality may be evaluated that measures how users respond to each other’s reasoning within the shared simulation session.

[0178] As an example, for conflict resolution, when two or more users’ decisions are contradictory or cannot coexist within the simulation state, conflict may be resolved based on the simulation’s rules, the relative strength of the users’ arguments asATTORNEY DOCKET NO.: 38562.0001P1determined by their assessment scores, and the narrative logic of the simulation scenario. For example, if User A decides to support a trade embargo while User C decides to expand trade, the relative strength of each user's argument may be evaluated based on their assessment scores, the narrative logic of the simulation scenario, and predefined simulation rules to determine appropriate outcomes. For example, outcomes may be produced where one user’s decision prevails, where a compromise outcome emerges, or where the conflict itself becomes a narrative element with consequences for all participants.

[0179] As an example, for interaction effect detection, instances where one user’s decision creates opportunities, constraints, or consequences for other users may be identified. For example. User B’s decision to publish a legal pamphlet may create new dialogue options for User A and new threats for User C. As an example, these interaction effects may be propagated through the shared simulation session, updating the available actions and information states for affected users.

[0180] As an example, for collaborative competency evaluation, inter-user reasoning quality that emerges from the collective decision-making process may be evaluated. This evaluation measures each user’s ability to synthesize another participant’s argument into their own reasoning, identify and respond constructively to logical weaknesses in another participant’s position, and formulate positions that account for multiple stakeholder perspectives. As an example, the collaborative competency data may represent a form of critical thinking measurement.

[0181] At step 760, a shared simulation session update may be generated based on the synthesized assessment outputs. A unified simulation state update may be generated that reflects the collective effects of all user decisions as processed at step 750. The simulation state update may comprise an updated narrative position within the simulation scenario, cumulative effects of user decisions on environmental conditions and nonplayer character states, modified resource or constraint conditions, and updated relationship states between characters. In an example, updates to persistent learning profiles may be triggered, storing the competency scores and collaborative assessment data generated during the current turn for each user.

[0182] At step 770, character-specific narrative content may be generated for each concurrent user session. The unified simulation state update may be received and customized narrative output may be generated for each user based on that correspondingATTORNEY DOCKET NO.: 38562.0001P1user’s character role and information state. The narrative content may reflects what each character would know and experience within the shared simulation session, maintaining information asymmetry between users occupying different character roles. For example, for User A playing a merchant character, trade-relevant consequences and economic dialogue options may be generated. For example, for User B playing a law er character, legal and philosophical consequences and argumentation opportunities may be generated. For example, for User C playing a loyalist character. Crown-loyalty consequences and political pressures may be generated.

[0183] In an example, the narrative content may comprise character-specific dialogue and scene descriptions that reflect each character’s perspective, character- appropriate decision options for the next simulation turn, feedback integrated into the narrative as incharacter reactions from non-player characters that reflect the quality of each user’s reasoning, and character-specific environmental descriptions that reflect each user’s information state. In an example, curriculum context data may be incorporated to include references to prior learning and cross-unit conceptual connections within the generated narrative.

[0184] At step 780, the narrative may be output for each user via that corresponding user’s respective interface. The character-specific narrative content may be output to each concurrent user session via a communication channel. The narrative output may be delivered to each user’s client interface for display (e.g., an electronic device 102) within the shared simulation session. In an example, environment configuration operations may be coordinated to update interaction points and visual elements within browser-based immersive displays based on the assessment output and narrative state.

[0185] As an example, the process 700 may be implemented with variable numbers of parallel assessment paths corresponding to the number of active users in a simulation session. While FIG. 7 shows three parallel paths for Users A, B, and C, the process 700 may scale to accommodate two users, four users, larger groups, and the like depending on simulation configuration and system capacity. The parallel assessment architecture enables linear scaling of assessment throughput with available processing resources.

[0186] In one example, the process 700 may be implemented in a distributed configuration where steps 721, 722, and 723 execute on separate server instances or processing cores, with steps 740 through 770 executing on a coordinating server that aggregates results and performs synthesis. In another example, all steps of the processATTORNEY DOCKET NO.: 38562.0001P1700 may execute on a single server system with multi-threaded processing for the parallel assessment operations. The process 700 may also be implemented in hybrid configurations where assessment operations are distributed across cloud computing resources while synthesis and narrative generation execute on dedicated application servers.

[0187] In another example, the process 700 may alternatively be implemented with asynchronous processing patterns where assessment operations proceed independently without waiting for all users to submit inputs simultaneously. For example, at step 710, rolling collection windows may be implemented that trigger synthesis operations when sufficient inputs are received or when timeout thresholds are reached. Asynchronous processing may enable users to be accommodated with varying response times while maintaining simulation progression.

[0188] The process 700 represents a technical improvement over prior adaptive learning workflows that addresses specific technical problems inherent in existing systems. Prior systems that support multiple users typically process inputs sequentially, resulting in latency that scales linearly with user count and prevents real-time collaborative experiences. Prior systems also lack mechanisms for synthesizing multiple user decisions into coherent narrative outcomes while maintaining per-user assessment integrity and information asymmetry’. The arrangement and sequencing of operations in the process 700 address these technical problems through parallel independent assessment, synthesis and conflict resolution, and information-asymmetry-aware narrative generation. This operational sequence enables multi-user adaptive simulation with dynamic competency assessment and real-time narrative adaptation that was not achievable with prior system workflows.

[0189] FIG. 8 shows an example management system 800 for implementing shared simulation sessions with multiple concurrent users. As shown in FIG. 8, the management system 800 may comprise a shared simulation session 810 that may comprise simulation state information (e g., information states for each concurrent user session), including all narrative events, environmental conditions, non-player character states, resource conditions, and the cumulative effects of all user decisions made during the shared simulation session 810. For example, the information state for each concurrent user session may comprise: base character knowledge defined by a character role; information acquired through a user's own decisions and exploration during the shared simulationATTORNEY DOCKET NO.: 38562.0001P1session 810; information shared by other characters through in-simulation interactions; and information explicitly withheld based on the character's role and the shared simulation session’s 810 design. The shared simulation session 810 may serve as an authoritative data structure from which character-specific views are derived, ensuring that all users participate in a coherent shared experience even when their individual perspectives differ based on character role assignments.

[0190] The shared simulation session 810 may comprise three overlapping knowledge regions representing character-specific information subsets. User A knowledge 812 may comprise an information state accessible to a user assigned to a first character role. The User A knowledge 812 may define what information, perspectives, and contextual details are available to User A based on the user’s character’s identity, background, relationships, and actions taken during the shared simulation session 810. For example, in a historical simulation where User A plays a colonial merchant, the User A knowledge 812 may comprise detailed information about trade economics, commodity prices, shipping routes, and commercial relationships while excluding information about military strategies or legal proceedings that the merchant character would not have access to.

[0191] User B knowledge 816 may comprise an information state accessible to a user assigned to a second character role. The User B knowledge 816 may define a distinct subset of the shared simulation session 810 that corresponds to what the second character would know within the shared simulation session 810. In the historical simulation example, if User B plays a Patriot lawyer, the User B knowledge 816 may comprise detailed information about legal arguments, philosophical treatises, political alliances, and courtroom proceedings while excluding detailed trade economics or Crown military7intelligence.

[0192] User C knowledge 814 comprise an information state accessible to a user assigned to a third character role. The User C knowledge 814 may define another distinct subset of the shared simulation session 810 corresponding to the third character's perspective and available information. Continuing the historical simulation example, if User C plays a British loyalist, the User C knowledge 814 may comprise information about Crown policies, military dispositions, loyalist networks, and official proclamations while excluding details of Patriot organizing activities or colonial merchant trade secrets.ATTORNEY DOCKET NO.: 38562.0001P1

[0193] The three knowledge regions 812, 814, and 816 may overlap comprising shared knowledge 818. The shared knowledge 818 may comprise information that is accessible to all characters within the simulation scenario, such as publicly known events, common environmental conditions, and information that has been explicitly shared between characters through in-simulation interactions. The shared knowledge 818 may provide a common ground that enables meaningful interaction between users occupying different character roles, ensuring that users can reference shared events and conditions even when each users’ broader information states differ substantially.

[0194] A narrative generation process 804 may receive data from the shared simulation session 810. The narrative generation process 804 may be configured to generate character-specific narrative content for each user based on a unified simulation state and the information state associated with each user’s character role. The narrative generation process 804 may be configured to transform the shared simulation session 810 into appropriately constrained narrative outputs that maintain information asymmetry between users. As an example, the narrative output content for each user may comprise: character-specific dialogue and scene descriptions; character-appropriate decision options for a next turn; feedback integrated into the narrative (e.g., in-character reactions from non-player characters that reflect the quality of the user’s reasoning); and character-specific environmental descriptions that reflect the user’s information state.

[0195] The narrative generation process 804 may access the shared simulation session 810 and apply information state filters corresponding to each user’s character role. When generating narrative content for User A, the narrative generation process 804 may constrain its output to information within the User A knowledge 812 region, presenting events, dialogue, and decision options that the merchant character would have access to while omitting or obscuring information outside that region. The narrative generation process 804 may perform analogous filtering operations for User B and User C, ensuring that each user receives narrative content consistent with that corresponding user’s character’s perspective and knowledge boundaries.

[0196] The management system 800 may comprise one or more user interfaces (e.g.. User A interface 802, User B interface 808, User C interface 806, etc.). The User A interface 802 may receive narrative content filtered through the User A knowledge 812 information state. The User B interface 808 may receive narrative content filteredATTORNEY DOCKET NO.: 38562.0001P1through the User B knowledge 816 information state. The User C interface 806 may receive narrative content filtered through the User C knowledge 814 information state.

[0197] The user interfaces 802, 806, and 808 may be configured to display the character-specific narrative content to users through client-side application instances. In an example, the user interfaces 802, 806, and 808 may be implemented as web browserbased applications that render narrative content, dialogue options, and environmental descriptions within immersive display environments. The user interfaces 802, 806, and 808 may also receive user input data corresponding to decisions made by each user within the simulation, transmitting this input data back to the system for assessment and processing (e.g., as described with reference to the process 700 of FIG. 7).

[0198] In an example, management system 800 may be configured to further support assessment of perspective-taking and empathy competencies. For example, when a user demonstrates awareness of information or perspectives that their character should not have access to, the assessment process may flag this as a potential issue. Conversely, when a user demonstrates an ability to reason effectively despite having limited information, this provides positive evidence of critical thinking under uncertainty.

[0199] In one example, the system 800 may be implemented in a distributed configuration where the shared simulation session 810 is maintained on a central server system while the user interfaces 802, 806, and 808 execute on separate user devices connected via network communication. The narrative generation process 804 may execute on the central server system, generating character-specific content that is transmitted to user devices for display. In another example, portions of the narrative generation process 804 may execute on user devices (e.g., electronic devices 102), with the shared simulation session 810 providing filtered data streams to each device based on character role assignments.

[0200] In another example, the system 800 may be implemented with dynamic information state boundaries that change during simulation progression. As users make decisions and interact with the simulation, their character’s information state may expand to include newly acquired knowledge or may be constrained based on narrative events. The narrative generation process 804 may track these dynamic boundaries and adjust its filtering operations accordingly, ensuring that narrative content remains consistent with each character’s evolving information state throughout the simulation session.ATTORNEY DOCKET NO.: 38562.0001P1

[0201] As an example, the system 800 may alternatively be implemented with varying numbers of character roles and corresponding knowledge regions. While FIG. 8 shows three users with three distinct knowledge regions, the system 800 may scale to accommodate two users, four users, larger groups, and the like with corresponding adjustments to the knowledge region structure within the shared simulation session 810. The overlap patterns between knowledge regions may also vary based on simulation design, with some simulations featuring substantial shared knowledge and others featuring minimal overlap to create more pronounced information asymmetry.

[0202] The system 800 represents a technical improvement over prior multi-user simulation systems that addresses specific technical problems inherent in existing platforms. Prior systems that support multiple users in shared environments typically provide uniform information access to all participants, preventing the creation of character-specific perspectives that enable meaningful role-playing and perspectivetaking assessment. Prior systems also lack mechanisms for maintaining coherent shared state while delivering asymmetric information views, resulting in either fully shared experiences without character differentiation or fully separate experiences without meaningful interaction. The arrangement and operation of the components in the system 800 address these technical problems through the centralized shared simulation session 810 that maintains authoritative state, character-specific knowledge regions that define information boundaries, and a narrative generation process that produces filtered content for different user interfaces. This structural arrangement enables multi-user adaptive simulation with information asymmetry management that was not achievable ith prior system architectures.

[0203] FIG. 9 shows an example assessment framework data flow 900 for evaluating user responses and generating competency assessments within a shared simulation session. The assessment framework data flow 900 may comprise a competency framework configuration 910 that represents input data received from a user during a shared simulation session. The competency framework configuration 910 may comprise one or more decisions, actions, or textual responses provided by a user through the user’s user interface. The competency framework configuration 910 may comprise (e.g., organized into) assessment dimensions that correspond to different categories of competency dimensions applicable to the user’s input. The competency dimensions may comprise a know ledge dimension 912, a theme dimension 914, and a skill dimensionATTORNEY DOCKET NO.: 38562.0001P1916. The knowledge dimension 912 may comprise data indicative of factual content domains that may be assessed based on the competency framework configuration 910. such as historical facts, scientific principles, or domain-specific information relevant to the shared simulation session. The theme dimension 914 may comprise data indicative of cross-cutting conceptual understandings that span multiple knowledge areas, such as cause and effect relationships, continuity and change over time, or systems thinking. The skill dimension 916 may comprise data indicative of cognitive and analytical competencies, such as argumentation, evidence evaluation, perspective synthesis, or critical reasoning abilities. In an example, each competency dimension may comprise a unique identifier, a descriptive title, a detailed description specifying what the dimension measures and how the dimension should be evaluated, and scoring criteria defining a scale for evaluation. In an example, The competency dimensions may be configurable per shared simulation session and per assessment method, allowing the same shared simulation session to be assessed against different competency standards by swapping an active dimension configuration with another dimension configuration.

