Adversarial Hardening of Problems for Automated Responses

Abstract

A system, computer program product, and method for conducting a test with adversarial challenge of a problem for automated response are presented. The method includes electronically receiving an original problem. A response to the original problem is to be submitted electronically by a human. The method also includes generating a modified problem by modifying the original problem. The modified problem is configured to be understandable by a human and not properly answered via electronic means without human support.

Description

[Technical Field]

[0001] The present disclosure relates to conducting tests, and more particularly to improving adversarial hardening of problems for automated responses. [Background technology]

[0002] Many known testing events are conducted online, where test takers are remotely located and in-person proctoring is impractical. As the internet continues to grow in presence in our society, access to information becomes easier and the internet serves as the first destination for questions. For example, answers to questions can be quite conveniently obtained through well-known and nearly ubiquitous artificial intelligence (AI) agents, such as digital assistants and chatbots, which are easily accessed through all forms of computing devices, including mobile devices. Therefore, some form of remote proctoring is necessary to maintain the integrity of the testing process, given that test takers may utilize such technologies to generate automated responses to test questions. At least some known methods of remotely administering and proctoring online exams involve invasive monitoring, such as video-based monitoring of test takers, and associated screen and screen sharing. Both methods require the full, focused, and continuous attention of a human proctor. Additionally, both methods use additional computer resources that can impact the performance of test hardware and software, which can be a burden for time-sensitive tests. Furthermore, both methods increase the cost of testing through the use of human proctors and additional hardware and software. Summary of the Invention

[0003] Systems, computer program products, and methods for conducting tests involving adversarial hardening of problems against automated responses are provided.

[0004] In one aspect, a computer system for conducting a test involving adversarial challenge of a problem for automated response is provided. The system includes one or more processing devices and at least one memory device operatively coupled to the one or more processing devices. The one or more processing devices are configured to electronically receive an original problem. A response to the original problem is to be submitted electronically by a human. Additionally, the one or more processing devices are configured to generate modified problems by modifying the original problem. The modified problems are configured to be understandable by a human and not to be properly responded to via electronic means without human assistance.

[0005] In another aspect, a computer program product for conducting a test involving adversarial challenge of a problem against an automated response is provided. The computer program product includes one or more computer-readable storage media and program instructions co-stored on the one or more computer storage media. The product also includes program instructions for electronically receiving an original problem. A response to the original problem is to be submitted electronically by a human. The product also includes program instructions for generating a modified problem by modifying the original problem. The modified problem is configured to be understandable by a human and not to be properly responded to via electronic means without human assistance.

[0006] In yet another aspect, a computer-implemented method for conducting a test with adversarial difficulty of a question for automated response is provided. The method includes electronically receiving an original question. A response to the original question is to be submitted electronically by a human. The method also includes generating a modified question by modifying the original question. The modified question is configured to be understandable by a human and not to be properly answered via electronic means without human support.

[0007] This summary is not intended to describe each aspect, every implementation, or every embodiment, or combination thereof, of the present disclosure. These and other features and advantages will become apparent from the following detailed description of the present embodiment(s), which is provided in conjunction with the accompanying drawings.

[0008] The drawings included in this application are incorporated into and form a part of this specification. The drawings illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the disclosure. The drawings illustrate particular embodiments and are not intended to limit the disclosure. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram illustrating a cloud computing environment, according to some embodiments of the present disclosure. [Figure 2] FIG. 1 is a block diagram illustrating a set of functional abstraction model layers provided by a cloud computing environment, in accordance with some embodiments of the present disclosure. [Figure 3] FIG. 1 is a block diagram illustrating a computer system / server that may be used as a cloud-based support system for implementing the processes described herein, according to some embodiments of the present disclosure. [Figure 4] FIG. 1 is a block diagram illustrating a computer system configured to conduct testing involving adversarial hardening of problems for automated responses, according to some embodiments of the present disclosure. [Figure 5A] 1 is a flowchart illustrating a process for conducting a test with adversarial hardening of a problem for an automated response, according to some embodiments of the present disclosure. [Figure 5B] FIG. 5B illustrates a continuation of the flowchart of FIG. 5A in accordance with some embodiments of the present disclosure. [Figure 5C] FIG. 5C illustrates a continuation of the flowchart of FIG. 5B in accordance with some embodiments of the present disclosure. [Figure 6]Figure 6(A) is an image diagram showing an example image associated with a test question according to some embodiments of the present disclosure. Figure 6(B) is an image diagram of Figure 6(A) showing an example of an at least partially altered image associated with an at least partially altered test question according to some embodiments of the present disclosure. Figure 6(C) is an image diagram of Figure 6(A) showing an example of a further altered image associated with a further altered test question according to some embodiments of the present disclosure. Figure 6(D) is an image diagram of Figure 6(A) showing an example of an altered image associated with an altered test question according to some embodiments of the present disclosure. [Figure 7] Figure 7(A) is a diagram illustrating a text representation of an example text test question according to some embodiments of the present disclosure. Figure 7(B) is a block diagram illustrating an example audio test question based on the text test question of Figure 7(A) according to some embodiments of the present disclosure. Figure 7(C) is a block diagram illustrating an example audio test question based on the text test question of Figure 7(A) that has been at least partially modified according to some embodiments of the present disclosure. Figure 7(D) is a block diagram illustrating an example audio test question based on the text test question of Figure 7(A) to which an audio noise signal has been added according to some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0010] While the present disclosure is susceptible to various modifications and alternative forms, specifics thereof have been shown by way of example in the drawings and will be described in detail. It should be understood, however, that the intention is not to limit the disclosure to the particular embodiments described. On the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure.

[0011] It will be readily understood that the components of the present embodiments, as generally described herein and illustrated in the Figures, could be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the present apparatus, system, method, and computer program product embodiments, as presented in the Figures, is not intended to limit the scope of the present embodiments as claimed, but merely represents selected embodiments. In addition, it should be understood that, while specific embodiments have been described herein for purposes of illustration, various modifications may be made without departing from the scope of those embodiments.

[0012] References throughout this specification to "selected embodiments," "at least one embodiment," "one embodiment," "another embodiment," "other embodiments," or "an embodiment," and similar language, mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment. Thus, the appearances of the phrases "selected embodiments," "at least one embodiment," "one embodiment," "another embodiment," "other embodiments," or "an embodiment" in various places throughout this specification are not necessarily referring to the same embodiment.

