Real-time network online education trainee management information method and system

Abstract

The invention relates to the technical field of educational training management, and discloses a real-time network online educational training personnel information management method and system, and the method comprises the steps: obtaining behavior telemetering data of a mouse, a keyboard and the like generated when a student interacts with a learning interface; a preset deep learning model is adopted to process the data into student state vectors representing dimensions such as concentration degree and confusion degree in real time; and triggering a management intervention based on the vector. In a preferable scheme, through generating a content cognitive fingerprint, the student state is compared with a standard cognitive mode of teaching content, and cognitive deviation is calculated to realize accurate tutoring. In another preferred scheme, the cognitive coordination degree of the cooperative group is evaluated and guided by calculating a group cognitive synchronization index. According to the method, the intrinsic cognitive state of the student is dominated, multi-dimensional, real-time and predictive management from individuals to groups is realized, and the intervention efficiency and teaching quality of online education are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of education and training management technology, specifically to a real-time online education and training personnel management information method and system. Background Technology

[0002] In recent years, distance learning models centered on online education platforms have experienced rapid development. They have greatly alleviated the limitations of time and space in educational resources, providing learners with unprecedented flexibility and convenience. Existing online education management systems possess mature functions in course distribution, progress tracking, and outcome assessment, capable of recording a series of objective events such as student login time, video viewing completion rate, assignment submission status, and final test scores. This data provides a quantifiable foundation for teaching management.

[0003] However, beneath this apparent prosperity, existing technologies generally suffer from a fundamental limitation in their ability to gain insights into the depth and breadth of the learning process. The data these systems rely on is essentially discrete, high-latency "outcome-oriented" data; it can answer "whether the learner completed the task," but it cannot reveal "how the learner completed the task." The learner's actual cognitive load during the learning process, the subtle fluctuations in attention, the confusion when encountering obstacles to understanding, and the intellectual pleasure at the moment of sudden enlightenment—this series of intrinsic processes that constitute the core of the learning experience—is a completely unobservable "black box" for traditional systems.

[0004] This lack of observational ability directly leads to superficial and delayed management intervention. Managers or systems often only realize a problem after a learner submits an incorrect answer, fails a test, or actively sends a request for help. By then, valuable learning opportunities have often been missed, and intervention becomes more of a post-hoc remedy than real-time guidance during the process. Furthermore, this limitation is amplified in collaborative learning scenarios. While systems can record who speaks and when in a group discussion, they cannot perceive deeper collaborative dynamics such as whether the team's understanding is synchronized, whether the discussion is focused, or whether "social laziness" exists among members, rendering the management of the online collaborative process virtually nonexistent.

[0005] Therefore, the core dilemma of current technology lies in its failure to effectively establish a mapping relationship between the massive, low-level interactive behaviors between learners and the learning interface (such as mouse hovering and pausing, keyboard input and deletion) and their high-level, intrinsic cognitive states. This leaves administrators with vast amounts of data yet still unable to truly "understand" the learning process, hindering genuine personalized instruction and precise intervention. Summary of the Invention

[0006] The technical problem that this invention aims to solve is that existing online education and training management technologies mainly rely on discrete recording of students' operational events, which cannot provide real-time and in-depth insights into students' learning processes and cognitive states, resulting in superficial, delayed, and untargeted management interventions.

[0007] To address the aforementioned technical problems, this invention provides an online education management method and system based on behavioral telemetry and cognitive state inference. This method can quantify and make explicit the invisible cognitive states of students, and conduct collaborative analysis in conjunction with teaching content and group dynamics, thereby achieving refined, real-time, and predictive management intervention.

[0008] The first aspect of this invention provides an online education management method based on behavioral telemetry and cognitive state inference, the method comprising:

[0009] Acquire behavioral telemetry data generated by students during their interaction with the online education learning interface; the behavioral telemetry data is non-semantic data characterizing the interaction between students and the underlying learning interface, and in some embodiments, it may specifically include at least one of the following:

[0010] Mouse dynamic data: such as mouse pointer screen coordinates, movement speed, acceleration, trajectory entropy value, click frequency, and scroll wheel behavior;

[0011] Keyboard interaction rhythm data: such as the time interval between keys, the duration of key presses, and the frequency of use of the delete / backspace keys;

[0012] Window interaction data: such as the focus state and focus switching frequency of the learning window.