[0204] The assessment framework data flow 900 may comprise an artificial (Al) assessment process 920 (e.g., one of the Al modules 154) that receives the competency framework configuration 910 with its associated dimensional organization. The Al assessment process 920 may be configured to evaluate the user input based on the different competency dimensions defined in the competency framework configuration 910 for the simulation. The Al assessment process 920 may also receive a user response 930 and curriculum context data 940. The user response 930 may represent additional contextual information about the user’s response, including a simulation state of the shared simulation session at the time of response, a character role occupied by the user, and a specific decision point or prompt to which the user was responding. The curriculum context data 940 may provide information from a broader course surrounding the shared simulation session, including prior unit content, previously covered concepts and skills, course-wide learning objectives, and upcoming instructional targets. The curriculum context data 940 may enable the Al assessment process 920 to evaluate crossunit learning progression, determining whether the user is building on knowledge from prior units and making conceptual connections across a curriculum associated with the curriculum context data 940 rather than evaluating the response in isolation against only the current simulation content.ATTORNEY DOCKET NO.: 38562.0001P1

[0205] The Al assessment process 920 may process the competency framework configuration 910, the user response 930, and the curriculum context data 940 to generate a structured assessment output. The Al assessment process 920 may output assessment prompts to Al module services (e.g., stored on one or more of the servers 106) via application programming interfaces, with the prompts including the competency framework configuration dimensions, user response data, simulation context, character role information, curriculum context, and a predefined output schema that constrains the Al service to generate validated structured data (e g., assessment outputs) conforming to specified typed fields. As an example, the predefined output schema ensures that the Al service produces consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations.

[0206] The assessment framework data flow 900 may comprise a structured assessment output 950 that represents the structured assessment output generated by the Al assessment process 920. The structured assessment output 950 may produce detailed assessment data conforming to the predefined output schema. The structured assessment output may comprise an overall score represented as a numerical value indicating an aggregate quality7of the user response. The structured assessment output may further comprise knowledge scores comprising arrays with title, score, applicability indicators, and justification fields for each knowledge dimension in the competency framework. Similarly, the structured assessment output may comprise theme scores and skill scores with corresponding arrays of dimensional assessments. The structured assessment output 950 may generate relevance values ranging from 0 to 1 for knowledge relevance, theme relevance, and skill relevance, indicating the degree to which the user response provides meaningful evidence for each category of competency dimensions. The structured assessment output 950 may output commentary as a string providing personalized feedback to the user, a Boolean indicator for whether the user is struggling, and a struggle context string describing the nature of any identified difficulties when the struggle indicator is true.

[0207] The assessment framework data flow 900 may comprise a relevance filter 960 that receives the structured assessment output from the structured assessment output 950.The relevance filter 960 may apply a configurable threshold to the relevance weights generated by the structured assessment output 950, separating relevant competency dimensions from non-relevant dimensions for a specific user response being evaluated.ATTORNEY DOCKET NO.: 38562.0001P1Competency dimensions having relevance weights exceeding the threshold may pass through the relevance filter 960 and receive scores for a current turn. Competency dimensions falling below the threshold may be excluded from scoring for that particular turn, ensuring that the user’s persistent learning profile is not diluted by assessments that lack contextual relevance to the specific response provided.

[0208] The assessment framework data flow 900 may comprise a persistent learning profile 970 that receives the filtered assessment data from the relevance filter 960. The persistent learning profile 970 may aggregate the generated competency scores with prior scores for each dimension, maintaining cumulative competency data across simulation sessions. The persistent learning profile 970 may implement selective aggregation wherein prior scores for each dimension are aggregated only from turns where that respective dimension was determined to be relevant by the relevance filter 960. For each competency dimension, the persistent learning profile 970 may maintain the cumulative score derived from turns where that dimension was determined to be relevant, the number of turns on which the dimension was assessed, and the average relevance weight across assessed turns. This selective aggregation may produce persistent learning profiles that reflect what the user actually demonstrated in contextually appropriate situations, providing higher-fidelity learning data compared to profiles built from uniform scoring across all dimensions regardless of relevance.

[0209] The assessment framework data flow 900 may comprise a dashboard 980 that receives data from the persistent learning profile 970. The dashboard 980 may display the aggregated competency data to one or more users, such as instructors, administrators, or other authorized users, for example for review and analysis of user performance across simulation sessions. For example, the dashboard 980 may display visualizations of competency progression over time, comparisons between individual and class-level performance, identification of common misconceptions or skill gaps, and recommendations for differentiated instruction based on the competency data maintained in the persistent learning profile 970.

[0210] In an example, the assessment framework data flow 900 may support a plurality of assessment methods corresponding to different competency frameworks. For example, a historical simulation may be assessed using an AP History framework with knowledge dimensions corresponding to historical periods, theme dimensions corresponding to historical themes, and skill dimensions corresponding to historical reasoning skills. TheATTORNEY DOCKET NO.: 38562.0001P1same shared simulation session may alternatively be assessed using a competency-based framework, with different dimension configurations. As an example, teachers or administrators may select or switch an active assessment method without modifying the shared simulation session content, and can apply a new assessment method retroactively to existing user performance data.

[0211] In one example, the assessment framework data flow 900 may be implemented in a distributed configuration where the Al assessment process 920 executes on cloudbased server infrastructure while the relevance filter 960 and persistent learning profile 970 are maintained on application servers closer to the user interface layer. In another example, all components of the adaptive environment generation system 900 may execute on a single server system with shared memory access. The dashboard 980 may be implemented as a web-based application accessible through standard browsers, enabling instructors to review competency data from any network-connected device.

[0212] In another example, the assessment framework data flow 900 may be implemented with configurable relevance thresholds that vary based on simulation type, assessment framework, or educator preferences. A higher threshold produces more selective scoring with fewer dimensions assessed per turn, while a lower threshold produces broader scoring with more dimensions assessed. The threshold configuration enables educators to balance assessment granularity against the desire for comprehensive competency coverage.

[0213] In another example, the assessment framework data flow 900 may support multiple assessment frameworks that can be applied to the same simulation content. The knowledge dimension 912, theme dimension 914, and skill dimension 916 may be configured differently for different assessment methods, enabling the same user response to be evaluated against different competency standards by swapping the active framework configuration. The persistent learning profile 970 may maintain separate competency records for each competency framework, or may aggregate scores across frameworks using mapping rules that relate dimensions across different frameworks.

[0214] The assessment framework data flow 900 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing assessment platforms. Prior systems uniformly apply all assessment dimensions to every user response regardless of whether a given response actually provides meaningful evidence for a particular competency, resulting in dilutedATTORNEY DOCKET NO.: 38562.0001P1competency profiles with reduced accuracy and actionability . Prior systems also evaluate user responses in isolation against only the current activity content, preventing assessment of cross-unit learning progression and curriculum-coherent skill development. The arrangement and operation of the components in the adaptive environment generation system 900 address these technical problems through the dimensional organization of a user response into a knowledge dimension, a theme dimension, and a skill dimension, the integration of curriculum context data into an Al assessment process, the generation of per-dimension relevance weights by the Al assessment process, threshold-based filtering by a relevance filter, and the selective aggregation by the persistent learning profile. This structural arrangement enables dynamic competency assessment with per-turn relevance determination and curriculum-aware evaluation that was not achievable with prior system architectures.

[0215] FIG. 10 shows an example process 1000 for determining competency relevance based on evaluating user competency within a shared simulation session using per-turn relevance determination and selective scoring. A user response 1002 may be received that comprises input data received from a user during a shared simulation session. The user response 1002 may serve as a primary' input to an assessment processing pipeline. For example, the user response 1002 may comprise decisions, actions, or textual responses submitted by a user through the user’s user interface during shared simulation session interaction. The user response 1002 may include free-form text responses, selection of predefined options, or combinations thereof depending on the shared simulation session design and a decision point presented to the user via the user interface. The user response 1002 may be transmitted from the user’s client device to the assessment processing components via network communication protocols, with the response data structured according to a predefined input format that facilitates downstream processing.

[0216] A competency framework 1004 may be received that comprises a plurality of scored competency dimensions organized into categorical groupings. The competency framework 1004 may define assessment criteria against which user responses are evaluated. The competency framework 1004 may comprise knowledge dimensions representing factual content domains, theme dimensions representing cross-cutting conceptual understandings, and skill dimensions representing cognitive and analytical competencies. Each competency dimension within the competency framework 1004 mayATTORNEY DOCKET NO.: 38562.0001P1comprise a unique identifier, a descriptive title, a detailed description specifying what the dimension measures and how it should be evaluated, and scoring criteria defining the scale for evaluation. The competency framework 1004 may be configurable per simulation and per assessment method, enabling the same shared simulation session content to be assessed against different competency standards by swapping the active framework configuration. In an example, the competency framework 1004 may be generated by artificial (Al) services (e.g., stored on one or more of the servers 106) based on course descriptions, subject area specifications, and grade level parameters, enabling rapid creation of assessment frameworks without manual authoring of individual competency dimensions. As an example, for younger learners (e.g., K-5), more generous scoring may be applied, wherein the process 600 may focus on participation and basic understanding and use age-appropriate feedback language. As an example, for advanced learners (e.g., university, workforce, etc.), the process 600 may expect sophisticated analysis, professional vocabulary7, and complex reasoning. As an example, the grade-level calibration may affect both the scoring rubric and the relevance determination, ensuring that the same simulation content produces age-appropriate assessment while maintaining consistent competency dimension coverage.

[0217] Curriculum context data 1006 may be received that provides information from the broader course surrounding the active simulation. The curriculum context data 1006 may comprise prior unit content, previously covered concepts and skills, course-wdde learning objectives, and upcoming instructional targets. The curriculum context data 1006 may enable the assessment process to evaluate cross-unit learning progression, determining whether the user is building on knowledge from prior units and making conceptual connections across the curriculum rather than evaluating the response in isolation against only the current simulation content. In an example, the curriculum context data 1006 may be retrieved from a curriculum data store that maintains hierarchical content organization including course, unit, episode, and character level information with associated learning materials and objectives at each level.

[0218] An Al assessment module 1008 (e.g., one of the Al modules 154) may receive the user response 1002. the competency framework 1004, and the curriculum context data 1006 as inputs. The Al assessment module 1008 may be configured to evaluate the user response 1002 against the competency dimensions defined in the competency framew ork 1004 while considering a broader curriculum context provided by theATTORNEY DOCKET NO.: 38562.0001P1curriculum context data 1006. The Al assessment module 1008 may output assessment prompts to external Al model services via application programming interfaces, with the prompts including the user response data, simulation context, character role information, curriculum context, and a predefined output schema that constrains the Al service to generate validated structured data conforming to specified typed fields. For example, a predefined output schema may define fields for: an overall numerical score; perdimension scores for each knowledge, theme, and skill dimension; category-level relevance weights indicating the overall applicability of knowledge assessment, theme assessment, and skill assessment to the current response; per-dimension applicability flags indicating whether individual dimensions were relevant; justification text explaining the scoring rationale; and a commentary field providing personalized feedback to the user. The predefined output schema ensures that the Al service produces consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations. The Al assessment module 1008 may generate structured assessment output comprising per-dimension scores, per-dimension relevance weights indicating the degree to which the user response provides evidence relevant to each competency dimension, category-level relevance weights for knowledge, theme, and skill categories, commentary' providing personalized feedback, and struggle indicators identifying when a user appears to be experiencing genuine confusion or misunderstanding.

[0219] A filter 1010 may receive the structured assessment output from the Al assessment module 1008. The filter 1010 may implements threshold-based filtering to determine which competency dimensions should receive scores based on their relevance weights. For example, the filter 1010 may apply a configurable relevance threshold to the per-dimension relevance weights generated by the Al assessment module 1008, separating relevant competency dimensions from non-relevant dimensions for the specific user response being evaluated. Competency dimensions having relevance weights exceeding the threshold may pass through the filter 1010 and are included in the scoring for the current turn. Competency dimensions falling below the threshold may be excluded from scoring for that particular turn, ensuring that the user’s persistent learning profile is not diluted by assessments that lack contextual relevance to the specific response provided. The filter 1010 may implement comparison algorithms that evaluate numerical relevance weights against the configurable threshold value to produce binaryATTORNEY DOCKET NO.: 38562.0001P1inclusion decisions for each competency dimension, outputting a filtered set of dimension scores that represent only the contextually relevant assessments.

[0220] The filter 1010 may output selective scores 1012 comprising competency scores generated only for those dimensions determined to be relevant to the user response 1002.The selective scores 1012 may include per-dimension score values for each competency dimension that passed the relevance threshold, along with the associated relevance weights and justification text explaining the scoring rationale. Scores for dimensions that were determined to be non-relevant may be excluded from the selective scores 1012, thereby preventing dilution of the persistent learning profile with assessments that do not provide meaningful evidence of user competency in those particular dimensions. In an example, the selective scores 1012 may be structured as data records that can be efficiently stored and aggregated with prior assessment data.

[0221] A weighted overall score computation 1014 may receive input from the selective scores 1012 and compute an overall score for the user response 1002 as a weighted average of category -level scores. The weighted overall score computation 1014 may also receive input from the competency framework 1004 to obtain category -level relevance weights for normalizing the weighted average calculation. The weighted overall score computation 1014 may compute the overall score by weighting the knowledge category score, theme category score, and skill category score according to their respective category -level relevance values, normalized to sum to 1.0. For example, if the knowledge relevance weight is 0.7, the theme relevance weight is 0.8, and the skill relevance weight is 0.9, the normalized weights would be approximately 0.29, 0.33, and 0.38 respectively, and the overall score would be computed as the sum of each category score multiplied by its normalized weight. As an example, a student user’s decision to negotiate a trade agreement between colonial factions may produce the following relevance weights: "Persuasive Communication”: 0.9 (highly relevant because the response requires and demonstrates persuasive reasoning); “Diverse Perspectives”: 0.8 (relevant because the negotiation involves considering multiple viewpoints); “Historical Knowledge: Colonial Economics”: 0.7 (relevant because the negotiation draws on economic knowledge); “Historical Knowledge: Military Strategy”: 0.1 (not relevant because the response involves no military reasoning); “Synthesis”: 0.3 (marginally relevant). Only dimensions with relevance weights exceeding a configurable threshold (e.g., 0.5) receive scores on that turn. This weighted computation ensures that the overallATTORNEY DOCKET NO.: 38562.0001P1score for a turn appropriately reflects the relative importance of each assessment category to the specific user response rather than treating all categories as equally relevant regardless of the response content.

