Assistance system, computer-implemented method, and computer program

Abstract

To achieve appropriate support of user's work by information technology SOLUTION: A disclosed system is a support system that provides a work environment for a user and supports work of the user, the support system including an agent that performs vertical support for the user and performs horizontal support for the user, in which the agent is configured to output a suggestion related to the work of the user for the vertical support, and the agent is configured to search a knowledge base included in the agent on the basis of work content in the work environment and determine knowledge to be presented to the user for the horizontal support.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a support system, computer-implemented method, and computer program for supporting collaborative learning or other user activities. [Background technology]

[0002] As shown in the field of proximal development in psychology, it is believed that at the root of human learning activities, learning through dialogue and interaction with others is important for acquiring knowledge. In the field of learning science, this knowledge has been developed and active learning, a type of collaborative learning that allows learners to take the initiative in their learning activities, has become widespread.

[0003] As indicated by the Learning by Doing theory, learners can deepen their understanding not by simply reading textbooks or listening to lectures, but by engaging with others in practical learning and actually doing the learning. However, many beginners still have difficulty reconstructing knowledge and advancing discussions appropriately on their own, and so they often proceed with the help of advice and guidance from an expert, such as a teacher, who acts as a kind of scaffolding. In such classroom group work (collaborative learning), one teacher provides facilitation, including advice, to each group of learners as appropriate. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2004-527859 Summary of the Invention

[0005] However, in large classes, it is difficult to provide this type of learning to all learners. Providing consistent facilitation to each learner in a limited amount of time is also difficult from a human resources perspective. Furthermore, in order to provide individual support for each learner's unique impasses and problems, it is necessary to constantly monitor the learner's condition. In light of these points, it is an important challenge to use information technology to provide learning support for learners who are engaged in learning.

[0006] Furthermore, these issues are not limited to situations where learners are learning, but also arise in situations of intellectual group work other than learning, such as discussions or brainstorming.Furthermore, these issues are not limited to situations where multiple people work together, such as cooperative learning or other group work, but also arise in situations where students work alone.

[0007] Therefore, it is desirable to realize appropriate support for user work using information technology.

[0008] One aspect of the present disclosure is a system. The disclosed system is a support system that provides a work environment for a user and supports the user's work, and includes an agent that provides vertical support to the user and horizontal support to the user, wherein for the vertical support, the agent is configured to output suggestions related to the user's work, and for the horizontal support, the agent is configured to search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to present to the user.

[0009] Another aspect of the present disclosure is a method. The disclosed method is a computer-implemented method executed by a computer to support a user's work, comprising: providing a work environment for the user; and an agent providing vertical support and horizontal support to the user, wherein for the vertical support, the agent is configured to output suggestions related to the user's work, and for the horizontal support, the agent searches a knowledge base held by the agent based on work content in the work environment to determine knowledge to present to the user.

[0010] Another aspect of the present disclosure is a computer program for causing a computer to operate as a support system that provides a work environment for a user and supports the user's work, the support system including an agent that provides vertical support to the user and horizontal support to the user, the agent being configured to output suggestions related to the user's work for the vertical support, and the agent being configured to search a knowledge base held by the agent based on work content in the work environment and determine knowledge to present to the user for the horizontal support.

[0011] Further details will be described in the following embodiments. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a configuration diagram of a support system according to an embodiment. [Figure 2] FIG. 2 is a hardware configuration diagram of the support system and the client device according to the embodiment. [Figure 3] FIG. 3 is a diagram showing a screen interface provided to the user by the support system. [Figure 4] FIG. 4 shows the initial state of the collaborative work process. [Figure 5] FIG. 5 is a diagram showing a pre-created concept map. [Figure 6] FIG. 6 is a diagram showing a collaborative work process A. [Figure 7] FIG. 7 is a diagram showing the collaborative work process B. [Figure 8] FIG. 8 is a diagram showing a collaborative work process C. [Figure 9] FIG. 9 is a diagram showing a collaborative work process D. [Figure 10] FIG. 10 is a diagram showing a collaborative work process E. [Figure 11] FIG. 11 is a diagram showing a collaborative work process F. [Figure 12] FIG. 12 is a diagram showing a collaborative work process G. [Figure 13] FIG. 13 is a diagram showing the collaborative drawing area and the presentation unit in the collaborative work process G. [Figure 14] FIG. 14 is a diagram showing the chat area in the collaborative work process G. DETAILED DESCRIPTION OF THE INVENTION

[0013] <1. Overview of the Support System, Computer Implementation Method, and Computer Program>

[0014] (1) A support system according to an embodiment can provide a work environment for a user and support the user's work. The support system according to an embodiment can include an agent that provides vertical support to the user and horizontal support to the user. For the vertical support, the agent can be configured to output suggestions related to the user's work. For the horizontal support, the agent can be configured to search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to present to the user. According to the support system according to an embodiment, the user's work is appropriately supported by the agent's vertical support and horizontal support.

[0015] In one aspect, the work environment may include a collaborative work environment that can be used by multiple users collaboratively. In another aspect, the work environment may include a drawing area where work including drawing can be performed, and a chat area where chat can be performed.