[0013] Then, based on the behavioral telemetry data, a student state vector representing the student's cognitive state is generated in real time using a pre-defined cognitive state inference model. The cognitive state inference model is preferably a pre-trained deep learning sequence model, such as a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU), to effectively capture temporal dependencies in the behavioral data.

[0014] The student state vector is a standardized multidimensional vector, denoted as L. t Its generation process can be expressed by the following formula:

[0015] L t =M cog (S t );

[0016] Among them, S t M represents the sequence of behavioral telemetry data within a time window preceding time t. cog This represents the cognitive state inference model.

[0017] In some embodiments, the student state vector L t The dimensions can specifically include at least one of the following to comprehensively characterize the learner's cognitive and emotional state:

[0018] Focus d focus (t): Represents the learner's level of concentration on the current learning task.

[0019] Confusion level d conf (t): Represents the degree of obstacle encountered by learners in understanding the learning content.

[0020] Cognitive load d load (t): Represents the amount of psychological resources required by the learner to process the current information.

[0021] Emotional valence d val (t): Represents the positive or negative aspect of the learner's current learning emotions.

[0022] Finally, based on the student's state vector, a preset management intervention is triggered. This management intervention is a response action automatically executed according to the student's real-time state. In some embodiments, it may specifically include at least one of the following: pushing personalized learning materials to the student, sending prompts to the student or teacher, or adjusting the presentation of teaching content.

[0023] In a preferred embodiment, the method further includes the step of generating and comparing a content cognitive fingerprint. First, for a preset teaching content, a content cognitive fingerprint (CCF) is generated based on the student state vectors generated by a reference student group when learning the teaching content. k (τ) represents the standard learner's performance on a specific teaching content C. k The expected pattern of how cognitive states change with the normalization progress τ. Its generation process can be expressed by the following equation:

[0024] CCF k (τ)=F agg ({L i (τ)|i∈G ref});

[0025] Among them, G ref For reference student groups, L i (τ) is the student state vector of student i in the group at progress τ, F agg For statistical aggregation operators.

[0026] Subsequently, the student state vector L of the current student j is calculated in real time. j (t) and the content cognitive fingerprint CCF k Cognitive bias Δ of (τ(t)) at the corresponding learning progress τ(t) cog,j,k(t). Accordingly, the steps to trigger management interventions can be further based on this cognitive bias to achieve precise management interventions related to the specific progress of the teaching content.

[0027] In another preferred embodiment, when this method is applied to a collaborative learning scenario involving groups of learners, it further includes a step of analyzing group cognitive synchronicity. First, the learner state vectors of each member within the group are acquired in real time. Then, based on these learner state vectors, a group cognitive synchronicity index characterizing the degree of cognitive coordination within the group is calculated. The group cognitive synchronicity index is GCS. t (G) can be obtained through a synchronization analysis function F sync This is obtained by processing the set of student state vectors for all members within group G. For example, one dimension can be calculated as the variance of a specific cognitive dimension (such as focus) among the student state vectors of all members within the group. Accordingly, the steps to trigger management intervention can be further based on this group cognitive synchronicity index to achieve macro-level guidance and management intervention for the student group.

[0028] A second aspect of this invention provides an online education management system based on behavioral telemetry and cognitive state inference, the system comprising:

[0029] The data acquisition module is used to perform the data acquisition steps in the above method to acquire behavioral telemetry data generated by students when interacting with the online education learning interface;

[0030] The state inference module is used to generate a student state vector representing the student's cognitive state in real time based on the behavioral telemetry data and a preset cognitive state inference model.

[0031] The intervention execution module is used to trigger preset management interventions based on the student's state vector.

[0032] This invention establishes a mapping relationship between students' underlying interactive behaviors and their higher-level cognitive states through the above-mentioned technical solutions, and further introduces the collaborative analysis dimensions of content cognitive fingerprints and group cognitive synchronicity, thereby achieving in-depth, dynamic and multi-level insights into the learning process and significantly improving the real-time, accuracy and intelligence level of online education management.

[0033] This invention provides a method and system for managing information on online education and training personnel in real time. It has the following beneficial effects:

[0034] 1. This invention enables a deep understanding of the learner's learning process, completely changing the previous situation where management was limited to superficial data such as login time and video playback completion. By collecting non-semantic behavioral telemetry data such as mouse trajectories and keyboard rhythms, and transforming them into multi-dimensional learner state vectors containing attention and confusion levels, this invention, for the first time, makes explicit and quantifiable the learner's invisible and dynamically changing internal cognitive state during the learning process, providing managers with an unprecedented window into the "black box" of learning.