[0222] A persistent learning profile 1016 may receive the selective scores 612 and the weighted overall score from the weighted overall score computation 1014. The persistent learning profile 1016 may aggregate the generated competency scores with prior scores for each dimension, maintaining cumulative competency data across simulation sessions. The persistent learning profile 1016 may be configured to implement selective aggregation, wherein prior scores for each dimension are aggregated only from turns where that respective dimension was determined to be relevant by the filter 1010 (e.g., relevance weight exceeding a threshold). For each competency dimension, the persistent learning profile 1016 may maintain the cumulative score derived from turns where that dimension was determined to be relevant, the number of turns on which the dimension was assessed, and the average relevance weight across assessed turns. The persistent learning profile 1016 may be configured to implement weighted aggregation algorithms that combine new scores with prior scores using relevance-weighted averaging, maintaining running statistics for each dimension including count, sum, and average relevance. This selective aggregation may produce competency profiles that reflect what the user actually demonstrated in contextually appropriate situations, providing higher-fidelity learner data compared to profiles built from uniform scoring across all dimensions regardless of relevance.

[0223] In one example, the process 1000 may be implemented in a distributed configuration where the Al assessment module 1008 executes on cloud-based server infrastructure while the filter 1010, weighted overall score computation 1014, and persistent learning profile 1016 are maintained on application servers closer to the user interface layer. In another example, the process 1000 may be executed on a single server system with shared memory access. The Al assessment module 1008 may interface with multiple Al model providers (e.g., the Al modules 154) via application programming interfaces, with load balancing and failover mechanisms to ensure reliable assessment processing. The persistent learning profile 1016 may be implemented using database systems that support efficient read and write operations for aggregating scores across potentially large numbers of simulation turns and sessions.ATTORNEY DOCKET NO.: 38562.0001P1

[0224] In another example, the process 1000 may be implemented with configurable relevance thresholds that vary based on simulation type, assessment framework, or educator preferences. A higher threshold produces more selective scoring with fewer dimensions assessed per turn, while a lower threshold produces broader scoring with more dimensions assessed. The threshold configuration enables educator / teacher users to balance assessment granularity against the desire for comprehensive competency coverage. In an example, the threshold may be dynamically adjusted based on the user’s progression through the shared simulation session or based on the cumulative assessment data in the persistent learning profile 1016.

[0225] In an example, the process 1000 may support retroactive assessment where previously stored user responses are re-evaluated against a different competency framework 1004. This retroactive capability enables evaluation of the same user performance against multiple competency standards without requiring the user to repeat the simulation, providing flexibility for educator / teach users to apply new assessment frameworks to existing learner data. The persistent learning profile 1016 may maintain separate competency records for each assessment framework applied, or may aggregate scores across frameworks using mapping rules that relate dimensions across different frameworks.

[0226] The process 1000 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing assessment platforms. Prior systems uniformly apply all assessment dimensions to every’ user response regardless of w hether a given response actually provides meaningful evidence for a particular competency, resulting in diluted competency profiles with reduced accuracy and actionability. Prior systems also evaluate user responses in isolation against only the current activity content, preventing assessment of cross-unit learning progression and curriculum-coherent skill development. Prior systems further compute overall scores using uniform weighting across assessment categories regardless of the contextual relevance of each category’ to the specific response. The arrangement and operation of the components in the process 1000 address these technical problems through the integration of a competency framework and curriculum context data into an Al assessment module, the generation of per-dimension relevance weights by the Al assessment module, threshold-based filtering by a filter to produce selective scores, a relevance-weighted overall score computation, and selective aggregation by a persistentATTORNEY DOCKET NO.: 38562.0001P1learning profile. This structural arrangement enables dynamic competency assessment with per-turn relevance determination, curriculum-aware evaluation, and relevance-weighted scoring that was not achievable with prior system architectures.

[0227] FIG. 11 shows an example process 1100 for generating an assessment output, a narrative, and coaching responses. User input 1110 may be received that is associated with input data received from a user during shared simulation session. The user input 1110 may be utilized for the assessment and narrative generation processes. The user input 1110 may comprise data indicative of decisions, actions, or textual responses submitted by a user through the user’s user interface during interaction with the shared simulation session. The user input 1110 may comprise free-form text responses, selection of predefined options, or combinations thereof depending on the simulation design and a specific decision point presented to the user. The user input 1110 may be transmitted from the user’s user device (e.g., one of the electronic devices 102) via network communication protocols.

[0228] Curriculum context data 1120 may provide information from a broader course surrounding the shared simulation session. The curriculum context data 1120 may comprise prior unit content, previously covered concepts and skills, course-wide learning objectives, and upcoming instructional targets. In an example, the curriculum context data 1120 may be retrieved from a curriculum data store that maintains hierarchical content organization including course, unit, episode, and character level information with associated learning materials and objectives at each level. The curriculum context data 1120 may enable both the assessment process and the coaching process to operate with awareness of the broader instructional context, supporting curriculum-coherent evaluation and guidance that references prior learning and connects current challenges to previously covered material.

[0229] An artificial intelligence (Al) assessment module 1140 (e.g., one of the Al modules 154) may receive the user input 1110 and the curriculum context data 1120 as inputs. The Al assessment module 1140 may be configured to evaluate the user input 1110 during the shared simulation session and generate structured assessment output comprising competency scores across applicable competency dimensions, relevance weights indicating the degree to which the user response provides evidence for each dimension, identified reasoning patterns including logical structure and evidence cited, narrative adjustment signals indicating how the user's decision should influenceATTORNEY DOCKET NO.: 38562.0001P1simulation progression, and user struggle indicators that identify when a user appears to be experiencing genuine confusion or misunderstanding. The Al assessment module 1140 may transmit assessment prompts to Al module services (e.g., stored on one or more of the servers 106) via application programming interfaces, with the prompts including the user input data, simulation context, character role information, curriculum context, and a predefined output schema that constrains the Al service to generate validated structured data conforming to specified Aped fields. The predefined output schema ensures that the Al module service produces consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations. The Al assessment module 1140 may process the curriculum context data 1120 to evaluate cross-unit learning progression, determining whether the user is building on knowledge from prior units and making conceptual connections across the curriculum rather than evaluating the response in isolation against only the current simulation content.

[0230] An Al narrative generation module 1150 may receive the structured assessment output from the Al assessment module 1140. The Al narrative generation module 1150 may be configured to generate adaptive narrative content 1180 based on the structured assessment output, producing customized dialogue, scene descriptions, decision options, and feedback content that reflects the user’s character role, information state, and assessed competency level within the shared simulation session. The Al narrative generation module 1150 may process the narrative adjustment signals from the structured assessment output to determine how the shared simulation session narrative should progress based on the user’s decisions and reasoning quality. The Al narrative generation module 1150 may also receive curriculum context data 1120 to incorporate references to prior learning and cross-unit conceptual connections within the generated narrative content, enabling curriculum-coherent educational experiences that build upon previously covered material.

[0231] An Al coaching module 1160 (e.g., one of the Al modules 154) may be configured to operate as a separate process from the Al assessment module 1140 and the Al narrative generation module 1150. For example, the Al coaching module 1160 may have read-only access to the structured assessment output from the Al assessment module 1140. For example, the Al coaching module 1160 may not have authorization to write to or modify the structured assessment output. As an example, this architecturalATTORNEY DOCKET NO.: 38562.0001P1separation ensures that coaching guidance does not contaminate the assessment’s measurement of unassisted user reasoning. The Al coaching module 1160 may be configured to monitor the structured assessment output across a plurality of user turns associated with the shared simulation session and to proactively initiate coaching interactions when the assessment output satisfies predefined intervention conditions. The predefined intervention conditions may comprise a struggle indicator being set to true, consecutive assessment scores falling below a defined threshold, identified misconceptions recurring across multiple turns, and engagement pattern changes such as significantly reduced response length or increased response time. In an example, predefined intervention conditions may further comprise completion of a simulation milestone warranting a reflective coaching prompt. The predefined intervention conditions may be configurable per simulation, per assessment framework, and per educator preference. In an example, the Al coaching module 1160 may receive a structured output: a boolean indicator (“studentstruggling'’) that is set to true when the Al detects signs of genuine confusion or misunderstanding (as distinguished from low effort or disengagement); and a descriptive field (“struggleContext”) specifying what concept or skill the user (e.g., student) appears to be struggling with. The Al coaching module 1160 may receive the curriculum context data 1120, enabling the coaching process to reference prior learning, connect current challenges to previously covered material, and provide guidance (e.g., message configured to output to the corresponding user) that is coherent with the broader instructional sequence.

[0232] As an example, a student giving dismissive or minimal-effort responses may not be struggling and may be disengaged, and proactive coaching may be counterproductive. The Al coaching module 1160 may be configured to distinguish between these conditions by evaluating: whether the student is making attempts that reflect misunderstanding versus non-attempts; whether response patterns indicate confusion versus indifference; and whether scores are declining despite apparent effort.

[0233] In an example, predefined intervention conditions may further comprise: consecutive assessment scores below a defined threshold; identified misconceptions recurring across multiple turns; engagement pattern changes (such as significantly reduced response length or increased response time); and completion of a simulation milestone warranting a reflective coaching prompt. The trigger conditions are configurable per simulation, per assessment framework, and per educator preference.ATTORNEY DOCKET NO.: 38562.0001P1

[0234] An external memory service 1130 may interface with the Al coaching module 1160 to maintain persistent conversational state (e.g., via the messages to the corresponding user) for the coaching process across simulation sessions. The external memory service 1130 may store conversation history between a coaching assistant and each user, derived insights about user learning patterns and communication preferences, and contextual information about user simulation states and assessment performance. When a user returns to the platform for a new simulation session, the Al coaching module 1160 may retrieve stored context from the external memory service 1130, enabling the coaching process to reference prior interactions, recall user-specific learning patterns, and maintain continuity in its coaching relationship with the user across multiple simulation modules. The external memory service 1130 may operate with observed and unobserved participant classifications, where user messages contribute to insight derivation while coaching messages are stored but do not recursively influence the insight derivation process. The external memory service 1130 may be implemented as a cloud-based database service, a dedicated memory management server, or a local storage system depending on deployment requirements and privacy considerations.

[0235] A user interface 1170 may be configured to present shared simulation session content and coaching messages / responses / interactions to the user. The user interface 1170 may comprise a simulation narrative 1172 and a coaching panel 1174 that are visually and functionally distinct from one another. The simulation narrative 1172 may receive adaptive narrative content from the Al narrative generation module 1150. The simulation narrative 1172 may display the shared simulation session, dialogue, scene descriptions, and decision options to the user via a communication interface, or channel, for example. The coaching panel 1174 may receive coaching messages from the Al coaching module 1160. The coaching panel 1174 may display coaching messages / responses / interactions via a separate communication interface, or channel, such as a chat panel display alongside the simulation narrative 1172, that is distinct from the simulation narrative output, for example. The visual and functional separation between the simulation narrative 1172 and the coaching panel 1174 ensures that users can distinguish between simulation content and coaching guidance, and enables the system to maintain clear boundaries between assessed simulation performance and coaching-assisted learning.ATTORNEY DOCKET NO.: 38562.0001P1

[0236] In one example, the process 1100 may be implemented in a distributed configuration where the Al assessment module 1140. the Al narrative generation module 1150, and the Al coaching module 1160 execute on separate server instances that communicate via network protocols. In another example, the Al assessment module 1140, the Al narrative generation module 1150, and the Al coaching module 1160 may be implemented as software processes executing on a single server system with shared memory access. The external memory service 1130 may be implemented as a cloudbased database service accessible via application programming interfaces, enabling scalable storage of conversational state across large numbers of users and sessions. The user interface 1170 may be implemented as a web browser-based application that renders the simulation narrative 1172 and the coaching panel 1174 within a single browser window with distinct visual regions, or as separate interface elements that can be independently positioned or minimized based on user preference.

[0237] In some embodiments, the process 1100 may be implemented with configurable intervention conditions that vary based on simulation type, assessment framework, educator preferences, or individual learner accommodations. The Al coaching module 1160 may calibrate its tone, vocabulary7complexity, and support level based on documented user learning accommodations including Individualized Education Programs, English Language Learner status, Section 504 plans, and Gifted and Talented designations. To maintain privacy compliance, user personally identifiable information may be anonymized via a hashing function before transmission to the Al coaching module 1160 and the external memory service 1130, with the mapping between identifiers and anonymized tokens maintained only within a secure data store.

[0238] In an example, the external memory7service 1130 may be configured to maintain separate conversational state records for each user, enabling the Al coaching module 1160 to provide individualized coaching that reflects each user’s unique learning history and communication patterns.

[0239] The process 1100 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing coaching and assessment platforms. Prior systems that provide coaching assistance during assessed activities cannot distinguish between unassisted user reasoning and coached performance, compromising measurement validity7and producing assessment data that conflates learner ability with coaching effectiveness. Prior systems also lack mechanismsATTORNEY DOCKET NO.: 38562.0001P1for maintaining persistent coaching relationships across sessions, resulting in coaching interactions that cannot reference prior conversations or adapt to user-specific learning patterns over time. Prior systems further provide coaching that operates in isolation from curriculum context, preventing guidance that connects current challenges to prior learning or upcoming instructional targets. The arrangement and operation of the components in the process 1100 addresses these technical problems through the architectural separation between an Al assessment module and an Al coaching module with read-only access to assessment data, a persistent conversational state maintained by an external memory service, and the integration of curriculum context data into both assessment and coaching processes. This structural arrangement enables adaptive simulation with auxiliary Al coaching that maintains measurement validity, provides personalized coaching with cross-session continuity, and delivers curriculum-coherent guidance that was not achievable with prior system architectures.