[0013] The illustrated embodiments will be best understood by referring to the drawings, in which like parts are designated by like numbers throughout. The following descriptions are intended to be merely exemplary and merely illustrate certain selected embodiments of devices, systems, and processes consistent with the embodiments claimed herein.

[0014] Although this disclosure includes a detailed description of cloud computing, it should be understood that implementation of the teachings described herein is not limited to cloud computing environments. Rather, embodiments of the present disclosure may be implemented in conjunction with any other type of computing environment now known or later developed.

[0015] Cloud computing is a service delivery model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, network bandwidth, servers, processing, memory, storage, applications, virtual machines, and services) that can be rapidly provisioned and released with minimal administrative effort or interaction with the service provider. This cloud model may include at least five characteristics, at least three service models, and at least four deployment models.

[0016] The characteristics are as follows:

[0017] On-demand self-service: Cloud consumers can unilaterally provision computing capacity, such as server time and network storage, as needed automatically and without human interaction with the provider of the service.

[0018] Broad network access. Functionality is available across the network and accessed through standard mechanisms that facilitate use by a variety of thin and thick client platforms (e.g., cell phones, laptops, and PDAs).

[0019] Resource Pooling. Provider computing resources are pooled to serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically allocated and reallocated according to demand. Consumers generally have no control or knowledge over the exact location of the resources provided, but there is a sense of location independence in that they may be able to specify a location at a higher level of abstraction (e.g., country, state, or data center).

[0020] Rapid Elasticity. Capabilities can be rapidly and elastically provisioned, sometimes automatically, to quickly scale out, and rapidly released to quickly scale in. To the consumer, the capabilities available for provisioning often appear unlimited, and any amount can be purchased at any time.

[0021] Metered services. Cloud systems automatically control and optimize resource usage by utilizing metering capabilities at some level of abstraction appropriate to the type of service (e.g., storage, processing, bandwidth, and active user accounts). Resource usage can be monitored, controlled, and reported, providing transparency to both providers and consumers of the services used.

[0022] The service model is as follows:

[0023] Software as a Service (SaaS). The functionality offered to the consumer is the use of the provider's applications running on a cloud infrastructure. The applications are accessible from a variety of client devices through thin-client interfaces, such as web browsers (e.g., web-based email). The consumer does not manage or control the underlying cloud infrastructure, including networks, servers, operating systems, storage, or even individual application functions, except possibly for limited user-specific application configuration settings.

[0024] Platform as a Service (PaaS). The functionality offered to the consumer is the deployment onto a cloud infrastructure of applications created or acquired by the consumer, written using programming languages ​​and tools supported by the provider. The consumer does not manage or control the underlying cloud infrastructure, including the network, servers, operating systems, or storage, but does have control over the deployed applications and, in some cases, the application hosting environment configuration.

[0025] Infrastructure as a Service (IaaS). The functionality provided to the consumer is the provisioning of processing, storage, network, and other basic computing resources onto which the consumer can deploy and run any software, which may include operating systems and applications. The consumer does not manage or control the underlying cloud infrastructure, but does have control over the operating system, storage, deployed applications, and sometimes limited control over the selection of networking components (e.g., host firewalls).

[0026] The deployment model is as follows:

[0027] Private cloud: This cloud infrastructure is operated solely for an organization. It may be managed by that organization or a third party and may reside on-premise or off-premise.

[0028] Community Cloud: This cloud infrastructure is shared by multiple organizations to support a specific community with common concerns (e.g., mission, security requirements, policy, and compliance considerations). It may be managed by those organizations or a third party and may reside on-premises or off-premises.

[0029] Public cloud: This cloud infrastructure is made available to the general public or large industry groups and is owned by an organization that sells cloud services.

[0030] Hybrid cloud: This cloud infrastructure is a composite of two or more clouds (private, community, or public) that remain their own entities but are bound together by standard or proprietary technologies that enable data and application portability (e.g., cloud bursting for load balancing between clouds).

[0031] Cloud computing environments are service-oriented and focus on statelessness, low coupling, modularity, and semantic interoperability. At the heart of cloud computing is an infrastructure that includes a network of interconnected nodes.

[0032] Referring now to FIG. 1 , an exemplary cloud computing environment 50 is shown. As shown, the cloud computing environment 50 includes one or more cloud computing nodes 10, with which local computing devices used by cloud consumers, such as a personal digital assistant (PDA) or mobile phone 54A, a desktop computer 54B, a laptop computer 54C, or an automotive computer system 54N, or combinations thereof, may communicate. The nodes 10 may also communicate with each other. These nodes may be physically or virtually grouped (not shown) in one or more networks, such as the private, community, public, or hybrid clouds described above, or combinations thereof. This enables the cloud computing environment 50 to provide infrastructure, platform, or software, or combinations thereof, as a service for which cloud consumers are not required to maintain resources on their local computing devices. It is understood that the types of computing devices 54A-N shown in FIG. 1 are intended to be exemplary only, and that computing node 10 and cloud computing environment 50 can communicate with any type of computing device through any type of network or network-addressable connection (e.g., using a web browser), or both.

[0033] Referring now to Figure 2, a set of functional abstraction layers provided by cloud computing environment 50 (Figure 1) is shown. It should be understood in advance that the components, layers, and functions shown in Figure 2 are intended to be merely exemplary, and embodiments of the present disclosure are not limited thereto. As shown, the following layers and corresponding functions are provided:

[0034] Hardware and software layer 60 includes hardware and software components. Examples of hardware components include mainframe 61, RISC (reduced instruction set computer) architecture-based servers 62, servers 63, blade servers 64, storage devices 65, and network and network forming components 66. In some embodiments, software components include network application server software 67 and database software 68.

[0035] The virtualization layer 70 provides an abstraction layer from which the following examples of virtual entities may be provided: virtual servers 71, virtual storage 72, virtual networks including virtual private networks 73, virtual applications and operating systems 74, and virtual clients 75.