[0035] 2. This invention elevates management intervention from the traditional "post-event remediation" to a completely new level of "real-time prediction." Because it can capture the precise moments when a student's confusion increases or concentration decreases, the system no longer needs to wait for the student to submit incorrect answers or fail the test before reacting. It can trigger precise, personalized interventions based on the student's specific cognitive state during the golden window of opportunity when they encounter difficulties, much like providing each student with a perceptive personal tutor, greatly improving the effectiveness and timeliness of the intervention.

[0036] 3. By introducing the innovative concept of content cognitive fingerprinting, this invention endows the system with the ability to understand the cognitive difficulty of the teaching content itself. Instead of judging whether a student is confused in isolation, the system compares their cognitive state with the "standard cognitive path." This analysis based on cognitive bias can effectively distinguish between learning difficulties caused by individual student factors and those due to universal difficulties inherent in the teaching content itself. This not only provides a more accurate basis for personalized tutoring but also accumulates valuable data assets for instructional designers to optimize and iterate course content.

[0037] 4. This invention effectively solves the problem of evaluating and managing online collaborative learning processes. By calculating group cognitive synchronicity indicators, managers can quantify the cognitive coordination level of a learning group in real time, intuitively determining whether the team is in a highly synchronized "resonance" state or a "out-of-focus" state where members are working independently. This quantitative assessment of group dynamics provides solid technical support for process guidance, problem warning, and effect evaluation of online collaborative tasks, filling a key technological gap in this field.

[0038] 5. This invention constructs a multi-dimensional collaborative management framework encompassing individuals, content, and groups, providing education administrators with a panoramic decision-making view. Administrators can flexibly drill down to observe the cognitive fluctuation curves of individual students; they can make horizontal comparisons to analyze the impact of specific content on different students; and they can also gain a macro-level overview to grasp the overall learning atmosphere and cognitive pace of a class or group. This multi-layered, three-dimensional information integration enables more scientific and efficient management decisions based on comprehensive and dynamic data. Attached Figure Description

[0039] Figure 1 This is a system architecture block diagram of the present invention;

[0040] Figure 2 Schematic diagram of the method of the present invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not constitute any limitation thereof.

[0042] See attached document Figure 1 This figure illustrates the functional module structure of an online education management system in one embodiment of the present invention. The present invention provides an online education management system based on behavioral telemetry and cognitive state inference. At the hardware level, this system may include a learning interface deployed on the student client, a backend processing server, a database, and a management interface used by teachers or administrators. The backend processing server logically integrates the core functional units of the present invention, specifically including: a state inference unit, a content fingerprint analysis unit, a group synchronization analysis unit, and an intervention decision unit.

[0043] See attached document Figure 2 The figure illustrates a flowchart of an online education management method according to one embodiment of the present invention. The method may include the following steps:

[0044] S1: The client's data acquisition module acquires real-time telemetry data of student behavior generated during interaction on the learning interface and sends the data to the backend processing server.

[0045] S2. The state inference unit in the backend processing server receives behavioral telemetry data and generates a student state vector that represents the individual cognitive state of the student.

[0046] S3, the content fingerprint analysis unit and the group synchronicity analysis unit receive the student state vector and, in combination with the preset teaching content model and student group information, generate cognitive bias and group cognitive synchronicity indicators respectively.

[0047] S4. The intervention decision-making unit integrates the trainees' state vectors, cognitive biases, and group cognitive synchronicity indicators to form specific management intervention instructions.

[0048] S5. Management intervention instructions are sent to the student's client or management interface and executed, completing one management loop.

[0049] Combined with appendix Figure 1 and attached Figure 2The overall workflow of this invention is described below. In a typical application, when a student logs into the learning interface and begins interacting with the teaching content, the data acquisition module embedded in the front end starts working. This module continuously captures the student's underlying interactive behavior, forming a continuous stream of behavioral telemetry data, and sends it to the backend processing server via a secure network connection.