[0240] FIG. 12 shows an example interface 1200 for displaying interaction points. An artificial intelligence (Al) image generation 1210 may be configured to produce visual assets for a shared simulation session. The Al image generation 1210 may be configured to generate environment images 1220 based on simulation context parameters including historical period, geographic location, time of day, weather conditions, narrative state, and emotional tone. The Al image generation 1210 may interface with external Al image generation services (stored on one or more of the servers 106) via application programming interfaces, transmitting generation prompts that specify the desired visual characteristics and receiving generated image data in response. The Al image generation 1210 may maintain a reference asset library that constrains a style, color palette, and artistic direction of generated visual assets, ensuring visual consistency across a shared simulation session’s environments. In an example, the Al image generation 1210 may generate character images, episode background images, and environmental detail images that can be regenerated or customized by curriculum editors to match specific instructional requirements. The system maintains a reference asset library that constrains the style, color palette, and artistic direction of generated assets, ensuring visual consistency across a simulation’s environments.

[0241] An image 1220 may comprise a display of the output of the Al image generation 1210. The image 1220 may comprise an equirectangular image, a cubemap projection, a cylindrical panorama, or other immersive display formats suitable for panoramic viewingATTORNEY DOCKET NO.: 38562.0001P1experiences. As an example, the image 1220 may comprise a specially distorted two-dimensional (2D) image that, when projected onto the interior surface of a sphere, appears naturally proportioned to an observer positioned at the sphere’s center. For example, the image 1220 may comprise an environment that enables a user to view all directions and zoom in and out from a fixed point of view. For example, the image 1220 may comprise a 3D browser-based 3D environment that allows omnidirectional viewing from a fixed point using modular 3D components with Al generated, or non-generated texturing, with 3D or 3D interactive characters. In an example, using generative Al (e.g., one of the Al modules 154), in addition to non-generated assets, the environment may be textured to resemble a desired setting such as a home interior, a large meeting hall, an outdoor environment, and the like. The environment may enable the user to move between different settings, completing various tasks as directed by a script associated with the shared simulation session and / or a character role. For example, users may be able to interact with various 2D or 3D interactive characters, which may use generative Al dialogue (e.g., via one of the Al modules 154) based on a script and a structure of gameplay associated with the shared simulation session. As an example, the image 1220 may be generated at a resolution of approximately 3840 by 1920 pixels and compressed to under 2 megabytes, enabling rapid loading on low-bandwidth networks and low-powered devices commonly found in educational settings including tablets and Chromebooks. The projection format enables the creation of immersive panoramic experiences without requiring complex three-dimensional (3D) modeling, texture mapping, or dynamic lighting calculations that would impose substantial computational overhead on user devices.

[0242] An assessment output 1230 may be configured to provide competency evaluation data from one or more Al assessment processes. The assessment output 1230 may comprise structured data (e.g., competency data) including competency scores across applicable competency dimensions, relevance weights indicating the degree to which user responses provide evidence for each dimension, identified reasoning patterns, and competency gaps or mastery areas determined through the assessment process. The assessment output 1230 may be utilized for configuring one or more adaptive elements of the shared simulation session, enabling the shared simulation session to respond to user performance by adjusting which interaction opportunities are presented and how they are configured.ATTORNEY DOCKET NO.: 38562.0001P1

[0243] An interaction point configuration 1240 may receive the assessment output 1230. The interaction point configuration 1240 may determine a configuration of one or more interactive points 1264 within the shared simulation session. The interaction point configuration 1240 may process competency evaluation data from the assessment output 1230 to determine which interaction points 1264 of the one or more interactive points 1264 should be active, what content should be loaded into each interaction point 1264, and how prominently each interaction pointl264 should be presented based on the user’s demonstrated competency levels and identified learning needs. When the assessment output 1230 identifies specific competency gaps, the interaction point configuration 1240 may emphasize interaction points 1264 targeting those competencies by adjusting their visual prominence, positioning, or availability. For example, if it is determined that a user has demonstrated weak understanding of economic concepts, the next environment may feature prominently placed interaction points related to trade and commerce. When mastery is demonstrated in particular competency dimensions, the interaction point configuration 1240 may surface more complex decision opportunities or optional exploration paths that challenge the user at an appropriate level.

[0244] Character assets 1250 may provide visual representations of characters within the simulation environment. The character assets 1250 may comprise stylized 2D images that are composited into the spherical environment as billboard elements. Billboard elements may comprise flat image planes that are oriented to always face the viewer regardless of the user’s orientation within the panoramic environment. As an example, this approach avoids the computational overhead of 3D character modeling, skeletal animation, texture mapping, and dynamic lighting while providing visually engaging character representations within the immersive environment. The character assets 1250 may be generated by the Al image generation 1210 based on character descriptions, or may be pre-authored assets stored in the simulation content repository. The character assets 1250 may connect to the interaction point configuration 1240, enabling character visual elements to be positioned and configured based on the narrative state and assessment-driven adaptation requirements.

[0245] A browser-based 3D viewer 1260 may be configured to display the shared simulation session for user interaction. As an example, the browser-based 3D viewer 1260 may be configured to utilize a 3D rendering library to create a sphere geometry, apply the image 1220 as an interior texture, and position a virtual camera at the center ofATTORNEY DOCKET NO.: 38562.0001P1the sphere, creating an immersive panoramic experience within a standard web browser. As an example, the browser-based 3D viewer 1260 may be configured to operate without requiring specialized plugins, native application installation, or high-performance graphics hardware, enabling deployment across the diverse device ecosystem found in educational settings. The browser-based 3D viewer 1260 may receive the image 1220 for rendering as the environment background, the character assets 1250 for compositing as billboard elements, and the interaction point configuration 1240 for placing and configuring interactive points 1264 within the environment according to a user viewpoint 1262

[0246] The user viewpoint 1262 may comprise the virtual camera position at the center of the spherical projection within the browser-based 3D viewer 1260. For example, the user viewpoint 1262 may define a perspective from which the user observes the shared simulation session, with the image 1220 appearing as a surrounding panorama when viewed from this central position. As an example, users may adjust their viewing direction through mouse, touch, or device orientation inputs, enabling exploration of the full panoramic environment while maintaining the user viewpoint 1262 at the sphere’s center.

[0247] The one or more interaction points 1264 may define locations within the spherical environment where users can interact with elements of the shared simulation session. As shown in FIG. 12, the interaction points 1264 may be positioned at one or more locations on the sphere surface surrounding the user viewpoint 1262, as hotspot H2, hotspot H3, and hotspot H4, for example. Each interaction point 1264 may be organized as a template with consistent behavior and game rules that can be loaded with Al-determined content. The interaction point templates may define an interaction mechanic, such as examine, converse, collect, or decide, while content loaded into each interaction point may be determined by the interaction point configuration 1240 based on a narrative state and the user’s assessment performance. The interaction points 1264 may be rendered as visual indicators within the browser-based 3D viewer 1260 that users may interact with and select in order to trigger the associated interaction.

[0248] In one example, the interface 1200 may be implemented in a distributed configuration where the Al image generation 1210 executes on cloud-based server infrastructure while the browser-based 3D viewer 1260 executes on the user’s local device. The image 1220 and character assets 1250 may be transmitted from serverATTORNEY DOCKET NO.: 38562.0001P1infrastructure to user devices via content delivery networks optimized for low-latency delivery of image assets. The interaction point configuration 1240 may execute on server infrastructure with configuration data transmitted to the browser-based 3D viewer 1260, or may execute locally on the user device (e.g., one or more of the electronic devices 102) with the assessment output 1230 transmitted from server-side assessment processes.

[0249] In another example, the interface 1200 may be implemented with progressive generative capability that transitions from template-based environment configuration to fully generative environment creation. In another example, environments are pregenerated or Al-assisted images with template-based interaction point placement by curriculum editors. In advanced embodiments, the Al image generation 1210 may generate environments procedurally in real-time based on narrative state, assessment needs, and user interaction history. In another example, the Al image generation 1210 may generate complete interactive environments including background images (e.g., equirectangular images, cubemap projections, cylindrical panoramas, etc.), character assets 1250, interaction point placement, and interaction content from narrative context alone without human authoring of environment assets.

[0250] In an example, the interface 1200 may support per-user environment personalization when operating in multi-user mode. The same base image 1220 may display different interaction points 1264, different character-specific visual elements from the character assets 1250, and different information overlays to each user based on that corresponding user’s character role and information state. The shared simulation session may serve as the shared spatial context while the interaction point configuration 1240 may maintain per-user customization that reflects each user’s character-specific knowledge and available actions within the simulation scenario. For example, all users may see the same tavern environment, but a merchant user may see trade-related interaction points and commodity prices, while a lawyer user may see legal documents and pamphlets, and a loyalist may see Crown proclamations and military dispatches.

[0251] The computing environment 1200 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in delivering immersive educational experiences in resource-constrained settings. Prior systems that provide immersive three-dimensional environments require substantial computational resources, high-bandwidth network connections, and specialized rendering hardware that are unavailable in many educational settings. Prior systems thatATTORNEY DOCKET NO.: 38562.0001P1provide lightweight web-based experiences lack the immersive quality that supports deeper engagement and contextual learning. Prior systems also lack mechanisms for adapting environment content based on real-time assessment data, resulting in static experiences that cannot respond to individual learner needs. The arrangement and operation of the components in the computing environment 1200 addresses these technical problems through Al image generation that produces optimized images, a browser-based 3D viewer that renders immersive panoramic experiences without specialized hardware requirements, and interaction point configuration that adapts environment content based on an assessment output. This structural arrangement enables adaptive immersive simulation environments that are accessible on resource-constrained devices while providing assessment-responsive learning experiences that was not achievable with prior system architectures.

[0252] FIG. 13 shows an example curriculum context integration process 1300 illustrating how curriculum context data integrates with assessment and coaching processes within a shared simulation session, enabling curriculum-coherent evaluation and guidance that references prior learning and connects current challenges to previously covered material within a broader instructional sequence.

[0253] The curriculum context data 1302 may provides information from a broader course surrounding the active simulation. The curriculum context data 1302 may comprise one or more categories of contextual information comprising one or more of prior unit content, previously covered concepts and skills, course-wide learning objectives, or upcoming instructional targets. The curriculum context data 1302 may be retrieved from a curriculum data store that maintains hierarchical content organization including course, unit, episode, and character level information with associated learning materials and objectives at each level. The curriculum context data 1302 may be structured as data records that can be efficiently queried and transmitted to Al processing modules alongside simulation-specific context, enabling curriculum-coherent evaluation and guidance without requiring the Al services to maintain persistent knowledge of the full course structure.

[0254] User input 1304 may be received from a user during a shared simulation session. The user input 1304 may comprise decisions, actions, or textual responses submitted by a user through the user's user interface during interaction with the shared simulation session. The user input 1304 may comprise free-form text responses, selection ofATTORNEY DOCKET NO.: 38562.0001P1predefined options, or combinations thereof depending on the simulation design and a specific decision point presented to the user. The user input 1304 may be transmitted from the user's user device via network communication protocols.

[0255] An Al assessment module 1306 may receive the user input 1304 and the curriculum context data 1302 as inputs. The Al assessment module 1306 may be configured to evaluate the user input 1304 during the shared simulation session and generate structured assessment output comprising competency scores across applicable competency dimensions, relevance weights indicating the degree to which the user response provides evidence for each dimension, identified reasoning patterns including logical structure and evidence cited, narrative adjustment signals indicating how the user's decision should influence simulation progression, and user struggle indicators that identify when a user appears to be experiencing genuine confusion or misunderstanding. The Al assessment module 1306 may include cross-unit connection detection functionality that evaluates whether the user is building on knowledge from prior units and making cross-unit conceptual connections rather than evaluating the response in isolation against only the current simulation content. For example, if a student in a colonial economics simulation references concepts from a prior unit on mercantilism, the Al assessment module 1306 may recognize this cross-unit connection and credit it in the competency evaluation. The Al assessment module 1306 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including the user input data, simulation context, character role information, curriculum context, and a predefined output schema that constrains the Al service to generate validated structured data conforming to specified typed fields.

[0256] A structured assessment output 1310 may comprise the output generated by the Al assessment module 1306. The structured assessment output 1310 may comprise typed fields including an overall score represented as a numerical value, per-dimension scores for each knowledge, theme, and skill dimension, category -level relevance weights indicating the overall applicability of knowledge assessment, theme assessment, and skill assessment to the current response, per-dimension applicability flags indicating whether individual dimensions were relevant, justification text explaining the scoring rationale, a commentary field providing personalized feedback to the user, a boolean indicator for whether the user is struggling, and / or a struggle context string describing the nature of any identified difficulties when the struggle indicator is true. The structured assessmentATTORNEY DOCKET NO.: 38562.0001P1output 1310 may be structured according to a predefined output schema that ensures consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations.

[0257] An Al coaching module 1308 may be configured to operate as a separate process from the Al assessment module 1306 and an Al narrative generation module 1312. The Al coaching module 1308 may receive the structured assessment output 1310 via read-only access. The Al coaching module 1308 may not have authorization to write to or modify the structured assessment output 1310. This architectural separation ensures that coaching guidance does not contaminate the assessment's measurement of unassisted user reasoning. The Al coaching module 1308 may also receive the curriculum context data 1302, enabling the coaching process to reference prior learning, connect current challenges to previously covered material, and provide guidance that is coherent with the broader instructional sequence. For example, the Al coaching module 1308 may generate a response such as "Remember when we explored the Stamp Act in the last unit? The same principles apply here." The Al coaching module 1308 may include curriculum-coherent guidance functionality that references prior learning and connects current challenges to previously covered material within the broader instructional sequence.

[0258] The Al coaching module 1308 may be configured to monitor the structured assessment output 1310 across a plurality of user turns associated with the shared simulation session and to proactively initiate coaching interactions when the structured assessment output 1310 satisfies predefined intervention conditions. The predefined intervention conditions may comprise a struggle indicator being set to true, consecutive assessment scores falling below a defined threshold, identified misconceptions recurring across multiple turns, engagement pattern changes such as significantly reduced response length or increased response time, and completion of a simulation milestone warranting a reflective coaching prompt. The predefined intervention conditions may be configurable per simulation, per assessment framework, and per educator preference. The Al coaching module 1308 may generate coaching messages 1316 based on the structured assessment output 1310 and the curriculum context data 1302.