[0036] In one example, the management layer 80 may provide the following functions: Resource provisioning 81 provides dynamic procurement of computing and other resources used to perform tasks within the cloud computing environment. Metering and pricing 82 provides cost tracking as resources are used within the cloud computing environment and billing or invoicing for the consumption of these resources. In one example, these resources may include application software licenses. Security provides identity verification for cloud consumers and tasks, as well as protection of data and other resources. User portal 83 provides access to the cloud computing environment for consumers and system administrators. Service level management 84 provides allocation and management of cloud computing resources to ensure required service levels are met. Service level agreement (SLA) planning and fulfillment 85 provides advance arrangements for and procurement of cloud computing resources where future demand is predicted by SLAs.

[0037] The Workload Layer 90 provides examples of functions for which a cloud computing environment may be used. Examples of workloads and functions that may be provided from this layer include mapping and navigation 91, software development and lifecycle management 92, virtual classroom instruction delivery 93, data analytics processing 94, transaction processing 95, and adversarial challenge to automated response problems 96.

[0038] 3, a block diagram of an example data processing system, referred to herein as computer system 100, is provided. System 100 may be embodied in a single-location computer system / server, or, in at least one embodiment, may be configured as a cloud-based system for sharing computing resources. For example, without limitation, computer system 100 may be used as cloud computing node 10.

[0039] To implement the systems, tools, and processes described herein, aspects of computer system 100 may be embodied in a single-location computer system / server, or, in at least one embodiment, configured as a cloud-based system sharing computing resources as a cloud-based support system. Computer system 100 may also operate with numerous other general-purpose or special-purpose computer system environments or configurations. Examples of well-known computer systems, environments, or configurations, or combinations thereof, that may be suitable for use with computer system 100 include, but are not limited to, personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputer systems, mainframe computer systems, and file systems (e.g., distributed storage environments and distributed cloud computing environments) that include any of the above systems, devices, and their equivalents.

[0040] Computer system 100 may be described in the general context of computer system-executable instructions, such as program modules, executed by computer system 100. Generally, program modules may include routines, programs, objects, components, logic, and data structures that perform particular tasks or implement particular abstract data types. Computer system 100 may also be practiced in a distributed cloud computing environment where tasks are performed by remote processing devices linked through a communications network. In a distributed cloud computing environment, program modules may be located in both local and remote computer system storage media, including memory storage devices.

[0041] As shown in FIG. 3, computer system 100 is illustrated in the form of a general-purpose computing device. Components of computer system 100 may include, but are not limited to, one or more processors or processing devices 104 (sometimes referred to as processors and processing units), such as hardware processors; system memory 106 (sometimes referred to as a memory device); and a communication bus 102 that couples various system components, including system memory 106, to processing device 104. Communication bus 102 represents any one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example and not limitation, such architectures include an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, an Enhanced ISA (EISA) bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus. Computer system 100 typically includes a variety of computer system-readable media. Such media may be any available media that can be accessed by computer system 100, including both volatile and nonvolatile media, removable and non-removable media. Additionally, computer system 100 may include one or more persistent storage devices 108, communications units 110, input / output (I / O) units 112, and displays 114.

[0042] Processing device 104 serves to execute instructions for software that may be loaded into system memory 106. Depending on the particular implementation, processing device 104 may be several processors, a multi-core processor, or some other type of processor. As used herein in reference to something, "several" means one or more of something. Furthermore, processing device 104 may be implemented using several heterogeneous processor systems, in which a main processor resides on a single chip along with secondary processors. As another illustrative example, processing device 104 may be a symmetric multiprocessor system including multiple processors of the same type.

[0043] System memory 106 and persistent storage 108 are examples of storage devices 116. A storage device may be any hardware capable of storing information, such as, but not limited to, data, functional program code, or other suitable information, or a combination thereof, on a temporary or persistent basis, or both. System memory 106 in these examples may be, for example, random access memory or any other suitable volatile or non-volatile storage device. System memory 106 may include a computer system-readable medium in the form of volatile memory, such as, for example, random access memory (RAM) or cache memory, or both.

[0044] Persistent storage 108 may take various forms, depending on the particular implementation. For example, persistent storage 108 may include one or more components or devices. For example, without limitation, persistent storage 108 may be provided for reading from and writing to a non-removable, non-volatile magnetic medium (not shown, commonly referred to as a "hard drive"). Although not shown, a magnetic disk drive may be provided for reading from and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive may be provided for reading from or writing to a removable, non-volatile optical disk, such as a CD-ROM, DVD-ROM, or other optical medium. In such cases, each may be connected to communication bus 102 by one or more data media interfaces.

[0045] The communications unit 110 in these examples may provide for communications with other computer systems or devices. In these examples, the communications unit 110 is a network interface card. The communications unit 110 may provide for communications using either or both physical and wireless communications links.

[0046] The input / output unit 112 may enable the input and output of data with other devices that may be connected to the computer system 100. For example, the input / output unit 112 may provide a connection for user input through a keyboard, a mouse, or some other suitable input device, or a combination thereof. Additionally, the input / output unit 112 may send output to a printer. The display 114 may provide a mechanism for displaying information to a user. Examples of input / output units 112 that facilitate establishing communications between various devices within the computer system 100 include, but are not limited to, a network card, a modem, and an input / output interface card. Additionally, the computer system 100 can communicate with one or more networks, such as, for example, a local area network (LAN), a general wide area network (WAN), or a public network (e.g., the Internet), or a combination thereof, via a network adapter (not shown in FIG. 3 ). It should be understood that other hardware and / or software components, not shown, may be used with the computer system 100. Examples of such components include, but are not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archival storage systems.

[0047] Instructions for the operating system, applications, and / or programs may be located in storage device 116, which communicates with processing device 104 over communication bus 102. In these illustrative examples, the instructions are in functional form in persistent storage 108. These instructions may be loaded into system memory 106 for execution by processing device 104. The processes of the different embodiments may be performed by processing device 104 using computer-implemented instructions, which may be located in a memory, such as system memory 106. These instructions are referred to as program code, computer-usable program code, or computer-readable program code, and may be read and executed by a processor of processing device 104. The program code in the different embodiments may be embodied in different physical or tangible computer-readable media, such as system memory 106 or persistent storage 108.