[0016] (2) The work environment preferably includes a drawing area where a plurality of users can work together, including drawing, and a chat area where the plurality of users can chat.

[0017] (3) It is preferable that the work environment includes a chat area where chatting is possible, and for the vertical support, the agent is configured to output the suggestions to the chat area.

[0018] (4) It is preferable that the work environment includes a drawing area where work including drawing can be performed, and for the horizontal support, the agent is configured to search a knowledge base that the agent possesses based on the content of the drawing in the drawing area and determine the knowledge to present to the user.

[0019] The work environment may include a drawing area where work can be done, including drawing a concept map.

[0020] (5) It is preferable that the working environment includes a drawing area where tasks including drawing a concept map can be performed, and for the horizontal support, the agent is configured to search a knowledge base possessed by the agent based on the concept map in the drawing area, determine knowledge to present to the user, and present to the user a concept map drawn based on the determined knowledge.

[0021] (6) The working environment preferably includes a chat area where multiple users can chat, and for the vertical support, the agent is configured to present a summary of the chat content to the user.

[0022] (7) It is preferable that the system further comprises a pre-work content presenting unit that presents the pre-work content of the user to the user.

[0023] (8) The user's work may be collaborative learning by multiple users.

[0024] (9) It is preferable that the agents include a first agent for the vertical support and a second agent for the horizontal support, and that a first avatar representing the first agent and a second avatar representing the second agent are presented to the user.

[0025] (10) A method according to an embodiment may be a computer-implemented method executed by a computer to support a user's work. The method according to an embodiment may include providing a work environment for the user, and an agent providing vertical support and horizontal support to the user. For the vertical support, the agent may output suggestions related to the user's work. For the horizontal support, the agent may search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to present to the user.

[0026] (11) A computer program according to an embodiment may be a computer program for operating a computer as a support system that provides a work environment for a user and supports the user's work. The support system may include an agent that provides vertical support to the user and horizontal support to the user. For the vertical support, the agent may be configured to output suggestions related to the user's work. For the horizontal support, the agent may be configured to search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to present to the user.

[0027] 2. Examples of Support Systems, Computer-Implemented Methods, and Computer Programs

[0028] Hereinafter, the embodiments will be described in more detail with reference to the drawings.

[0029] 2.1 Overview of the support system

[0030] 1 shows a support system 100 according to an embodiment. The support system 100 described below is, as an example, an intelligent learning support system in which an agent 120 of the support system 100 provides learning support by providing reflections while a pair of learners U1 and U2, who are users, proactively carry out explanation activities using a concept map that they jointly create.

[0031] In system 100, learners U1 and U2 create concept maps together with their collaborators while conversing with them via text chat. In system 100, agent 120 monitors the concept maps created by learners U1 and U2 and the content of the chat conversations, and provides advice and example concept maps to learners U1 and U2. Providing advice and example concept maps to learners U1 and U2 is performed, for example, through (1) knowledge inference using a cognitive architecture and (2) language analysis and sentence generation using LLMs (Large Language Models).

[0032] It is expected that the pair of learners U1 and U2 will reflect on the advice and generated examples provided by this system 100, thereby promoting their learning activities.

[0033] The support system 100 communicates with client devices 210 and 220 used by learners U1 and U2 via a network. The client devices 210 and 220 are provided with a client module 230 that operates on the client devices 210 and 220. The client module 230 communicates with the support system 100 and displays a screen that serves as a work environment (e.g., a collaborative learning environment) for learners U1 and U2.

[0034] The client module 230 is software installed on the client devices 210 and 220. The client module 230 is connected to the server 110 (described later) via a network such as a local area network, and transfers information via server-client socket communication.

[0035] Specifically, the client module 230 transmits (1) the drawing contents of the concept map created by the learners U1 and U2 and (2) the text of the text chat to the server 110. Then, the client module 230 receives information from the server 110, such as the concept map drawing module 150, presentation modules 140 and 160, and knowledge generation module 130 (to be described later), and displays (1) the drawing of the concept map and (2) the entries in the text chat field on the screen of the display (not shown) of the client device 220.

[0036] The support system 100 may include a server 110. The server 110 communicates with the client module 230 and manages data and collects logs for concept map drawing. The data and logs are stored in a database 111.

[0037] The support system 100 may include an agent 120. The agent 120 is software that performs support tasks for learners U1 and U2 who are users.

[0038] <2.2 Agent>

[0039] The agent 120 of the embodiment provides, as an example, vertical support to the learners U1 and U2 who are users, as well as horizontal support.

[0040] Vertical support is a support method centered on vertical facilitation, in which advice and instruction are given to users who are support recipients, such as learners U1 and U2. In vertical support, agents can be called teacher agents or tutor agents.

[0041] In vertical support, for example, a conversational agent that responds based on conversation (chat) analysis of collaborative learning activities between people is used to present metacognitive suggestions. This conversational agent (teacher or tutor agent) provides support by acting as a third party to a pair of learners who are explaining something to each other.