[0259] The AT narrative generation module 1312 may receive the structured assessment output 1310 from the Al assessment module 1306. The Al narrative generation module 1312 may be configured to generate adaptive narrative content 1314 based on the structured assessment output 1310, producing customized dialogue, scene descriptions,ATTORNEY DOCKET NO.: 38562.0001P1decision options, and feedback content that reflects the user's character role, information state, and assessed competency level within the shared simulation session. The Al narrative generation module 1312 may process the narrative adjustment signals from the structured assessment output 1310 to determine how the shared simulation session narrative should progress based on the user's decisions and reasoning quality. The adaptive narrative content 1314 may comprise character-specific dialogue and scene descriptions that reflect each character's perspective, character-appropriate decision options for the next simulation turn, feedback integrated into the narrative as in-character reactions from non-player characters that reflect the quality7of the user's reasoning, and character-specific environmental descriptions that reflect the user's information state.

[0260] A user interface 1318 may receive both the adaptive narrative content 1314 from the Al narrative generation module 1312 and the coaching messages 1316 from the Al coaching module 1308. The user interface 1318 may be configured to display shared simulation session content and coaching interactions to the user through distinct communication interfaces. For example, the adaptive narrative content 1314 may be displayed via a first communication interface that displays the shared simulation session, dialogue, scene descriptions, and decision options to the user. The coaching messages 1316 may be displayed via a second communication interface, such as a chat panel displayed alongside the simulation narrative, that is visually and functionally distinct from the simulation narrative output. The visual and functional separation between the simulation narrative and the coaching panel ensures that users can distinguish between simulation content and coaching guidance, and enables the system to maintain clear boundaries between assessed simulation performance and coaching-assisted learning, for example.

[0261] In one example, the process 1300 may be implemented in a distributed configuration where the Al assessment module 1306. the Al narrative generation module 1312, and the Al coaching module 1308 execute on separate server instances that communicate via network protocols. In another example, the Al assessment module 1306, the Al narrative generation module 1312, and the Al coaching module 1308 may be implemented as software processes executing on a single server system with shared memory access. The curriculum context data 1302 may be stored in a centralized curriculum data store accessible via application programming interfaces, or may be replicated across distributed storage systems for improved access latency. The userATTORNEY DOCKET NO.: 38562.0001P1interface 1318 may be implemented as a web browser-based application that renders the simulation narrative and the coaching panel within a single browser window with distinct visual regions, or as separate interface elements that can be independently positioned or minimized based on user preference.

[0262] In an example, the process 1300 may be implemented with hybrid processing configurations where lightweight Al operations such as coaching response generation are performed locally on user devices while computationally intensive operations such as cross-unit connection detection and competency evaluation are performed on server infrastructure. This hybrid approach enables responsive user experiences while leveraging centralized resources for complex processing tasks that require access to comprehensive curriculum context data.

[0263] The components of the process 1300 may be implemented using specific technical structures and processes rather than generic computing abstractions. The Al assessment module 1306 may employ structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing without error-prone text parsing. The Al coaching module 1308 may implement intervention trigger logic that evaluates the structured assessment output 1310 against configurable threshold conditions to determine when proactive coaching is appropriate. The user interface 1318 may implement distinct rendering pipelines for the simulation narrative and coaching panel, ensuring visual and functional separation while maintaining synchronized state with the underlying simulation session.

[0264] The process 1300 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing coaching and assessment platforms. Prior systems evaluate user responses in isolation against only the current activity content, preventing assessment of cross-unit learning progression and curriculum-coherent skill development. Prior systems also provide coaching that operates in isolation from curriculum context, preventing guidance that connects current challenges to prior learning or upcoming instructional targets. Additionally, prior systems that provide coaching assistance during assessed activities cannot distinguish between unassisted user reasoning and coached performance, compromising measurement validity7. The arrangement and operation of the components in the process 1300 address these technical problems through the integration of curriculum context data into both an Al assessment module and an Al coaching module, cross-unit connectionATTORNEY DOCKET NO.: 38562.0001P1detection functionality of an Al assessment module, curriculum-coherent guidance functionality of the Al coaching module, and a read-only access constraint that ensures the Al coaching module cannot modify a structured assessment output. This structural arrangement enables curriculum- aware adaptive simulation with auxiliary Al coaching that maintains measurement validity, provides curriculum-coherent guidance, and evaluates cross-unit learning progression, which was not achievable with prior system architectures that evaluated responses in isolation and provided coaching without curriculum context integration.

[0265] FIG. 14 shows an example teacher-mediated feedback release system 1400 for managing the delivery of Al-generated assessment feedback to users within a shared simulation session. The system 1400 may be configured to provide technical infrastructure through which users (e.g., teachers, educators, administrators, etc.) may control the release of Al-generated feedback, enabling configurable review workflows that balance automated efficiency with human oversight and quality assurance.

[0266] The system 1400 may comprise an Al assessment module 1402 that evaluates user responses during shared simulation sessions. The Al assessment module 1402 may be configured to process user inputs and generate competency evaluations across multiple dimensions including knowledge, theme, and skill categories. The Al assessment module 1402 may implement artificial intelligence processing to analyze the quality, reasoning patterns, and evidence utilization demonstrated in user responses. The Al assessment module 1402 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including user response data, simulation context, character role information, and a predefined output schema that constrains the Al service to generate validated structured data conforming to specified typed fields.

[0267] The system 1400 may comprise Al-generated feedback 1404 that represents the output produced by the Al assessment module 1402. The Al-generated feedback 1404 may comprise personalized commentary addressing the user's reasoning quality, identified strengths and areas for improvement, competency-specific guidance, and suggestions for further exploration within the simulation context. The Al-generated feedback 1404 may be structured according to predefined formats that facilitate downstream processing and review operations.ATTORNEY DOCKET NO.: 38562.0001P1

[0268] The system 1400 may comprise a feedback review queue 1406 that receives the Al-generated feedback 1404 from the Al assessment module 1402. The feedback review queue 1406 may be configured to hold Al-generated assessment feedback for teacher review before release to student users. The feedback review queue 1406 may implement queuing mechanisms that organize pending feedback items by student, by simulation session, by timestamp, or by other configurable sorting criteria. The feedback review queue 1406 may maintain metadata associated with each feedback item including the originating user session, the simulation context, the assessment scores, and confidence indicators generated by the Al assessment module 1402.

[0269] The system 1400 may comprise a release mode configuration 1408 that determines the active pathway for feedback delivery. The release mode configuration 1408 may be configured to implement one or more operational modes comprising one or more of an automatic mode, a teacher-hold mode, or a threshold-based mode. The release mode configuration 1408 may be configurable per simulation, per class, per assignment, or per individual student based on educator preferences and institutional requirements. The release mode configuration 1408 may store configuration parameters that define the active mode and associated threshold values.

[0270] In the automatic mode, the Al-generated feedback 1404 may flow directly from the feedback review queue 1406 to a student interface 1410 without teacher intervention. The automatic mode may be appropriate for formative feedback during low-stakes simulation interactions where immediate feedback supports learning progression without requiring human review. The automatic mode may enable rapid feedback delivery that maintains engagement and supports real-time learning within the simulation session.

[0271] In the teacher-hold mode, all Al-generated feedback 1404 may be routed to a teacher review interface 1412 before release to students. The teacher review interface 1412 may be configured to display pending feedback items to educators with options to view, approve, modify, or reject each feedback item. The teacher review interface 1412 may display the original user response, the simulation context, the Al-generated feedback content, and / or associated assessment scores to provide educators with comprehensive information for review decisions. The teacher review interface 1412 may implement approval workflows that enable educators to approve feedback as generated, modify feedback content before release, and / or reject feedback entirely. In an example,ATTORNEY DOCKET NO.: 38562.0001P1rejected feedback may be routed to rejected feedback 1414 for logging and analysis purposes.

[0272] In the threshold-based mode, a confidence threshold gate 1416 may evaluate the Al assessment module's 1402 confidence in its assessment and route feedback accordingly. The confidence threshold gate 1416 may receive confidence indicators from the Al assessment module 1402 that quantify the model's certainty in its evaluation. When the confidence indicator exceeds a defined threshold, the feedback may be automatically released to the student interface 1410 without teacher intervention. When the confidence indicator falls below the threshold, the feedback may be routed to the teacher review interface 1412 for human review before release. The confidence threshold gate 1416 may implement comparison algorithms that evaluate numerical confidence values against configurable threshold parameters to produce routing decisions.

[0273] The student interface 1410 may be configured to present released feedback to users within the shared simulation session context. The student interface 1410 may display feedback content alongside the user's original response and the simulation narrative, enabling users to understand the feedback in context. The student interface 1410 may implement visual formatting that distinguishes feedback content from simulation narrative content, ensuring users can identify assessment feedback within the broader simulation experience.

[0274] The teacher review interface 1412 may be configured to provide educators with comprehensive review capabilities. The teacher review interface 1412 may display pending feedback items in a queue format with filtering and sorting options. The teacher review interface 1412 may enable educators to view the full context of each feedback item including the user's response history, the simulation state at the time of response, and the Al's reasoning for its assessment. The teacher review interface 1412 may provide editing tools that enable educators to modify feedback content, adjust tone, add clarifications, or incorporate additional guidance before release. The teacher review interface 1412 may implement batch operations that enable educators to approve multiple feedback items simultaneously when appropriate.

[0275] The rejected feedback 1414 may be configured to store feedback items that educators have determined should not be released to students. The rejected feedback 1414 may maintain records of rejected items along with rejection reasons provided by educators, enabling analysis of Al assessment quality’ and identification of patterns thatATTORNEY DOCKET NO.: 38562.0001P1may warrant system calibration. The rejected feedback 1414 may provide data for continuous improvement of the Al assessment module 1402 through identification of systematic assessment errors or inappropriate feedback generation.

[0276] In one example, the system 1400 may be implemented in a distributed configuration where the Al assessment module 1402 executes on cloud-based server infrastructure while the feedback review queue 1406, the student interface 1410, and the teacher review interface 1412 are implemented as web-based applications accessible through standard browsers. In another example, the system 1400 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The release mode configuration 1408 may be stored in a configuration database that supports real-time updates, enabling educators to switch between modes without system restart or session interruption.

[0277] In an example, the system 1400 may be implemented with granular configuration options that enable different release modes for different feedback types. Formative feedback during simulation interactions may be configured for automatic release while summative feedback at simulation completion may be configured for teacher-hold mode. The system 1400 may support per-student configuration that enables educators to apply stricter review requirements for students who may benefit from more carefully curated feedback.

[0278] In an example, the system 1400 may be implemented with notification mechanisms that alert educators when feedback items are pending review. The teacher review interface 1412 may integrate with email, messaging platforms, or learning management system notification systems to ensure educators are aware of pending review items. The system 1400 may implement escalation logic that automatically releases feedback after a configurable timeout period if teacher review has not occurred, preventing indefinite feedback delays.

[0279] The components of the system 1400 may be implemented using specific technical structures and processes. The Al assessment module 1402 may employ structured output schemas that constrain Al model responses to validated data formats with ty ped fields including confidence indicators, enabling reliable downstream routing decisions. The feedback review queue 1406 may implement database-backed queuing with transactional integrity to ensure feedback items are not lost or duplicated during routing operations. The confidence threshold gate 1416 may implement comparisonATTORNEY DOCKET NO.: 38562.0001P1algorithms with configurable threshold parameters that can be adjusted without code modification. The teacher review interface 1412 may implement optimistic locking mechanisms that prevent concurrent modification of the same feedback item by multiple educators.

[0280] The system 1400 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing feedback delivery platforms. Prior systems that generate Al feedback typically deliver all feedback automatically without human oversight, preventing educators from catching inappropriate or inaccurate feedback before student exposure. Prior systems that require human review of all feedback create bottlenecks that delay feedback delivery and increase educator workload, reducing the timeliness benefits of automated assessment. Prior systems also lack mechanisms for confidence-based routing that would enable automatic release of high-confidence feedback while flagging uncertain assessments for human review. The arrangement and operation of the components in the system 1400 address these technical problems through a feedback review queue that holds feedback for configurable review, a release mode configuration that enables educators to select appropriate review workflows, a confidence threshold gate that routes feedback based on Al confidence levels, and a teacher review interface that provides comprehensive review and modification capabilities. This structural arrangement enables teacher-mediated feedback release with configurable automation that balances efficiency with quality assurance, which was not achievable with prior system architectures that implemented either fully automatic or fully manual feedback delivery without confidence-based routing.

[0281] FIG. 15 shows an example multi-modal assignment system architecture 1500 for integrating traditional assessment methods with simulation-based assessment w ithin a unified competency evaluation framework. The system architecture 1500 may be configured to provide technical infrastructure through which multiple assessment modalities including essay assignments, short response questions, and multiple choice questions are evaluated against a common competency framework and aggregated with simulation assessment data to produce unified competency profiles that capture comprehensive learner performance across diverse assessment types.

[0282] The system architecture 1500 may comprise an assignment configuration 1502 that defines parameters for assessment activities outside of simulation interactions. TheATTORNEY DOCKET NO.: 38562.0001P1assignment configuration 1502 may be configured to specify assignment types, timing relationships to simulation sessions, competency dimension mappings, and scoring criteria for each assessment activity. The assignment configuration 1502 may enable educators to configure assignments as pre-simulation assessments to establish baseline knowledge, post-simulation assessments to evaluate learning outcomes, standalone assessments independent of simulation timing, and / or end-of-unit assessments that synthesize learning across multiple simulation episodes. The assignment configuration 1502 may store configuration parameters in a structured data format that facilitates retrieval and application during assessment processing operations.

[0283] The system architecture 1500 may comprise essay assignments 1504 that represent extended written response assessments configured through the assignment configuration 1502. The essay assignments 1504 may be configured to evaluate complex reasoning, argumentation quality, evidence utilization, and synthesis capabilities that require extended written expression. The essay assignments 1504 may implement Al rubric evaluation using category-specific scoring criteria that align with the competency dimensions defined in the competency framework. The essay assignments 1504 may receive user-submitted written responses and transmit these responses to downstream assessment processing components along with associated rubric configurations and scoring parameters.

[0284] The system architecture 1500 may comprise short response questions 1506 that represent brief written response assessments configured through the assignment configuration 1502. The short response questions 1506 may be configured to evaluate focused knowledge recall, concept application, and targeted reasoning within constrained response formats. The short response questions 1506 may implement criteria-based scoring that evaluates responses against predefined correctness criteria, partial credit rubrics, and competency dimension mappings. The short response questions 1506 may receive user-submitted responses and transmit these responses to downstream assessment processing components along with associated scoring criteria and expected response parameters.