[0048] Program code 118 may be located in a functional form on a selectively removable computer-readable medium 120 and may be loaded or transmitted to computer system 100 for execution by processing device 104. In these examples, program code 118 and computer-readable medium 120 may form computer program product 122. In one example, computer-readable medium 120 may be computer-readable storage medium 124 or computer-readable signal medium 126. Computer-readable storage medium 124 may include, for example, an optical or magnetic disk that is inserted into or placed into a drive or other device that is part of persistent storage 108 for transmission to a storage device, such as a hard drive that is part of persistent storage 108. Additionally, computer-readable storage medium 124 may take the form of persistent storage, such as a hard drive, thumb drive, or flash memory that is connected to computer system 100. In some cases, computer-readable storage medium 124 need not be removable from computer system 100.

[0049] Alternatively, program code 118 may be communicated to computer system 100 using computer-readable signal medium 126. Computer-readable signal medium 126 may be, for example, a propagated data signal containing program code 118. For example, computer-readable signal medium 126 may be an electromagnetic signal, an optical signal, or any other suitable type of signal, or a combination thereof. These signals may be transmitted over a communications link, such as, for example, a wireless communications link, an optical fiber cable, a coaxial cable, a wire, or any other suitable type of communications link, or a combination thereof. In other words, in the illustrative example, the communications link and / or connection may be physical or wireless.

[0050] In some exemplary embodiments, program code 118 may be downloaded over a network from another device or computer system via computer-readable signal medium 126 to persistent storage 108 for use within computer system 100. For example, program code stored on a computer-readable storage medium of a server computer system may be downloaded over a network from the server to computer system 100. The computer system providing program code 118 may be a server computer, a client computer, or some other device capable of storing and transmitting program code 118.

[0051] Program code 118 may include, by way of example and not limitation, one or more program modules (not shown in FIG. 3 ) that may be stored in system memory 106, as well as an operating system, one or more application programs, other program modules, and program data. Each of the operating system, one or more application programs, other program modules, and program data, or any combination thereof, may comprise an implementation of a networking environment. The program modules of program code 118 generally perform the functions and / or methods of the embodiments described herein.

[0052] The different components illustrated for computer system 100 are not meant to provide architectural limitations to the manner in which different illustrative embodiments may be implemented. Different illustrative embodiments may be implemented in computer systems including components in addition to or instead of the components illustrated for computer system 100.

[0053] The present disclosure may be a system, method, or computer program product, or combination thereof, integrated at any possible level of technical detail. The computer program product may include a computer-readable storage medium (or media) having computer-readable program instructions for causing a processor to perform aspects of the present disclosure.

[0054] A computer-readable storage medium may be a tangible device capable of retaining and storing instructions for use by an instruction execution device. A computer-readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer-readable storage media includes portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, mechanically encoded devices such as punch cards or raised structures in grooves with recorded instructions, and any suitable combination of the foregoing. As used herein, computer-readable storage media should not be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission medium (e.g., light pulses passing through fiber optic cable), or electrical signals transmitted through wires.

[0055] The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to each computing / processing device, or may be downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, or a wireless network, or a combination thereof. The network may include copper transmission cables, fiber optic transmission cables, wireless transmission cables, routers, firewalls, switches, gateway computers, or edge servers, or a combination thereof. A network adapter card or network interface within each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to a computer-readable storage medium within the respective computing / processing device for storage.

[0056] The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to each computing / processing device, or may be downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, or a wireless network, or a combination thereof. The network may include copper transmission cables, fiber optic transmission cables, wireless transmission cables, routers, firewalls, switches, gateway computers, or edge servers, or a combination thereof. A network adapter card or network interface within each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to a computer-readable storage medium within the respective computing / processing device for storage.

[0057] Computer-readable program instructions for carrying out the operations of the present disclosure may be assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, configuration data for an integrated circuit, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk or C++, and procedural programming languages ​​such as the "C" programming language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., over the Internet using an Internet Service Provider). In some embodiments, electronic circuitry, including, for example, a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may execute computer-readable program instructions by using state information of the computer-readable program instructions to personalize the electronic circuitry to perform aspects of the present disclosure.

[0058] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0059] These computer-readable program instructions may be provided to a processor of a computer or other programmable data processing apparatus to produce a machine, whereby the instructions, executed by the processor of the computer or other programmable data processing apparatus, cause means for implementing the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams. These computer-readable program instructions may also be stored on a computer-readable storage medium that can instruct a computer, programmable data processing apparatus, or other device, or combination thereof, to function in a particular manner, whereby the computer-readable storage medium on which the instructions are stored includes instructions that implement aspects of the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[0060] The computer-readable program instructions may also be loaded into a computer, other programmable data processing apparatus, or other device to cause the computer, other programmable apparatus, or other device to perform a series of operational steps to produce a computer-implemented process, where the instructions executed on the computer, other programmable apparatus, or other device implement the functions / acts specified in one or more blocks of the flowcharts and / or block diagrams.

[0061] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowcharts or block diagrams may represent a module, segment, or portion of instructions, which includes one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions shown in the blocks may occur in a different order than that shown in the figures. For example, two blocks shown in succession may actually be accomplished as a single step, may be executed concurrently, may be executed substantially concurrently in a partially or fully overlapping manner, or the blocks may sometimes be executed in the reverse order, depending on the functionality involved. In addition, it should be noted that each block of the block diagrams and / or flowchart diagrams, and combinations of blocks in the block diagrams and / or flowchart diagrams, may be implemented by a specific-purpose hardware-based system that performs the specified functions or operations or executes a specific-purpose hardware and computer instruction combination.

[0062] Many known testing events are conducted online, where test takers are remotely located and in-person proctoring is impractical. As the internet continues to grow in presence in our society, access to information becomes easier and the internet serves as the first destination for questions. For example, answers to questions can be quite conveniently obtained through well-known and nearly ubiquitous artificial intelligence (AI) agents, such as digital assistants and chatbots, which are easily accessed through all forms of computing devices, including mobile devices. One example is the use of reverse image search engines to respond to image-based test questions. Therefore, given the possibility that test takers may utilize such technologies to generate automated responses to test questions, some form of remote proctoring is necessary to maintain the integrity of the testing process. At least some known methods of remotely administering and proctoring online exams involve invasive monitoring, such as video-based monitoring of test takers, and associated screen and screen sharing. Both methods require the full, focused, and continuous attention of a human proctor. In addition, both methods use additional computer resources that can affect the performance of testing hardware and software, which can be burdensome for time-limited tests. Furthermore, both methods increase the cost of testing by requiring a human proctor and additional hardware and software. At least some known methods of content manipulation to at least partially address the use of AI agents include watermarking text and image content and adding classical adversarial noise to audio content, but because these methods emphasize preserving semantic content, the modifications are constrained to be imperceptible and only slightly impede recognition by AI agents. Additionally, when questions for a test are created, the possibility of test takers using AI agents may limit or preclude many question formats, such as recognizing music, images, or text phrases.