[0050] Once the data arrives at the backend, it is first processed by the state inference unit. This unit is the starting point of the entire analysis chain, and it uses a pre-trained cognitive state inference model M. cog The raw and irregular behavioral telemetry data sequence S t Transformed into a structured, standardized student state vector L t This process can be described by the following formula:

[0051] L t =M cog (S t );

[0052] Among them, S t L represents the sequence of behavioral telemetry data within a time window preceding time t. t This is the generated student state vector, which contains quantitative values ​​of cognitive state in multiple dimensions such as focus and confusion.

[0053] The student state vector L generated by the state inference unit t It is not only a preliminary result describing the individual's state, but also the foundational input for all subsequent advanced analyses. It is distributed to other analytical units within the system to initiate deeper, collaborative analyses.

[0054] The content fingerprint analysis unit receives L t Then, it will retrieve the pre-stored Content Cognition Fingerprint (CCF) of the current teaching content from the database. k (τ), and calculates the cognitive bias Δ between the student's current learning progress τ(t) and the standard mode in real time. cog,j,k (t). Meanwhile, if a student is in a collaborative learning group G, the group synchronization analysis unit will aggregate the student state vectors {L} of all members within that group. i (t)|i∈G}, and calculate the group cognitive synchronicity index (GCS) that characterizes the current cognitive coordination of the group. t (G).

[0055] Ultimately, all these analytical results from different dimensions represent the immediate individual state L. t Cognitive bias based on content interaction Δ cog,j,k (t) and the synchronicity index GCS reflecting group dynamics t(G) – This information is transmitted to the intervention decision-making unit. This unit acts as the brain of the system, and based on a set of preset decision-making logic, it comprehensively evaluates this information and generates one or more specific, context-aware management intervention instructions.

[0056] These instructions are ultimately transmitted back to the corresponding execution endpoints. For example, a micro-adjustment instruction for an individual student is sent to the student's learning interface, presented as a floating notification or highlighted content; while a macro-level alert regarding group learning efficiency is pushed to the teacher's management interface. Simultaneously, all key data generated throughout the process—including raw data, student status vector time series, and generated analytical indicators—is stored in a database for subsequent teaching research, model iteration, and record-keeping queries, thus forming a data-driven, continuously optimizing intelligent management ecosystem.

[0057] In a specific embodiment of the present invention, the core method flow is described below. This flow is the foundation for the system to achieve intelligent management, systematically transforming students' unconscious and fragmented interactive behaviors into management guidelines with clear guidance.

[0058] First, in the behavioral telemetry data acquisition and preprocessing stage, the system accomplishes this through a lightweight script (e.g., JavaScript) deployed on the learner's front-end interface. This script captures the learner's underlying interaction with the interface at a preset frequency (e.g., several times per second) and quantifies these behaviors into a series of specific data metrics. These metrics are carefully selected to retain, to the greatest extent possible, cues reflecting cognitive states. Specifically, they may include: mouse trajectory coordinates, movement speed, and acceleration to characterize the learner's attention focus and exploratory behavior; mouse trajectory entropy to quantify the learner's hesitation or disorientation; and keyboard key press intervals, duration of keystrokes, and frequency of delete key usage to reflect the learner's operational rhythm and corrective behavior.

[0059] Before being sent to the backend, or after arriving at the backend, the collected raw behavioral data undergoes a necessary preprocessing process. This process includes data cleaning to remove abnormal or invalid data points; data normalization to map index values ​​of different dimensions to the same numerical range (e.g., [0,1]) to eliminate the impact of dimensional differences on subsequent model analysis; and time window segmentation to divide the continuous data stream into sequences S with a fixed length W. t Each sequence contains behavioral data from the past W time steps, serving as a complete input to the cognitive state inference model.

[0060] Next, in the deep learning-based individual cognitive state inference stage, the preprocessed behavioral data sequence S tThe data is fed into a state inference unit deployed on a backend server. In this embodiment, the core of this unit is a pre-trained deep learning sequence model M capable of capturing temporal information. cog For example, a Long Short-Term Memory (LSTM) network. This LSTM model consists of an input layer, one or more hidden layers, and an output layer. Its input layer receives a sequence S with the same dimension as the behavioral data vector. t It learns and memorizes the dependencies and long-term patterns of behavioral data over time through its internal gating mechanisms (input gate, forget gate, output gate).