[0285] The system architecture 1500 may comprise multiple choice questions 1508 that represent selection-based assessments configured through the assignment configuration 1502. The multiple choice questions 1508 may be configured to evaluate knowledge recognition, concept discrimination, and application of learned principles throughATTORNEY DOCKET NO.: 38562.0001P1selection among predefined response options. The multiple choice questions 1508 may implement question variant generation for retake integrity, producing alternative question formulations that assess equivalent competencies while preventing memorization-based performance inflation on repeated attempts. The multiple choice questions 1508 may receive user selections and transmit these selections to downstream assessment processing components along with correct answer mappings and competency dimension associations.

[0286] The system architecture 1500 may comprise an Al assessment process 1510 that receives assessment data from the essay assignments 1504, the short response questions 1506, and the multiple choice questions 1508. The Al assessment process 1510 may be configured to evaluate user responses across all assessment modalities using artificial intelligence processing that analyzes response quality, reasoning patterns, and evidence utilization. The Al assessment process 1510 may implement modality-specific evaluation algorithms that apply appropriate assessment methodologies to each response type while generating standardized competency evaluation output that enables cross-modality aggregation. The Al assessment process 1510 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including user response data, assessment context, rubric configurations, and predefined output schemas that constrain Al service responses to validated structured data formats.

[0287] The system architecture 1500 may comprise a competency framework 1512 that provides competency dimension definitions and scoring criteria to the Al assessment process 1510. The competency framework 1512 may be configured to define knowledge dimensions, theme dimensions, and skill dimensions against which user responses are evaluated. The competency framew ork 1512 may specify scoring rubrics, dimension descriptions, and relevance w eighting parameters that guide the Al assessment process 1510 in generating competency evaluations. The competency framework 1512 may be configurable per course, per unit, or per assignment, enabling educators to apply different competency standards to different assessment activities while maintaining consistency within defined assessment contexts.

[0288] The system architecture 1500 may comprise a relevance filter 1514 that receives competency evaluation output from the Al assessment process 1510. The relevance filter 1514 may be configured to apply threshold-based filtering to per-dimension relevance weights generated by the Al assessment process 1510, separating relevant competencyATTORNEY DOCKET NO.: 38562.0001P1dimensions from non-relevant dimensions for each specific user response. The relevance filter 1514 may implement comparison algorithms that evaluate numerical relevance weights against configurable threshold parameters to produce binary inclusion decisions for each competency dimension. Competency dimensions having relevance weights exceeding the threshold may pass through the relevance filter 1514 and receive scores for the current assessment, while competency dimensions falling below the threshold may be excluded from scoring for that particular response.

[0289] The system architecture 1500 may comprise a unified competency profile 1516 that receives filtered assessment data from the relevance filter 1514. The unified competency profile 1516 may be configured to aggregate competency scores from multiple assessment modalities into a single coherent competency representation for each user. The unified competency profile 1516 may implement weighted aggregation algorithms that combine scores from essay assignments, short response questions, and multiple choice questions according to configurable weighting schemes that reflect the relative evidence value of each assessment modality. The unified competency profile 1516 may maintain per-dimension statistics including score counts, score sums, and average relevance weights across all contributing assessments.

[0290] The system architecture 1500 may comprise simulation assessment data 1518 that provides competency evaluation data from simulation-based assessments to the unified competency profile 1516. The simulation assessment data 1518 may represent structured assessment output generated during shared simulation session interactions, including per-dimension competency scores, relevance weights, and reasoning pattern indicators. The simulation assessment data 1518 may enable the unified competency profile 1516 to integrate simulation performance with traditional assignment performance, producing comprehensive competency profiles that capture learner capabilities demonstrated across both interactive simulation experiences and conventional assessment formats.

[0291] The system architecture 1500 may comprise a persistent learning profile 1520 that receives aggregated competency data from the unified competency profile 1516. The persistent learning profile 1520 may be configured to maintain longitudinal competency records that track learner performance across multiple assessment activities and simulation sessions over time. The persistent learning profile 1520 may implement selective aggregation wherein competency scores are aggregated only from assessmentsATTORNEY DOCKET NO.: 38562.0001P1where the corresponding dimension was determined to be relevant by the relevance filter 1514. The persistent learning profile 1520 may store cumulative competency data that enables adaptive curriculum adjustment, personalized learning pathway recommendations, and longitudinal progress tracking across extended instructional sequences.

[0292] In one example, the system architecture 1500 may be implemented in a distributed configuration where the Al assessment process 1510 executes on cloud-based server infrastructure while the assignment configuration 1502, the relevance filter 1514, and the unified competency profile 1516 are maintained on application servers. In another example, the system architecture 1500 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The competency framework 1512 may be stored in a framework repository that supports versioning and retroactive application of updated frameworks to historical assessment data. The persistent learning profile 1520 may be implemented using database systems that support efficient read and write operations for aggregating scores across potentially large numbers of assessments and simulation sessions.

[0293] In an example, the system architecture 1500 may be implemented with modality-specific weighting configurations that assign different evidence weights to different assessment types based on the depth of reasoning they elicit. Essay assignments 1504 may receive higher weighting for skill dimensions that require extended argumentation, while multiple choice questions 1508 may receive higher weighting for knowledge dimensions that assess factual recall. The weighting configurations may be adjustable per competency framework, enabling educators to calibrate the relative contribution of each assessment modality to the unified competency profile 1516.

[0294] In an example, the system architecture 1500 may be implemented with question variant generation capabilities that produce alternative formulations of multiple choice questions 1508 for retake scenarios. The variant generation may implement semantic equivalence algorithms that ensure alternative questions assess the same competency¬ dimensions while varying surface-level question content to prevent memorization-based performance. The variant generation may maintain mappings between original questions and generated variants to enable consistent competency tracking across retake attempts.

[0295] The components of the system architecture 1500 may be implemented using specific technical structures and processes. The Al assessment process 1510 may employATTORNEY DOCKET NO.: 38562.0001P1structured output schemas that constrain Al model responses to validated data formats with typed fields, enabling reliable downstream processing and aggregation. The relevance filter 1514 may implement comparison algorithms with configurable threshold parameters that can be adjusted without code modification. The unified competency profile 1516 may implement weighted aggregation algorithms that combine scores from multiple modalities using relevance-weighted averaging, maintaining running statistics for each competency dimension. The persistent learning profile 1520 may implement efficient storage schemas that support rapid insertion of new assessment records and retrieval of historical performance data for adaptive system behavior.

[0296] The system 1500 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing multimodal assessment platforms. Prior systems that support multiple assessment types typically maintain separate competency records for each modality, preventing unified analysis of learner capabilities across assessment formats. Prior systems also lack mechanisms for relevance-weighted aggregation that would prevent dilution of competency profiles with assessment data from dimensions not meaningfully evaluated by particular response types. Additionally, prior systems that integrate simulation and assignment assessments do not apply consistent competency frameworks across modalities, resulting in incompatible competency data that cannot be meaningfully aggregated. The arrangement and operation of the components in the system 1500 address these technical problems through a unified competency profile that aggregates data from multiple assessment modalities, a relevance filter that applies threshold-based filtering to prevent profile dilution, a competency framework that provides consistent evaluation criteria across modalities, and a integration of simulation assessment data with traditional assignment assessments through a common aggregation pathway. This structural arrangement enables multi-modal assessment with unified competency profiling that was not achievable with prior system architectures that maintained separate assessment records without relevance-weighted cross -modality aggregation.

[0297] FIG. 16 shows an example multi -character perspective system 1600 for enabling a single user to experience a simulation episode through multiple character viewpoints, thereby supporting assessment of perspective synthesis competencies and cross-character reasoning capabilities. The system 1600 may be configured to provide technical infrastructure through which users engage with the same simulation scenarioATTORNEY DOCKET NO.: 38562.0001P1from different character perspectives across sequential playthroughs, with performance data aggregated to evaluate the user's ability’ to synthesize diverse viewpoints and demonstrate understanding across multiple stakeholder positions within the simulation context.

[0298] The system 1600 may comprise a simulation episode 1602 that represents a discrete simulation scenario configured for multi-perspective engagement. The simulation episode 1602 may define the narrative context, environmental conditions, decision points, and non-player character interactions that constitute the simulation experience. The simulation episode 1602 may be structured to support meaningful differentiation between character perspectives, ensuring that each character viewpoint provides distinct information access, decision options, and narrative experiences while maintaining coherence within the shared scenario context. The simulation episode 1602 may store scenario parameters including historical period, geographic setting, conflict dynamics, and stakeholder relationships that inform the character-specific experiences generated for each playthrough.

[0299] The system 1600 may comprise a plurality of character definitions. As an example, the system 1600 may comprise character definitions for a character A 1604, a character B 1606, and a character C 1608. Each character definition may specify the identity, background, motivations, knowledge state, and available actions associated with a particular perspective within the simulation episode 1602. Character A 1604 may correspond to a merchant character with knowledge of trade economics, commercial relationships, and market dynamics. Character B 1606 may correspond to a lawyer character with knowledge of legal arguments, philosophical treatises, and political alliances. Character C 1608 may correspond to a loyalist character with knowledge of Crown policies, official proclamations, and loyalist networks. The character definitions may be configured to create meaningful information asymmetry between perspectives, ensuring that each character viewpoint provides access to distinct subsets of the simulation's full information state.

[0300] The system 1600 may comprise a plurality of playthrough instances including playthrough A 1610, playthrough B 1612, and playthrough C 1614. Each playthrough instance may represent a complete engagement with the simulation episode 1602 from the perspective of a corresponding character. Playthrough A 1610 may instantiate the simulation episode 1602 with the user occupying the role defined by character A 1604,ATTORNEY DOCKET NO.: 38562.0001P1receiving narrative content, decision options, and assessment feedback appropriate to the merchant perspective. Playthrough B 1612 may instantiate the same simulation episode 1602 with the user occupying the role defined by character B 1606, experiencing the scenario from the lawyer perspective with corresponding differences in available information and decision opportunities. Playthrough C 1614 may instantiate the simulation episode 1602 with the user occupying the role defined by character C 1608, providing the loyalist perspective on the same underlying scenario events.

[0301] A user 1622 may engage with each of the play through instances sequentially. For example, the user 1622 may engage with playthrough A 1610, playthrough B 1612, and playthrough C 1614 across separate sessions, experiencing the simulation episode 1602 from each character perspective in turn. The sequential engagement pattern may enable the user 1622 to develop an understanding of how' the same scenario events appear from different stakeholder positions, building capacity for perspective synthesis and multi-viewpoint reasoning. The user 1622 may receive assessment feedback during each playthrough that evaluates performance against competency dimensions relevant to the active character perspective.

[0302] The system 1600 may comprise cross-character performance tracking 1616 that receives assessment data from each of the playthrough instances. The cross-character performance tracking 1616 may be configured to aggregate competency scores, reasoning patterns, and performance indicators across the user's engagements with different character perspectives within the same simulation episode 1602. The crosscharacter performance tracking 1616 may maintain records of how the user 1622 performed when occupying each character role, enabling comparison of competency demonstration across perspectives and identification of patterns in how7the user engages with different stakeholder viewpoints. The cross-character performance tracking 1616 may implement data structures that associate assessment records with both the user identity and the character perspective, enabling queries that retrieve performance data filtered by character role or aggregated across all perspectives.

[0303] The system 1600 may comprise a perspective synthesis evaluation 1618 that receives aggregated performance data from the cross-character performance tracking 1616. The perspective synthesis evaluation 1618 may be configured to evaluate the user's ability to synthesize understanding across multiple character perspectives, assessing competencies that emerge from multi-viewpoint engagement rather than single-ATTORNEY DOCKET NO.: 38562.0001P1perspective performance. The perspective synthesis evaluation 1618 may implement evaluation algorithms that analyze patterns across playthrough data, identifying indicators of perspective synthesis capability including recognition of how different stakeholders perceive the same events, understanding of how character-specific knowledge shapes reasoning and decision-making, ability to identify common ground and points of conflict between perspectives, and capacity to formulate positions that account for multiple stakeholder interests. The perspective synthesis evaluation 1618 may generate perspective synthesis scores that quantify the user's demonstrated ability to integrate understanding across character viewpoints.

[0304] The system 1600 may comprise a persistent learning profile 1620 that receives evaluation output from the perspective synthesis evaluation 1618. The persistent learning profile 1620 may be configured to store longitudinal competency records that include both per-character performance data and cross-character synthesis evaluations. The persistent learning profile 1620 may maintain cumulative records of the user's perspective-taking capabilities demonstrated across multiple simulation episodes, enabling tracking of growth in multi-viewpoint reasoning competencies over time. The persistent learning profile 1620 may implement aggregation algorithms that combine perspective synthesis scores with traditional competency scores, producing comprehensive learner profiles that capture both domain-specific knowledge and crosscutting reasoning capabilities.

[0305] In one example, the system 1600 may be implemented in a distributed configuration where the simulation episode 1602 and character definitions are stored on content servers while the playthrough instances execute on application servers that coordinate with Al assessment services for real-time evaluation. In another example, the system 1600 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The cross-character performance tracking 1616 may be implemented using database systems that support efficient storage and retrieval of assessment records indexed by both user identity and character perspective. The perspective synthesis evaluation 1618 may be implemented as an Al-powered analysis module that processes aggregated playthrough data to identify patterns indicative of perspective synthesis capability.

[0306] In an example, the system 1600 may be implemented with configurable playthrough sequencing that varies the order in which users experience differentATTORNEY DOCKET NO.: 38562.0001P1character perspectives. The sequencing configuration may enable educators to specify whether users should experience perspectives in a fixed order, a randomized order, or a user-selected order based on pedagogical objectives. The system 1600 may track the sequencing pattern for each user to enable analysis of how playthrough order affects perspective synthesis development.

[0307] In an example, the system 1600 may be implemented with inter-pl ay through reflection prompts that encourage users to explicitly compare their experiences across character perspectives. Between playthroughs, the system 1600 may present reflection questions that ask users to identify differences in how events appeared from different viewpoints, articulate the reasoning behind decisions made from each perspective, and synthesize understanding across the perspectives experienced thus far. Responses to reflection prompts may be evaluated by Al assessment processes and incorporated into the perspective synthesis evaluation 1618.