[0063] Disclosed and described herein are systems, computer program products, and methods directed at improving test sequence parameters for exclusively testing actual human test takers by utilizing adversarial machine learning to encourage users to create test questions that cannot be easily solved by artificial intelligence (AI) agents. In at least some embodiments, restructuring the delivery of questions to test takers, including manipulating the question content, makes it substantially more difficult, if not completely impossible, to use any AI agent to generate automated test question responses. The extent to which questions can be manipulated is limited to preserve semantic content, i.e., to ensure that the test taker can fully understand the question. However, unlike pure watermarking and classical adversarial noise insertion, which require limited modifications to be imperceptible to the test taker, this disclosure describes clearly perceptible modifications to question content. Specifically, in contrast to pure steganography, which attempts to completely hide information while ensuring that the content is recoverable, this disclosure describes methods that remove at least some non-essential information from the question but do not allow for complete obfuscation. In some embodiments, content manipulation is multimodal, i.e., modifications to a problem are not limited to maintaining the same mode. For example, a text description can be converted to speech and then adversarial noise can be added.

[0064] Referring to FIG. 4, a block diagram illustrating a computer system, namely, a test question automated response adversarial hardening system 400 (herein referred to as system 400), configured to administer a test involving adversarial hardening of questions for automated response is presented. System 400 includes one or more processing devices 404 (only one shown) communicatively and operatively coupled to one or more memory devices 406 (only one shown). In addition, system 400 includes a data storage system 408 communicatively coupled to processing device 404 and memory device 406 through a communication bus 402. In one or more embodiments, communication bus 402, processing device 404, memory device 406, and data storage system 408 are similar to their counterparts, namely, communication bus 102, processing device 104, system memory 106, and persistent storage device 108, respectively, shown in FIG. 3. System 400 further includes one or more input devices 410 and one or more output devices 412 communicatively coupled to communication bus 402. In addition, system 400 includes one or more internet connections 414 (only one shown) to one or more artificial intelligence (AI) agents 416 (only one shown).

[0065] In one or more embodiments, an adversarial automated response obfuscation engine 420 (referred to herein as engine 420) resides within memory device 406. Engine 420 includes a checking module 422, a filtering / editing module 424, a semantic services module 426, and an AI agent communication module 428. These modules are further discussed with respect to Figures 4-7. Stored data 430 is maintained within data storage system 408 for access by memory device 406.

[0066] In one or more embodiments, the actual implementation of system 400, including engine 420, is as a cloud service, as described with respect to FIGS. 1 and 2 herein. As a cloud service, engine 420 may reside within any computing device within a cloud-based infrastructure for delivering the services described herein. Generally, no portion of engine 420 resides on a test taker's device; that is, the test taker's device is configured to receive the final version of the modified original test question (as further discussed herein) and provide a non-automated response from the test taker. Thus, in such an embodiment, the cloud service receives the original question and one of the more acceptable responses from its author through input device 410. The cloud-based service modifies the original question using engine 420. The modified question may be returned to the author through output device 412 for manual review, and if acceptable to the author, the modified question may be stored on the author's computing device or as part of stored data 430 in data storage system 408. In some embodiments, the modified questions may be sent directly to one or more test takers via output device 412. In some embodiments, engine 420 is a static algorithm with minimal, if any, customization. In some embodiments, engine 420 may be more flexibly customized by authors. Thus, only the bandwidth required for delivery of the modified questions and test taker responses is required.

[0067] In at least one embodiment, the actual implementation of system 400, including engine 420, is as a standalone system embedded within a computer system directly accessible to the author, such as, but not limited to, the author's personal or employer-provided computer system, including desktop computer 54B and laptop computer 54C (both shown in FIG. 1) and server 63 (shown in FIG. 2). In such an embodiment, as with the cloud-based implementation described above, the test taker's device should not receive or store anything other than the final modified questions.

[0068] Referring to FIG. 5A, a flowchart illustrating a process 500 for conducting a test through adversarial challenge of questions for automated responses is provided. Referring also to FIG. 4, in an embodiment, engine 420 electronically receives (502) an original question, where test takers are expected to submit responses electronically as part of the remote testing process. The original test question is authored by a test administrator and comprises one or more modalities of an image, one or more text phrases, and an audio clip. The author of the original question may have concerns that test takers might use an AI agent to generate automated responses. For example, commercially available AI agents, such as image search engines or digital assistants, may respond to such questions by providing, for example, a caption for the image or a statement explaining the information sought in the test question. Thus, in an embodiment, the author may utilize engine 420 to modify the original question before delivering it to one or more test takers by submitting the original question and response to engine 420 via input device 410.

[0069] The original question is sent to the check module 422 of the engine 420. The check module 422 is configured to perform an operation 502 of receiving the original question and then an operation of receiving the modified question as part of an iterative cycle described further herein. As part of the receiving operation, the check module 422 is further configured to examine the incoming question for any adversarial features provided as described herein, and then initiate the iterative cycle if they are not found in the original question. Additionally, when at least some adversarial features are determined for the current test question, the check module 422 further determines whether the adversarial features in the original question are fully satisfactory or fully unsatisfactory. Furthermore, in performing the iterative cycle of transforming the original question into a final modified question, the check module 422 facilitates continuing the iterative cycle until the criteria for sending the modified question to the test taker are met.

[0070] Thus, in an embodiment, the original question is modified (504), thereby generating an at least partially modified question. In order to transmit the modified question to the test taker, the at least partially modified question is configured to satisfy two requirements: the final modified question must be understandable by a human test taker, and the final modified question must not be responsive via electronic means without direct human support, i.e., the test taker must be able to respond to the modified question directly without the aid of an AI agent. Thus, the original question is transformed (506) into an at least partially modified question. More specifically, the original question is transmitted to one or more filtering components, i.e., the filtering / editing module 424, which is at least part of an iterative cycle from original question to final modified question that includes the question transformation operation 506.