[0061] After processing the entire sequence, the hidden state of the model is passed to a fully connected output layer. This output layer maps the high-dimensional hidden state to a low-dimensional, normalized student state vector L. t Each dimension of this vector corresponds to a specific cognitive indicator, and its value is ensured to fall within the [0,1] interval by an activation function (such as the Sigmoid function), facilitating subsequent comparison and analysis. Therefore, this model completes the crucial transformation from the complex and varied behavioral representations of the physical world to a concise and clear state description at the psychological level.

[0062] Finally, in the basic management intervention phase based on individual status, the intervention decision-making unit receives the real-time generated trainee status vector L. t And make decisions based on a set of pre-set basic intervention rules. These rules will L t Different dimensions are associated with specific intervention actions. For example, a specific rule could be set as follows: if the learner's confusion dimension d is detected... conf The value of (t) persists for a preset time length T. duration Within this range, all values ​​are above a configurable perplexity threshold θ. conf If so, the system determines that the student has encountered a persistent learning obstacle.

[0063] Based on this assessment, the intervention decision-making unit will immediately trigger a pre-defined management intervention action. For example, the system can automatically retrieve the most relevant supplementary explanations or simplified examples from the knowledge base and push them to the learner in the form of a non-intrusive pop-up or sidebar. This intervention based on a single learner's state vector constitutes the most direct and fundamental management loop in this invention, ensuring rapid response and support for individual learning difficulties.

[0064] In a preferred embodiment of the present invention, to address the problem that judging solely based on individual states cannot distinguish between individual student problems and general content difficulties, the present invention further introduces an analysis and intervention mechanism that combines content cognitive fingerprinting. This mechanism correlates the dynamic cognitive state of students with the static cognitive attributes of the teaching content, thereby achieving a deeper and more accurate management insight.

[0065] In this embodiment, the system first performs an offline fingerprint generation process for each individual piece of teaching content (e.g., a video, a document chapter) in the database. For a specific piece of teaching content C... k The system will select a statistically representative reference student group G. ref The system collects the complete learner state vector (LSV) time series of all learners in the group while learning this content. To eliminate individual differences in learning pace, the system maps each learner's physical learning time t to a normalized content progress τ, where τ∈[0,1].

[0066] Subsequently, the content fingerprint analysis unit employs a statistical aggregation operator F. agg (For example, piecewise mean or Gaussian smoothing) The state vectors of all reference learners at the same content progress point τ are processed to extract a trajectory that can represent the standard cognitive response, namely the content cognitive fingerprint (CCF). k (τ). Its generation process can be expressed by the following formula:

[0067] CCF k (τ)=F agg ({L i (τ)|i∈G ref});

[0068] Among them, CCF k (τ) is the cognitive fingerprint vector of content k at progress point τ, L i (τ) is the reference group G ref The student state vector of student i at progress point τ. The generated CCF. k (τ) is stored in the database as an inherent attribute of the teaching content. It describes the general pattern of how the concentration, confusion and other states of ordinary learners change as the content progresses.

[0069] When a new student J begins learning this course content C k At this point, the system enters the real-time online comparison phase. At each moment t during a student's learning process, the system generates their instantaneous student state vector L. j While tracking the learning progress τ(t), the system will also retrieve the corresponding standard cognitive fingerprint (CCF) from the database. k(τ(t)). Next, the system calculates the cognitive bias Δ between the two individuals using a distance metric function D (e.g., Euclidean distance or Mahalanobis distance). cog,j,k (t).

[0070] Δ cog,j,k (t)=D(L j (t),CCF k (τ(t)));

[0071] This cognitive bias Δ cog,j,k (t) is a quantitative value that precisely reflects the degree to which learner j's learning experience deviates from the standard model at a specific progress point τ(t) of content k.

[0072] Based on this cognitive bias, the intervention decision-making unit can perform judgments and operations that are far more precise than basic interventions. For example, when the system detects a cognitive bias Δ... cog,j,k The value of (t) exceeds a preset deviation threshold θ dev At that moment, the system not only knows that the student is confused, but also knows exactly where this confusion occurs within the teaching content, and that the level of confusion significantly exceeds the normal range. Based on this, the system can trigger a precise intervention strongly correlated with the progress of that content. For example, it can automatically highlight difficult sentences and paragraphs in that section, or push annotations and discussions left by other students or teachers specifically targeting that knowledge point, thereby guiding the granularity of management intervention from the macro-level state to the micro-level content tutoring.