[0308] The components of the system 1600 may be implemented using specific technical structures and processes. The simulation episode 1602 may implement scenario definition schemas that specify character-agnostic scenario parameters alongside character-specific information state configurations. The character definitions may implement role specification formats that define knowledge boundaries, available actions, and perspective-specific narrative content for each character. The crosscharacter performance tracking 1616 may implement multi-dimensional indexing that enables efficient retrieval of assessment records filtered by user, character, simulation episode, or competency dimension. The perspective synthesis evaluation 1618 may implement pattern recognition algorithms that identify indicators of cross-perspective understanding in aggregated playthrough data.

[0309] The system 1600 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing singleperspective simulation platforms. Prior systems that provide simulation-based learning typically offer only a single character perspective per scenario, preventing assessment of perspective-taking and synthesis competencies that require multi-viewpoint engagement. Prior systems also lack mechanisms for tracking performance across multiple playthroughs of the same scenario from different perspectives, resulting in fragmented assessment data that cannot support cross-character analysis. Additionally, prior systems that support multiple character options do not implement perspective synthesisATTORNEY DOCKET NO.: 38562.0001P1evaluation that analyzes patterns across character-specific performances to assess higher-order reasoning capabilities. The arrangement and operation of the components in the system 1600 address these technical problems through a simulation episode that supports multi-perspective instantiation, character definitions that create meaningful information asymmetry between viewpoints, cross-character performance tracking that aggregates assessment data across playthroughs, and a perspective synthesis evaluation that analyzes aggregated data to assess cross-perspective reasoning capabilities. This structural arrangement enables multi -character perspective assessment with synthesis evaluation that was not achievable with prior system architectures that provided singleperspective simulation experiences without cross-character performance analysis.

[0310] FIG. 17 shows an example Al-generated assessment framework creation system 1700 for automatically generating competency assessment frameworks from course parameters within a shared simulation session. The system 1700 may be configured to provide technical infrastructure through which educators may rapidly create assessment frameworks by providing high-level course descriptions, subject area specifications, and grade level parameters, with artificial intelligence processing generating structured competency dimensions that can be applied to simulation assessment without requiring manual authoring of individual competency definitions.

[0311] A course description 1702 may comprise textual input describing educational content, learning objectives, and instructional focus of a course or unit for which an assessment framework is to be generated. The course description 1702 may comprise narrative descriptions of course content, lists of topics to be covered, statements of learning outcomes, and references to source materials or curriculum standards that inform the course design. The course description 1702 may be provided by users (e.g., teachers, educators, administrators, etc.) via an administrative interface. As an example, the course description 1702 may vary in length and specificity based on educator preferences and available course documentation. The course description 1702 may be structured as free-form text or may conform to predefined templates that facilitate downstream processing by Al components.

[0312] A subject area 1704 may comprise categorical input specifying academic discipline or domain to which the course belongs. The subject area 1704 may comprise designations such as history, science, mathematics, language arts, social studies, or other academic disciplines recognized within educational curricula. The subject area 1704 mayATTORNEY DOCKET NO.: 38562.0001P1inform the Al framework generation process by providing domain-specific context that shapes the types of competency dimensions generated, the vocabulary used in dimension descriptions, and the assessment criteria appropriate for the specified discipline. The subject area 1704 may be selected from a predefined list of supported disciplines or may be specified as custom text for specialized or interdisciplinary courses.

[0313] A grade level 1706 may comprise input specifying an educational level of the target learner population. The grade level 1706 may comprise designations such as elementary, middle school, high school, Advanced Placement, undergraduate, graduate, or professional development levels. As an example, the grade level 1706 may inform the Al framework generation process by calibrating the complexity of competency dimensions, the sophistication of expected reasoning, and the vocabulary used in assessment criteria to be appropriate for the specified learner population. As an example, the grade level 1706 may enable the same course description to produce different assessment frameworks when applied to different grade levels, with higher grade levels producing more demanding competency expectations and more nuanced assessment criteria.

[0314] The system 1700 may comprise an Al framework generation module 1708 that receives the course description 1702, the subject area 1704, and the grade level 1706 as inputs. The Al framework generation module 1708 may be configured to process these inputs using artificial intelligence techniques to generate structured competency framew orks without requiring manual authoring of individual competency dimensions. The Al framework generation module 1708 may implement natural language processing to analyze the course description 1702 and extract key concepts, learning objectives, and skill requirements. The Al framework generation module 1708 may apply domainspecific knowledge associated with the subject area 1704 to ensure generated competency dimensions align with disciplinary conventions and assessment practices. The Al framework generation module 1708 may calibrate the complexity and expectations of generated dimensions based on the grade level 1706 to produce age-appropriate assessment criteria. The Al framework generation module 1708 may transmit generation prompts to external Al model services via application programming interfaces, with the prompts including the course parameters and predefined output schemas that constrain Al service responses to validated structured data formats conforming to competency framew ork specifications.ATTORNEY DOCKET NO.: 38562.0001P1

[0315] The system 1700 may comprise a generated assessment framework 1710 that represents the output produced by the Al framework generation module 1708. The generated assessment framework 1710 may comprise structured competency dimensions organized into categorical groupings including knowledge chunks, theme chunks, and skill chunks. The knowledge chunks may define factual content domains relevant to the course, specifying what information learners should know and how knowledge should be assessed. The theme chunks may define cross-cutting conceptual understandings that span multiple knowledge areas, specifying thematic connections and analytical perspectives relevant to the course content. The skill chunks may define cognitive and analytical competencies that learners should demonstrate, specifying reasoning capabilities, evidence evaluation abilities, and synthesis skills appropriate for the course and grade level. Each competency dimension within the generated assessment framework 1710 may comprise a unique identifier, a descriptive title, a detailed description specifying what the dimension measures and how it should be evaluated, and scoring criteria defining the scale for evaluation.

[0316] The system 1700 may comprise a teacher review module 1712 that receives the generated assessment framework 1710 for educator evaluation and modification. The teacher review module 1712 may be configured to display the Al-generated competency dimensions to educators through an administrative interface that enables viewing, editing, approval, or rejection of individual dimensions. The teacher review module 1712 may enable educators to modify dimension titles, descriptions, or scoring criteria to better align with specific instructional objectives or institutional requirements. The teacher review module 1712 may enable educators to add custom dimensions not generated by the Al framework generation module 1708 or to remove dimensions that are not relevant to the specific course implementation. The teacher review module 1712 may implement version tracking that maintains records of modifications made to AI-generated frameworks, enabling analysis of how educators refine Al-generated content and informing improvements to the Al framework generation module 1708.

[0317] The system 1700 may comprise a simulation assessment configuration 1714 that receives approved assessment frameworks from the teacher review 1712 for application to active simulation sessions. The simulation assessment configuration 1714 may be configured to integrate the approved competency framework with simulation assessment processes, enabling user responses during simulation interactions to be evaluated againstATTORNEY DOCKET NO.: 38562.0001P1the generated competency dimensions. The simulation assessment configuration 1714 may transmit framework configurations to Al assessment modules that evaluate user inputs, providing dimension definitions, scoring criteria, and relevance weighting parameters that guide the assessment process. The simulation assessment configuration 1714 may enable real-time application of newly generated frameworks to ongoing simulation sessions without requiring system restart or session interruption.

[0318] The system 1700 may comprise stored simulation data 1716 comprising historical user response data from prior simulation sessions. The stored simulation data 1716 may be configured to receive approved assessment frameworks from the teacher review 1712 for retroactive application to previously collected user responses. As an example, the stored simulation data 1716 may enable educators to apply newly generated assessment frameworks to historical simulation data, generating competency evaluations against the new framework without requiring users to repeat simulation interactions. The stored simulation data 1716 may maintain associations between user responses and the assessment frameworks under which they were originally evaluated, enabling comparison of user performance across different framework versions and supporting analysis of how framework changes affect competency profiles.

[0319] In one example, the system 1700 may be implemented in a distributed configuration where the Al framework generation module 1708 executes on cloud-based server infrastructure while the teacher review 1712 is implemented as a web-based administrative interface accessible through standard browsers. In another example, the system 1700 may be implemented in a centralized configuration where all components execute on a single server system with shared memory access. The generated assessment framework 1710 may be stored in a framework repository that supports versioning, enabling tracking of framework evolution over time and rollback to prior versions if needed. The stored simulation data 1716 may be implemented using database systems that support efficient storage and retrieval of user response data indexed by simulation session, user identity, and assessment framework version.

[0320] In an example, the system 1700 may be implemented with iterative refinement capabilities that enable educators to provide feedback on generated frameworks that informs subsequent generation operations. The Al framework generation module 1708 may incorporate educator feedback patterns to improve the quality and relevance of generated competency dimensions over time. The system 1700 may implement machineATTORNEY DOCKET NO.: 38562.0001P1learning techniques that analyze patterns in educator modifications to Al-generated frameworks, identifying systematic adjustments that can be incorporated into the generation process to reduce the need for manual refinement.

[0321] In an example, the system 1700 may be implemented with framework template libraries that provide starting points for Al generation based on common course types or curriculum standards. The Al framework generation module 1708 may reference template libraries when generating frameworks for courses that align with recognized curriculum standards such as Advanced Placement, Common Core, or discipline-specific professional standards. The template libraries may accelerate framework generation and improve alignment with established assessment practices while still enabling customization based on specific course descriptions and educator preferences.

[0322] The components of the system 1700 may be implemented using specific technical structures and processes. The Al framework generation module 1708 may employ structured output schemas that constrain Al model responses to validated data formats with typed fields for dimension identifiers, titles, descriptions, and scoring criteria, enabling reliable downstream processing and integration with assessment systems. The teacher review module 1712 may implement optimistic locking mechanisms that prevent concurrent modification of the same framework by multiple educators. The simulation assessment configuration 1714 may implement framework switching logic that validates framework compatibility with active simulations and manages state transitions between framework versions. The stored simulation data 1716 may implement multi-version indexing that enables efficient retrieval of user responses for re-evaluation against different framework versions.

[0323] The system 1700 represents a technical improvement over prior adaptive learning systems that addresses specific technical problems inherent in existing assessment framework creation platforms. Prior systems that require manual authoring of competency frameworks impose substantial time burdens on educators, limiting the adoption of competency -based assessment and preventing rapid iteration on assessment criteria. Prior systems also lack mechanisms for calibrating assessment frameworks to specific grade levels, resulting in frameworks that may be inappropriately demanding or insufficiently rigorous for target learner populations. Additionally, prior systems that support framework creation do not enable retroactive application of new frameworks to historical data, forcing educators to choose assessment criteria before data collection andATTORNEY DOCKET NO.: 38562.0001P1preventing comparison of user performance across different competency standards. The arrangement and operation of the components in the system 1700 address these technical problems through an Al framework generation module that generates structured competency dimensions from high-level course parameters, grade level input that enables age-appropriate calibration of generated frameworks, a teacher review module that enables educator refinement of Al-generated content, and stored simulation data that enables retroactive framework application to historical user responses. This structural arrangement enables Al-assisted assessment framework creation with grade-level calibration and retroactive application capability that was not achievable with prior system architectures that required manual framework authoring without Al assistance or retroactive application support.

[0324] FIG. 18 shows an example struggle detection and coaching trigger logic system 1800 for determining when proactive coaching interventions are appropriate based on assessment output analysis within a shared simulation session. The system 1800 may be configured to provide technical infrastructure through which user response patterns are evaluated to distinguish between genuine learning struggles that warrant coaching support and disengagement patterns that may not benefit from proactive intervention, enabling targeted coaching delivery that improves learning outcomes while avoiding counterproductive interventions, for example.

[0325] A user response 1802 that may comprise input data received from a user during interaction with a shared simulation session. As an example, the user response 1802 may comprise decisions, actions, or textual responses submitted by a user through the user's user interface during simulation engagement. As an example, the user response 1802 may include free-form text responses, selection of predefined options, or combinations thereof depending on the simulation design and the specific decision point presented to the user. The user response 1802 may be transmitted from the user's user device to assessment processing components via network communication protocols, with the response data structured according to predefined formats that facilitate downstream evaluation and pattern analysis.

[0326] The system 1800 may comprise an Al assessment module 1804 that receives the user response 1802 for evaluation against competency frameworks associated with the shared simulation session. The Al assessment module 1804 may be configured to process user responses and generate competency evaluations across multiple dimensionsATTORNEY DOCKET NO.: 38562.0001P1including knowledge, theme, and skill categories. The Al assessment module 1804 may implement artificial intelligence processing to analyze the quality, reasoning patterns, and evidence utilization demonstrated in user responses. The Al assessment module 1804 may transmit assessment prompts to external Al model services via application programming interfaces, with the prompts including user response data, simulation context, character role information, and a predefined output schema that constrains the Al service to generate validated structured data conforming to specified typed fields.

[0327] The system 1800 may comprise a structured assessment output 1806 that represents the output produced by the Al assessment module 1804. The structured assessment output 1806 may comprise typed fields including an overall score represented as a numerical value, per-dimension scores for each knowledge, theme, and skill dimension, relevance weights indicating the applicability of each dimension to the current response, justification text explaining the scoring rationale, a commentary field providing personalized feedback, a boolean indicator for whether the user is struggling, and a struggle context string describing the nature of any identified difficulties when the struggle indicator is true. The structured assessment output 1806 may be structured according to a predefined output schema that ensures consistent, parseable output that can be reliably processed by downstream components without error-prone text parsing operations.

[0328] The system 1800 may comprise a response pattern evaluation 1808 that receives the structured assessment output 1806 for analysis of user engagement and learning patterns. The response pattern evaluation 1808 may be configured to distinguish between different types of low-performance responses by analyzing multiple indicators beyond simple score values. The response pattern evaluation 1808 may evaluate whether the user is making genuine attempts that reflect misunderstanding versus non-attempts that indicate disengagement, whether response patterns indicate confusion versus indifference, and whether scores are declining despite apparent effort. The response pattern evaluation 1808 may analyze response characteristics including response length, response time, vocabulary' usage, reasoning structure, and consistency with prior responses to determine the underlying cause of performance issues.