[0071] In one or more embodiments, the filtering / editing module 424 electronically receives the original questions, which may include one or a combination of the following modalities: original text questions, original image questions, and original audio questions. The original questions may be transformed (506) through one or more operations according to predetermined parameters. For example, the image may be blurred, with the degree of blurring governed by established guidelines and parameters. In addition, one or more aspects of the image may be subjected to one or more perturbations or manipulations. Furthermore, the image may be manipulated, for example, by adding adversarial noise to at least partially obscure one or more features of the image. Image alterations are further discussed with respect to Figures 6(A) through 6(D). Audio alterations may include perturbations or modulations with effects including, but not limited to, playing each audio file backward, adding echo, removing specific frequency ranges, adding background noise, splitting into small fragments and rearranging them, and reversing lyrics. Audio alterations are further discussed with respect to Figures 7(A) through 7(D). Text questions may be transformed by converting the modality of the question while preserving the essence of the original question by undergoing text-to-speech conversion (also shown in FIGS. 7(A)-7(D)) or text-to-image conversion. For text-to-speech conversion, the speech perturbations and modulations described above may be used. Additionally, the text may be converted into one or more images, where the text may be converted into images that are non-copyable and non-pastable. In addition to perturbing the text, in some embodiments, random words may be added to or removed from every other sentence. Thus, the transformation operation 506 is performed at least in part as a function of the modality of the original test question.

[0072] In at least some embodiments, the at least partially modified question and the original question are sent (510) to a semantic services module 426. The semantic services module 426 is configured to perform a similarity assessment (512) between the original question and the at least partially modified question to determine semantic similarity. Referring to FIG. 5B, a continuation of process 500 from FIG. 5A is provided. Still referring to FIG. 4, more specifically, the semantic services module 426 is configured to perform a determining operation 514 to determine whether the previous conversion preserves sufficient information to be understandable to a human test-taker. For example, the converting operation 506 may modulate an audio question to remove at least a portion of the audio file. However, if the question involves speech recognition and assignment to a specific individual, the addition of background noise may be implemented, and the semantic services module 426 would determine (514) whether the amplitude of the background noise frequencies drowns out the portion of the audio question to be analyzed by the test-taker. In some embodiments, the stored data 430 may include examples of human-readable and non-human-readable modified problems. In some embodiments, the semantic services module 426 ModificationThe process 500 then returns to the perform similarity assessment operation 512, where the transformed problem undergoes an iterative cycle of operations 512-518 until a "Yes" result is achieved at the decision operation 514. Additionally, feedback from the similarity assessment operation 512 and the determination operation 514 to determine the semantic similarity between the original problem and the at least partially modified problem is sent to the further transformation operation 518.

[0073] Further, in at least some embodiments, the at least partially modified problem is transmitted (520) to a predetermined AI agent 416 through the AI ​​agent communication module 428 and the internet connection 414. In some embodiments, the AI ​​agent 416 is external to the engine 420. In some embodiments, one or more AI agents 416 are included within the engine 420 or reside within one or more of the memory device 406 and the data storage system 408. The engine 420 is agnostic to the nature of the AI ​​agents 416. The AI ​​agents 416 are configured to perform (522) a trial labeling operation on the at least partially modified problem through one or more machine learning models residing within the respective AI agents. The AI ​​agents 416 return results of the trial labeling operation 522 to the AI ​​agent communication module 428, which is further configured to perform a determining operation 524 to determine whether the at least partially modified problem can be answered through electronic means without human support, i.e., human response. When the at least partially altered problem is not labeled, i.e., when the AI ​​agent 416 fails to assign a label to the at least partially altered problem, the result of decision operation 524 is "No," and a "No" response to decision operation 524 is discussed further below. When the at least partially altered problem is labeled, the result of decision operation 524 is "Yes." As a result of a "Yes" response to decision operation 524, the at least partially altered problem is sent (526) to the filtering / editing module 424, where the at least partially altered problem is further transformed (518).

[0074] Additionally, the AI ​​agent 416 in some embodiments, or the AI ​​agent communication module 428 in some embodiments, assigns a confidence score to the response from the AI ​​agent 416. In some embodiments, the confidence score is a number along a scale of 0% to 100%, where the confidence score indicates, at least in part, the reliability of the AI ​​agent 416 that it correctly identified and labeled the at least partially modified problem.

[0075] Further, in some embodiments, the AI ​​agent 416 may be prompted to provide a response to the question. Additionally, a decision operation 524 may be performed to determine whether the response provided by the AI ​​agent is sufficiently close to the answer (response) to the test question provided by the author in the receive operation 502. When the AI ​​agent 416 is able to provide a sufficiently accurate response, the result of the decision operation 524 is "Yes." When the AI ​​agent 416 is unable to provide a sufficiently accurate response, including providing an incorrect response, the result of the decision operation is "No." Process 500 returns to the perform trial labeling operation 522, and the further transformed question undergoes an iterative cycle of operations 520-524-518 until a "No" result is achieved at the decision operation 524. Additionally, feedback from the trial labeling operation 522 and the decision operation 524 is sent to the further transformation operation 518.

[0076] As discussed, two iterative cycles or loops are processed. The first iterative cycle includes operations 512-514-516 performed through the semantic services module 426 to generate a final modified test question that is understandable to a human test-taker. The second iterative cycle includes operations 522-524-526 performed through the AI ​​agent communication module 428 and the AI ​​agent 416 to generate a final modified test question to which the AI ​​agent 416 cannot respond. In some embodiments, the two iterative cycles may be performed in parallel. In some embodiments, the two iterative cycles may be performed sequentially in a cycle including first cycle-second cycle-first cycle, etc., until the requirements of both determining operations 514 and 524 are achieved. In some embodiments, the two iterative cycles may be performed in a manner such that one of the two cycles generates a satisfactory modified question before providing the modified question to the other cycle.

[0077] 5C, a continuation of process 500 from FIG. 5B is presented. Continuing with reference to FIGS. 4, 5A, and 5B, in these embodiments, when the result of decision operation 514 is "Yes" and the result of decision operation 524 is "No," the final modified question and its results are sent to a merging operation 530 associated with check module 422. As a result of merging operation 530, the fully modified question is sent 532 to the human test-taker via check module 422 and output device 412, and process 500 ends. In other embodiments, either a "No" result of decision operation 514 or a "Yes" result of decision operation 524 prompts check module 422 to prevent the modified question from being sent to the human test-taker, and further modification of the question is performed as described above.