[0073] In another preferred embodiment of the present invention, to extend the application scenario of the present invention from individual learning to team collaboration, the system introduces a group cognitive synchronization analysis mechanism for collaborative learning. This mechanism aims to provide objective and real-time decision-making basis for the process management of online collaborative learning by quantitatively assessing the degree of cognitive coordination among group members.

[0074] In this embodiment, when multiple learners are assigned to a pre-defined collaborative learning group G by the system, the group synchronization analysis unit in the backend processing server is activated. This unit collects the learner state vectors (LSVs) of all members i in group G at time t in real time, forming an instantaneous group state set {L... i (t)|i∈G}.

[0075] Subsequently, the unit employs a synchronization analysis function F. sync This set is processed to calculate the Group Cognitive Synchronicity Index (GCS), which characterizes the group's current state of cognitive coordination. t(G). This indicator itself is also a multi-dimensional vector, with each dimension corresponding to a cognitive dimension (such as focus, confusion, etc.) in the learner's state vector, and is used to quantify the consistency or dispersion of the group in the corresponding dimension. In this embodiment, the function F symc This can be achieved by calculating the statistical variance of group members across a specific cognitive dimension. For example, the synchronicity index g of a group on the dimension of attention. focus (t) can be calculated by the following formula: g focus (t)=Var({d focus,i (t)|i∈G})

[0076] Where, d foeus,i (t) is the focus component of member i in the student state vector at time t within group G. Based on this, a lower g focus The (t) value indicates that the focus of the group members is highly synchronized and the whole team is in a good state of collaborative work; conversely, a higher value means that the attention of the group members has diverged significantly, and some members are distracted or out of sync with the team rhythm.

[0077] Based on this quantitative indicator, the intervention decision-making unit can effectively assess and guide the health of group learning. Specifically, the intervention decision-making unit continuously monitors the group cognitive synchronicity indicator, GCS. t The change of (G). When the value of one of its dimensions, such as g foeus (t) exceeds a preset group divergence threshold θ symc If this happens, the system will determine that there is a potential risk to the group's collaboration efficiency.

[0078] At this point, the system will trigger a macro-level guidance and management intervention for the group. This intervention is not targeted at any individual, but rather at the entire group. For example, the system can automatically push a guiding discussion question in the group's shared discussion area, such as "Please summarize the consensus reached so far and clarify the next task assignments," to help members refocus; or, the system can also issue a warning sign requiring attention to the group on the teacher's management interface, prompting the teacher to intervene and provide manual guidance. This macro-level guidance based on group dynamics transforms online collaborative management from relying on member self-discipline and teacher oversight to a data-driven, proactive, and automated process control.

[0079] To more comprehensively illustrate how the various technical modules of this invention work together to generate practical value, this invention provides an embodiment of a comprehensive application process. In this scenario, student A is participating in an online course module that includes two phases: independent learning and group collaboration.

[0080] At the start of the course module, student A first needs to independently watch an instructional video on "Bayes' Theorem." From the moment student A clicks play, the data acquisition module deployed at the front end begins capturing their behavioral telemetry data and uploading it to the backend processing server in real time. In the first half of the video, the content is relatively basic, and student A is engaged in learning, with smooth mouse movements and few pauses. At this time, the state inference unit generates their student state vector L. A (t) indicates that its focus dimension value is high, while the confusion dimension value remains at a low level.

[0081] When the video reaches approximately 65% ​​completion (i.e., τ(t) = 0.65), the content delves into complex conditional probability calculations, which is a recognized challenge in this course. The system detects that student A's mouse begins to move aimlessly in small areas near the player's progress bar and clicks the back button twice. Based on this, the state inference unit quickly updates its student state vector L. A (t), whose perplexity component increases sharply.

[0082] At this point, the content fingerprint analysis unit is activated. It will take student A's current student state vector L. A (t), compared with the pre-stored standard content cognitive fingerprint (CCF) for this video at the progress point τ = 0.65 in the database. k (0.65) was compared. The system found that although the standard fingerprint itself has a peak in confusion at this point, student A's confusion at this moment was significantly higher than that peak, resulting in a cognitive bias Δ between the two. cog,A,k (t) exceeded the preset deviation threshold θ dev·

[0083] Upon receiving this high deviation signal, the intervention decision-making unit immediately implemented a precise management intervention. Instead of displaying a generic prompt, the system pushed an interactive component strongly correlated with the progress point τ = 0.65 next to the video player. The component read: "Many learners are confused about the calculation of 'posterior probability' here. Would you like to see a graphical breakdown of the steps for this concept?" After learner A clicked confirm, the system presented them with supplementary micro-learning material to help them overcome this learning obstacle.