[0329] The response pattern evaluation 1808 may branch into two distinct pathways based on the analysis results. A proactive coaching trigger 1810 may represent the pathway activated when the response pattern evaluation 1808 determines that the user isATTORNEY DOCKET NO.: 38562.0001P1experiencing genuine learning struggles that may benefit from coaching intervention. The proactive coaching trigger 1810 may be activated when indicators suggest the user is making sincere attempts to engage with the simulation content but is experiencing confusion, misunderstanding, or difficult}' applying relevant concepts. A no coaching intervention 1812 may represent the pathway activated when the response pattern evaluation 1808 determines that the user is disengaged rather than struggling, indicating that proactive coaching may be counterproductive. The no coaching intervention 1812 may be activated when indicators suggest the user is providing dismissive or minimal-effort responses that do not reflect genuine learning attempts.

[0330] The system 1800 may comprise trigger conditions 1814 that define the specific criteria evaluated to determine when proactive coaching should be initiated. The trigger conditions 1814 may comprise a struggle indicator being set to true in the structured assessment output 1806, consecutive assessment scores falling below a defined threshold, identified misconceptions recurring across a plurality of user turns associated with the shared simulation session, engagement pattern changes such as significantly reduced response length or incre...

Claims

1. ATTORNEY DOCKET NO.: 38562.0001P1CLAIMSWhat is claimed is:

1. A method comprising:receiving, by a computing device, via a user interface, user input associated with a simulation session;generating, by one or more artificial intelligence (Al) modules, based on the user input, a structured assessment output;generating, by the one or more Al modules, based on the structured assessment output, adaptive narrative content for the simulation session;causing the adaptive narrative content to output via the user interface; andcausing, based on the output of the adaptive narrative content, the simulation session to update.

2. The method of claim 1, wherein the user input comprises data indicative of one or more of a user decision, a user action, or a textual response.

3. The method of claim 1, wherein the user input comprises a user interaction with the simulation session.

4. The method of claim 1, wherein the structured assessment output comprises evaluation data comprising one or more competency scores across one or more competency dimensions.

5. The method of claim 4, further comprising:maintaining a persistent learning profile associated with a user of the user input, wherein the persistent learning profile comprises cumulative competency data aggregated across one or more simulation sessions; andupdating, based on the one or more competency scores, the persistent learning profile.

6. The method of claim 5, further comprising:determining, based on the persistent learning profile, one or more competency gapsATTORNEY DOCKET NO.: 38562.0001P1associated with the user; andadjusting, based on the one or more competency gaps, one or more parameters of a subsequent simulation session.

7. The method of claim 6, wherein adjusting the one or more parameters comprises one or more of modifying a difficulty level, reordering a sequence of simulation modules, or inserting a supplementary simulation associated with one or more of: one or more competency dimensions associated with the one or more competency gaps, or real-time narrative adaptation within an active simulation associated with the one or more competency gaps.

8. The method of claim 4, wherein the structured assessment output further comprises a relevance weight for each competency dimension of the one or more competency dimensions, wherein generating the structured assessment output comprises:determining the one or more relevance weights satisfy a threshold; andgenerating, based on the one or more relevance weights satisfying the threshold, the one or more competency scores for the one or more competency dimensions.

9. The method of claim 8, wherein competency scores are not generated for competency dimensions associated with relevance weights that do not satisfy the threshold.

10. The method of claim 1, wherein the one or more Al modules comprise a single Al module.

11. The method of claim 1, wherein the one or more Al modules comprises a first Al module and a second Al module,wherein generating, by the one or more Al modules, based on the user input, the structured assessment output comprises generating, by the first Al module, based on the user input, the structured assessment output, andwherein generating, by the one or more Al modules, based on the structured assessment output, the adaptive narrative content for the simulation session comprisesATTORNEY DOCKET NO.: 38562.0001P1generating, by the second AT module, based on the structured assessment output, the adaptive narrative content for the simulation session.

12. The method of claim 1, wherein the adaptive narrative content comprises one or more of dialogue, scene descriptions, or decision options that reflect the evaluation data.

13. The method of claim 1, wherein the simulation session comprises a shared simulation session accessible by a plurality of concurrent user sessions, wherein each concurrent user session of the plurality of concurrent user sessions is associated with data indicative of a character role within the shared simulation session, wherein the method further comprises:receiving a second user input associated with a second concurrent user session, of the plurality of concurrent user sessions, associated with a second character role; and generating, by an Al module of the one or more Al modules, second adaptive narrative content for the second concurrent user session, wherein the second adaptive narrative content is associated with the second character role and an information state associated with the second character role.

14. The method of claim 1, wherein the adaptive narrative content comprises one or more of text, audio, one or more images, video, an interactive three-dimensional environment, an interactive two-dimensional environment, a virtual reality environment, or an augmented reality environment, and wherein an output modality of the adaptive narrative content is selected based on one or more of a device capability associated with a user device associated with the user input or a deployment context of the simulation session.

15. The method of claim 1, further comprising:maintaining, by a first Al module of the one or more Al modules, a dynamic skills transcript associated with a user of the user input, wherein the dynamic skills transcript comprises a persistent profile comprising one or more of simulation session assessment data or non-simulation session assessment data ; outputting, by the first Al model, data from the dynamic skills transcript to a second Al module of the one or more Al modules; andATTORNEY DOCKET NO.: 38562.0001P1updating, by the second AT module, based on the data from the dynamic skills transcript and based on a cumulative profde of a user of the user input derived from one or more prior simulation sessions, the adaptive narrative content.

16. A method compri sing :generating, by a computing device, a shared simulation session comprising a plurality of concurrent user sessions, wherein each concurrent user session of the plurality of concurrent user sessions is associated with data indicative of a character role within the shared simulation session;receiving, from a user device associated with each concurrent user session of the plurality of concurrent user sessions, user input data;generating, based on the user input data received from each concurrent user session, an assessment output associated with each concurrent user session;updating, based on the plurality of assessment outputs, the shared simulation session; and generating, based on updating the shared simulation session and based on each character role of the plurality of concurrent user sessions, character content for each concurrent user session.

17. The method of claim 16, wherein the shared simulation session is accessible by the plurality of concurrent user sessions.

18. The method of claim 16, wherein the data indicative of each character role comprises data defining character-specific knowledge and one or more available actions, wherein the data indicative of each character role defines a subset of the shared simulation session according to a character of a corresponding character role of the plurality of character roles.

19. The method of claim 16, wherein each assessment output of the plurality of assessment outputs comprises evaluation data and one or more narrative adjustment signals.

20. The method of claim 16, wherein updating, based on the plurality of assessment outputs, the shared simulation session comprises updating the shared simulation session based one orATTORNEY DOCKET NO.: 38562.0001P1more of: resolving conflicts between mutually exclusive user decisions; detecting interaction effects of a user decision that creates consequences for other users; and evaluating collaborative reasoning quality.

21. The method of claim 16, wherein the character content for each concurrent user session comprises one or more of character-specific dialogue and scene descriptions, character-appropriate decision options, and feedback integrated into the shared simulation session as incharacter reactions from non-player characters.

22. The method of claim 16, further comprising:displaying, by the plurality of user devices, via a browser-based immersive environment comprising a spherical projection of an image, the character content for each concurrent user session within the shared simulation session,wherein the image comprises an equirectangular image, a cubemap projection, or a cylindrical panorama,wherein the browser-based immersive environment comprises a plurality of interaction points comprising content and availability based on the assessment output, and wherein the plurality of interaction points are configured to cause the shared simulation session to adapt to the plurality of character roles and a plurality of competency levels associated with a plurality of users associated with the plurality of concurrent user sessions.

23. The method of claim 16, further comprising:determining, based on the user input data, a plurality of relevance weights for a plurality of competency dimensions, wherein the plurality of competency dimensions comprise one or more of one or more knowledge dimensions, one or more theme dimensions, or one or more skill dimensions; andgenerating, based on one or more relevance weights of the plurality of relevance weights satisfying a threshold, one or more competency scores associated with one or more competency dimensions, of the plurality of competency dimensions, associated with the one or more relevance weights.ATTORNEY DOCKET NO.: 38562.0001P124. The method of claim 16, further comprising:generating, based on updating the shared simulation session, competency data for a plurality of users associated with the plurality of concurrent user sessions; and storing the competency data for each user of the plurality of users in a corresponding persistent learning profde.

25. A method comprising:determining, by a computing device, a persistent learning profile comprising a plurality of competency dimensions;receiving a user response associated with a shared simulation session;generating, by an artificial intelligence (Al) module, based on the user response and data indicative of the shared simulation session, a structured assessment output comprising a relevance weight for each competency dimension of the plurality of competency dimensions;generating, based on the structured assessment output, a plurality of competency scores for the plurality of competency dimensions; andcausing, based on the plurality of competency scores, the persistent learning profile to update the plurality of competency dimensions.

26. The method of claim 25, wherein the plurality of competency dimensions comprise one or more of one or more knowledge dimensions, one or more theme dimensions, or one or more skill dimensions.

27. The method of claim 25, wherein the structured assessment output further comprises one or more of an overall numerical score, per-dimension scores, category-level relevance weights, per-dimension applicability flags, justification text, or a commentary field.ATTORNEY DOCKET NO.: 38562.0001P128. The method of claim 25, wherein the persistent learning profile maintains, for each competency dimension of the plurality of competency dimensions, a cumulative score derived from user turns, wherein the corresponding competency dimension was determined to be relevant.

29. The method of claim 25, further comprising: receiving, from a broader course associated with the shared simulation session, curriculum context data comprising prior unit content, previously covered concepts and skills, and course-wide learning objectives.

30. The method of claim 29, wherein generating, by the Al module, based on the user response and data indicative of the shared simulation session, the structured assessment output comprises generating, by the Al module, based on the user response and data indicative of the shared simulation session and based on the curriculum context data, the structured assessment output.

31. The method of claim 25, further comprising:determining, based on the persistent learning profile, one or more additional competency scores associated with one or more additional shared simulation sessions; aggregating the one or more additional competency scores; andadjusting, based on the aggregated one or more competency scores, one or more curriculum elements associated with the shared simulation session.

32. The method of claim 31, wherein adjusting the one or more curriculum elements associated with the shared simulation session comprises one or more of reordering a sequence of simulation modules to address identified competency gaps, modifying emphasis within instructional materials based on identified weaknesses, inserting supplementary simulation episodes targeting specific competency dimensions, or generating customized assignments targeting competency dimensions identified as weak points in the persistent learning profile.

33. A method compri sing :ATTORNEY DOCKET NO.: 38562.0001P1receiving, by a computing device, from a user device, one or more user inputs associated with a shared simulation session, wherein the shared simulation session is output to the user device via a first communication interface;generating by a first artificial intelligence (Al) module, based on the one or more user inputs, a structured assessment output;generating, by a second Al module, based on the structured assessment output, adaptive narrative content;generating, by a third Al module, based on the structured assessment output, one or more responses;causing the user device to output the adaptive narrative content within the shared simulation session via the first communication interface; andcausing the user device to output the one or more responses via a second communication interface.

34. The method of claim 33, wherein the structured assessment output comprises one or more of one or more competency scores, one or more relevance weights, or one or more user struggle indicators.

35. The method of claim 33, wherein the third Al module receives the structured assessment output via read-only access.

36. The method of claim 33, wherein the third Al module generates the one or more responses across a plurality of user turns.

37. The method of claim 33, wherein the third Al module receives anonymized user personally identifiable information, wherein the user personally identifiable information is anonymized via a hashing function.

38. The method of claim 33, wherein generating, by the third Al module, based on the structured assessment output, the one or more responses comprises generating, by the third AlATTORNEY DOCKET NO.: 38562.0001P1module, based on the structured assessment output satisfying one or more conditions, the one or more responses.

39. The method of claim 38, wherein the one or more conditions comprise one or more of a struggle indicator being set to true, consecutive assessment scores falling below a defined threshold, identified misconceptions recurring across a plurality of user turns, or engagement pattern changes.

40. The method of claim 33, wherein generating, by the third Al module, based on the structured assessment output, the one or more responses comprises generating, by the third Al module, based on the structured assessment output and based on a persistent learning profile, the one or more responses.

41. The method of claim 33, wherein the one or more responses are generated according to one or more of tone, vocabulary complexity, or support level based on user learning accommodations.

42. A method comprising:generating, by a computing device, an image associated with a shared simulation session; displaying the image according to an interior surface of a virtual sphere within a browserbased three-dimensional (3D) viewer, wherein a virtual camera is positioned at a center of the virtual sphere to create a panoramic viewing experience; receiving an assessment output comprising competency evaluation data for a user; configuring, based on the assessment output, a plurality of interaction points within the shared simulation session; anddisplaying the configured plurality of interaction points within the shared simulation session for user interaction.

43. The method of claim 42, wherein the image comprises an equi rectangular image, a cubemap projection, or a cylindrical panorama.ATTORNEY DOCKET NO.: 38562.0001P144. The method of claim 42, wherein the image is generated by an artificial intelligence (Al) module based on simulation context parameters comprising one or more of historical period, geographic location, time of day, weather conditions, narrative state, or emotional tone.

45. The method of claim 42, further comprising displaying character visual assets as stylized two-dimensional billboard elements within the shared simulation session, wherein the billboard elements comprise an orientation facing the virtual camera.

46. The method of claim 42, wherein the plurality of interaction points are organized as templates defining interaction mechanics.

47. The method of claim 42, wherein an artificial intelligence (Al) module determines specific content inserted into each interaction point based on narrative state information and user assessment performance.

48. The method of claim 42, wherein each interaction point of the plurality of interaction points is associated with a character role associated with the user, wherein each character role is configured to cause the corresponding interaction point to display character-specific visual elements and information overlays to the corresponding user.

49. The method of claim 42, wherein configuring the plurality of interaction points within the shared simulation session comprises determining content and availability of each interaction point of the plurality of interaction points based on competency gaps or demonstrated mastery areas associated with the user.

50. The method of claim 42, wherein configuring the plurality of interaction points comprises emphasizing interaction points targeting competency dimensions identified as weak in the assessment output and surfacing more complex decision opportunities when mastery is demonstrated.

Images