[0078] As an example of modification, an image-based test question is presented. Referring to FIG. 6(A), an image diagram is presented showing an example of an original image 600 of a turtle associated with the original test question. While the turtle image 600 is shown in black and white, a color image may also be presented as the original question. Referring to FIG. 6(B), the original image 600 of FIG. 6(A) is provided, which shows an example of an at least partially modified image 610 associated with the at least partially modified test question. Referring also to FIGS. 4, 5A, 5B, and 5C, the semantic service module 426 analyzes the image 610 and determines that the result of decision operation 514 is "Yes" because the feature edges of the image 610 are preserved and sufficient information is retained in the image 610 to allow a human to recognize a turtle. However, the result of decision operation 524 is "Yes," indicating that the AI ​​agent 416 correctly identified (labeled) the image 610 as a turtle with a 99% confidence level. Thus, as described above, a "Yes" result at decision operation 514 and a "Yes" result at decision operation 524 will prohibit check module 422 from sending the altered question to a human test-taker, and further alteration of the question will be performed as described above.

[0079] Referring to FIG. 6(C), the turtle image diagram of FIG. 6(A) is presented, illustrating an example of a further altered image 620 associated with a further altered test question. Referring also to FIGS. 4, 5A, 5B, and 5C, semantic services module 426 analyzes image 620 and determines that the result of decision operation 514 is “No” because most of the features in image 600 are substantially absent from image 620, making it nearly impossible for a human to recognize a turtle. However, the result of decision operation 524 is “No,” whereby AI agent 416 is unable to identify (label) image 620 as a turtle and therefore no confidence level is provided. Thus, as discussed above, the “No” result of decision operation 514 and the “No” result of decision operation 524 result in the altered question being prohibited from being sent to a human test-taker via check module 422, and further alteration of the question is performed as discussed above.

[0080] Referring to FIG. 6(D), the turtle image diagram of FIG. 6(A) is presented, illustrating an example of a final modified image 630 associated with the modified test question. Semantic services module 426 analyzes image 630 and determines that the result of decision operation 514 is “Yes” because most of the feature edges of image 630 are preserved and sufficient information is preserved in image 630 to allow a human to recognize a turtle. The result of decision operation 524 is “No,” whereby AI agent 416 is unable to identify (label) image 630 as a turtle and therefore no confidence level is provided. As discussed above, the “Yes” result of decision operation 514 and the “No” result of decision operation 524 enable the final modified question and result sent to merge operation 530 to be sent 532 to a human test-taker via check module 422, thereby terminating process 500.

[0081] Thus, the check module 422 checks a simple AND condition through a merge operation 530: whether the semantic services module 426 deemed the semantic features of the modified question to be sufficiently preserved to remain human-understandable, and whether the AI ​​agent 416 provided an incorrect or substantially ambiguous answer, or did not provide an answer (completely disabled). If both conditions are met, the check module 422 interrupts the loop cycle and sends the fully modified question to the test taker (532); otherwise, the loop cycle continues.

[0082] An example of the modification of a text-to-audio test question is presented. Referring to FIG. 7(A), a text representation illustrating an example of a text test question and answer / response 700 is provided. Referring to FIG. 7(B), a block diagram illustrating an example of an audio test question 710 based on the text test question 700 of FIG. 7(A) is provided. The semantic service module 426 determines that the audio test question 710 is understandable by a human test taker. Additionally, the AI ​​agent 416 processes the audio test question 710 and provides a correct response with a 99% confidence value. Therefore, two requirements for the check module 422 to be able to send the modified test question to the test taker are not met. Referring to FIG. 7(C), a block diagram illustrating an example of an edited audio test question 720 based at least in part on the modified text test question 722 derived from the example of the text test question 700 of FIG. 7(A) is provided. Because the edited and converted audio question 720 is substantially unintelligible to both humans and the AI ​​agent 416, the edited and converted audio question 720 is deemed unable to be sent to the human test taker by the checking module 422. Referring to FIG. 7D, a block diagram is provided illustrating an example audio test question 730 based on the text test question 700 of FIG. 7A, to which an audio noise signal 732 of a particular frequency and amplitude has been added. The semantic services module 426 is unable to distinguish the converted audio from the noise 732, and the AI ​​agent 416 is unable to label the noisy converted audio question 730. As a result, the two requirements of the checking module 422 are met, and the noisy converted audio question 730 can be sent 532 to the human test taker.

[0083] In some embodiments, it may be possible that the results of both the semantic services module 426 and the AI ​​agent 416 cannot be reconciled, i.e., regardless of the number of iterations, only one of the two results of the decision operations 514 and 524 may reconcile to satisfy the requirements of the check module 422. For such a condition, the engine 420 includes an iteration counter and sufficient logic to stop the iteration after a predetermined number of iterations. When a loop iteration is stopped, a notification may be provided to the author, who may then make a decision on how to proceed, including, but not limited to, removing the test question from the test bank, manually modifying the question through trial and error, or using the most recent iteration of the modified question in anticipation that completely eliminating the possibility of using the AI ​​agent 416 for each question is not feasible.

[0084] Additionally, in some embodiments, the original question may be configured to meet the two requirements of the check module 422 without requiring any modification. In such embodiments, the filtering / editing module 424 may be configured to perform a first iteration with a "do nothing" command, whereby the unmodified exam question will result in a "Yes" outcome from decision operation 514 and a "No" outcome from decision operation 524, terminating the iterative loop and triggering merge operation 530 to send the unmodified exam question to the test taker (532).

[0085] The systems, computer program products, and methods disclosed herein facilitate overcoming the disadvantages and limitations of known systems and methods for administering remote testing to individuals that require test takers to have access to the internet. Specifically, the present disclosure describes an automated process and system for generating ambiguous questions from regular questions in a manner that tests the cognitive abilities of human test takers but avoids the current capabilities of AI agents. The testing mechanisms described herein facilitate reducing the need for extensive remote proctoring, thereby reducing the need for intrusive and resource-intensive computing environments, including the establishment and maintenance of stable, high-bandwidth communication links. Furthermore, many previous limitations on test modalities, including the use of music, images, and text phrases, may be lifted. Thus, through the present disclosure, significant improvements are realized over the implementation and integrity of known remote testing systems.