[0084] After completing independent learning, student A enters the second stage, joining students B and C in a virtual discussion room to form a collaborative group G. Together, they complete a case analysis task based on Bayes' theorem. At this point, the system's group synchronization analysis unit begins working, collecting and analyzing the student state vectors of group members A, B, and C in real time.

[0085] In the initial stages of the discussion, the three members actively exchanged opinions, and their participants' state vectors all showed high values ​​and similar values ​​in the focus dimension. Therefore, the focus variance g corresponding to the group cognitive synchronicity index... focus (t) remained at a low level. However, after about 15 minutes of discussion, participant C's attention began to wander, and the focus component in their participant state vector continued to decrease. This change led to a decrease in the overall focus variance g of the group. focus (t) increases rapidly and quickly exceeds the preset population divergence threshold θ. sync°

[0086] After detecting the group's desynchronization signal, the intervention decision-making unit implemented a macro-level guidance intervention for the group. Instead of directly pointing out participant C's lack of focus, the system automatically sent a guiding message to the group's public chat channel: "The discussion seems to have hit a bottleneck. We suggest everyone take a minute to restate the current core issue, and then continue the discussion." This neutral prompt effectively brought all members' attention back to the task itself, prompting the group to restore a highly efficient collaborative state, all while remaining natural and undisturbed for the participants.

[0087] As can be seen from this integrated application process, this invention seamlessly integrates real-time inference of individual cognitive states, in-depth interactive analysis based on content fingerprints, and dynamic group assessment for team collaboration. The system can automatically switch its analysis focus and execute the most appropriate management intervention based on the learner's different learning stage (independent or collaborative) and the specific problems encountered (individual difficulties or group disorientation), thereby achieving a complete, multi-layered, and highly accurate intelligent online education management closed loop.

[0088] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for managing personnel information in real-time online education and training, characterized in that, The following steps are involved: S1. Acquire behavioral telemetry data generated by students during their interaction with the online education learning interface; S2. Based on the behavioral telemetry data, a preset cognitive state inference model is used to generate a student state vector representing the student's cognitive state in real time. S3. Based on the student's state vector, trigger a preset management intervention.

2. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, The behavioral telemetry data includes at least one of the following: mouse dynamic data, keyboard interaction rhythm data, and window interaction data.

3. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, The student state vector is a multi-dimensional vector, and its dimensions include at least one of the following: focus, confusion, cognitive load, and emotional valence.

4. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, The management interventions include at least one of the following: pushing personalized learning materials to students, sending reminder messages to students or teachers, and adjusting the presentation of teaching content.

5. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, Also includes: For the preset teaching content, based on the student state vector generated by the reference student group when learning the teaching content, a content cognitive fingerprint of the teaching content is generated. The content cognitive fingerprint represents the cognitive state change pattern of the standard students on the teaching content. The cognitive deviation between the current student's student state vector and the content cognitive fingerprint at the corresponding learning progress is calculated in real time.

6. The method for managing personnel information in real-time online education and training according to claim 5, characterized in that, The step of triggering a preset management intervention based on the student's state vector further includes: Based on the aforementioned cognitive biases, precise management interventions related to the specific progress of the teaching content are triggered.

7. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, When the method is applied to a collaborative learning scenario involving groups of students, it further includes: The student state vector of each member in the student group is obtained in real time; Based on the student state vectors of each member in the student group, a group cognitive synchronicity index, which characterizes the degree of cognitive coordination in the group, is calculated.

8. The method for managing personnel information in real-time online education and training according to claim 7, characterized in that, The step of triggering a preset management intervention based on the student's state vector further includes: Based on the aforementioned group cognitive synchronicity index, macro-level guidance and management interventions are triggered for the student group.

9. The method for managing personnel information in real-time online education and training according to claim 1, characterized in that, The cognitive state inference model is a pre-trained deep learning sequence model.

10. A real-time online education and training personnel management information system, used to perform the method as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire behavioral telemetry data generated by students when interacting with the online education learning interface; The state inference module is configured to generate a student state vector representing the student's cognitive state in real time based on the behavioral telemetry data and a preset cognitive state inference model. The intervention execution module is used to trigger preset management interventions based on the student's state vector.

Images