[0086] The description of various embodiments of the present disclosure is presented for illustrative purposes, but is not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope of the described embodiments. The terms used herein have been selected to best explain the principles of the embodiments, practical applications, or technical improvements over technologies found in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. one or more processing devices, receiving an original problem electronically, wherein a response to the original problem is to be submitted electronically by a human; modifying the original problem, thereby generating a modified problem that includes information to ensure that the original problem remains unchanged to facilitate human-judged recognition, wherein the modified problem is understandable by the human; sending the modified problem to one or more artificial intelligence (AI) agents; performing a trial labeling operation on the modified problem through the one or more AI agents; and determining, based on the altered problem being human understandable and the trial labeling operation failing to assign a label to the altered problem, that the altered problem is not responsive via electronic means without human assistance; A system configured to:

2. one or more characteristics of the modified subject matter are retained; 2. The system of claim 1, wherein the determining comprises determining that the modified problem can be answered through electronic means without human support based on assigning a label to the modified problem in the trial labeling operation.

3. further comprising one or more filtering components communicatively coupled to the one or more processing devices, the one or more processing devices sending the original problem to the one or more filtering components, thereby performing one or more transformations on the original problem; The system of claim 1 or 2, further configured to:

4. further comprising one or more semantic services communicatively coupled to the one or more processing devices, the one or more processing devices sending the modified problem to the one or more semantic services; submitting the original problem to the one or more semantic services; performing a similarity assessment between the original problem and the modified problem using the information through the one or more semantic services; and The system of any one of claims 1 to 3, further configured to:

5. the one or more processing devices: The system of claim 4 , further configured to determine, through the one or more semantic services, that the modified problem is understandable by the human.

6. the one or more processing devices: assigning a label to the modified problem through the one or more AI agents; and assigning, through the one or more AI agents, a confidence value to the modified problem; generating a response to the modified problem through the one or more AI agents; and The system of claim 2 , further configured to perform one or more of the following:

7. the one or more processing devices: failing to assign a label to the modified problem through the one or more AI agents; and not generating a response to the modified problem through the one or more AI agents; generating, through said one or more AI agents, an erroneous response to said modified problem; The system of claim 1 , further configured to perform one or more of the following:

8. The original problem is The original text question and The original image problem and The original audio problem and The one or more processing devices are further configured to perform an alteration of the original problem, including an alteration of the modality of the original problem, the alteration altering the original text problem by: an at least partially altered audio problem that includes at least some adversarial noise; At least partially altered image issues and The system of any one of claims 1 to 7, comprising converting the data into one or more of:

9. one or more filtering components communicatively coupled to the one or more processing devices; one or more semantic services communicatively coupled to the one or more filtering components; access to one or more artificial intelligence (AI) agents communicatively coupled to the one or more filtering components; wherein the one or more processing devices further include converting the original problem into an at least partially modified problem through the one or more filtering components; iteratively submitting the original problem and the at least partially modified problem to the one or more semantic services; iteratively sending the at least partially modified problem to the one or more AI agents; repeatedly sending the at least partially modified problem to the one or more filtering components; repeatedly, determining, through the one or more semantic services, that the at least partially modified problem is understandable by the human; determining, through the one or more AI agents, that the at least partially modified problem cannot be adequately responded to through the electronic means without the human support, thereby establishing that the at least partially modified problem is a fully modified problem; transmitting said fully modified problem to said human over a network by said computer; The system of any one of claims 1 to 8, further configured to:

10. 1. A computer-implemented method comprising: receiving an original problem electronically, wherein a response to the original problem is to be submitted electronically by a human; modifying the original problem, thereby generating a modified problem that includes information to ensure that the original problem remains unchanged to facilitate human cognition, wherein the modified problem is understandable by the human; sending the modified problem to one or more artificial intelligence (AI) agents; performing a trial labeling operation on the modified problem through the one or more AI agents; and determining, based on the altered problem being human understandable and the trial labeling operation failing to assign a label to the altered problem, that the altered problem is not responsive via electronic means without human assistance; A method comprising:

11. one or more characteristics of the modified subject matter are retained; 11. The method of claim 10, wherein the determining comprises determining, based on the trial labeling operation assigning a label to the modified problem, that the modified problem can be answered through electronic means without human support.

12. generating the modified problem 12. The method of claim 10 or 11, comprising sending the original problem to one or more filtering components, thereby performing one or more transformations on the original problem.

13. sending the modified problem to one or more semantic services; submitting the original problem to the one or more semantic services; performing a similarity assessment between the original problem and the modified problem using the information through the one or more semantic services; and The method of claim 12 further comprising:

14. said performing said similarity assessment further comprising: The method of claim 13 , further comprising determining, through the one or more semantic services, that the modified problem is understandable by the human.

15. performing a trial labeling operation on the modified problem, assigning a label to the modified problem; assigning a confidence value to the modified problem; generating a response to the modified problem; The method of claim 11 , comprising one or more of:

16. performing a trial labeling operation on the modified problem, failing to assign a label to the modified problem; and not generating a response to the modified problem; generating a false response to the modified problem; The method of claim 10, comprising one or more of:

17. said receiving said original problem electronically comprising: The original text question and The original image problem and The original audio problem receiving the original problem having one or more modalities of: The modifying of the original problem includes modifying the modality of the original problem, the modifying of the modality comprising modifying the original text problem by: an at least partially altered audio problem that includes at least some adversarial noise; At least partially altered image issues and 17. The method of any one of claims 10 to 16, comprising converting into one or more of:

18. converting the original problem into an at least partially modified problem through the one or more filtering components; iteratively submitting the original problem and the at least partially modified problem to the one or more semantic services; iteratively sending the at least partially modified problem to the one or more AI agents; repeatedly sending the at least partially modified problem to the one or more filtering components; repeatedly, determining, through the one or more semantic services, that the at least partially modified problem is understandable by the human; determining that the at least partially modified problem cannot be adequately responded to through the electronic means without the human support, thereby establishing that the at least partially modified problem is a fully modified problem; transmitting said fully modified problem to said human over a network by said computer; The method of any one of claims 10 to 17, further comprising:

19. A computer-executable program for causing a computer system to execute the method according to any one of claims 10 to 18.

20. 20. A computer-readable storage medium storing the computer-executable program of claim 19.

Images