[0042] Examples of vertical support include a conversational agent providing learners with affective feedback, including positive emotions, during intervention (Y. Hayashi, “Designing affective pedagogical agents: How learners' and agents' gender and age influence emotion in an online tutoring,” Proceedings of the 20th International Conference on Computers in Education (ICCE2012), pp. 46-50, 2012). Multiple agents share roles and provide advice (Y. Hayashi, “Multiple pedagogical conversational agents to support learner-learner collaborative learning: Effects of splitting suggestion types,” Cognitive Systems Research, vol. 54, pp. 246-257, 2018). Furthermore, gaze feedback, which uses an eye tracker to display gaze information in real time, is also being proposed (Y. Hayashi, “Gaze awareness and metacognitive suggestions by a pedagogical conversational agent: an experimental investigation on interventions to support collaborative learning process and performance,” International Journal of Computer-Supported Learning). Collaborative Learning, vol.15, pp.469-98, Dec. 2020.)

[0043] The agent 120 of the embodiment not only provides the above-mentioned vertical support but also provides horizontal support in parallel. While vertical support by a conversational agent alone can cause some problems, providing horizontal support as well makes the support more effective.

[0044] In other words, in vertical support activities, learners and other workers basically play a central role in the explanation activity, and the agent acts as a third party to intervene in the interaction. As a result, some learners sometimes ignore the content of the agent's instructions, and as a result, the facilitation effect is not fully realized. This is due to the reliance on a "vertical" facilitation-centered method in which the agent unilaterally gives advice to the learner.

[0045] Therefore, rather than providing one-way knowledge instruction to learners as a third party, it would be useful to introduce a "horizontal support" method in which an agent participates as a member of the explanation activities of learners and other workers. To achieve this type of support, it would be useful to introduce knowledge teaching methods from learning science, such as mutual teaching and learning by teaching. This point will be discussed later.

[0046] 2.3 Support through mutual teaching with agents

[0047] In the field of learning science, a technique called worked-out example is being studied, in which learners observe examples and solution steps for a presented problem and then explain them themselves to promote comprehension (Michelene. Chi, Miriam. Bassok, Matthew. Lewis, Peter. Reimann, and Robert. Glaser, “Self-explanations: How students study and use examples in learning to solve problems,” Cognitive Science, vol. 13, no. 2, pp. 145-182, 1989).

[0048] A worked-out example consists of a problem formula, solution steps, or the solution itself, and is considered effective in fields such as mathematics, physics, and computer programming (UHM Alexander Renkl, Robert K. Atkinson and R. Staley, “From example study to problem solving: Smooth transitions help learning,” The Journal of Experimental Education, vol. 70, no. 4, pp. 293-315, 2002. https: / / doi.org / 10.1080 / 00220970209599510).

[0049] Worked-out examples also include erroneous worked-out examples, which are examples that contain errors (J. Sweller, JJG vanMerrienboer, and FGWC Paas, “Cognitive architecture and instructional design,” Educational Psychology Review, vol. 10, no. 3, pp. 251-296, 1998.).

[0050] It has been found that when there is a lot of prior knowledge, erroneous worked-out examples improve transfer performance, but when there is little prior knowledge, worked-out examples promote it (C.S. Grosse and A. Renkl, “Finding and fixing errors in worked examples: Can this foster learning outcomes?,” Learning and Instruction, vol.17, pp.612-634, 2007.).

[0051] In addition, a method called "Learning by Teaching" has also been devised in relation to the Worked-out example. The basic idea of ​​this teaching method is that students take on the role of teachers at certain times in the classroom (U. Hanke, "Learning by Teaching," pp. 1830-1832, Springer US, Boston, MA, 2012).

[0052] A student acting as a teacher can present new topics to other students, lead discussions, and help each other solve learning problems. Systems that implement these concepts in AI-based tutoring are currently being developed (N. Matsuda, W.W. Cohen, and K.R. Koedinger, “Teaching the teacher: tutoring simstudent leads to more effective cognitive tutor authoring,” International Journal of Artificial Intelligence in Education, vol. 25, no. 1, pp. 1–34, 2015.) (Yugo Hayashi and Tomoo Inoue, “Designing Collaborative Learning with Multiple Educational Conversational Agents,” IEICE Transactions on Education, vol. J98-A, no. 1, pp. 76–84, 2015.).

[0053] For example, a system has been devised in which learners can teach what they have learned to a knowledge-teaching agent named Betty in the form of a concept map that mimics a semantic network (N. Matsuda, W.W. Cohen, and K.R. Koedinger, “Teaching the teacher: tutoring simstudent leads to more effective cognitive tutor authoring,” International Journal of Artificial Intelligence in Education, vol. 25, no. 1, pp. 1-34, 2015.). Learners present problems to Betty, which then solves them by reasoning based on the concept map created by the learner. By looking at the concept map, learners reflect on it and revise their own thinking, deepening their understanding of the concept.

[0054] For these reasons, it is more useful to have a "horizontal" mutual learning approach such as Learning by Teaching, rather than a "vertical" facilitation approach in which an agent (AI tutor) provides one-sided instruction.

[0055] In this embodiment, based on the above, a new collaborative learning support system 100 has been constructed that incorporates both this "vertical" facilitation and "horizontal" mutual learning in the context of collaborative learning as an example of collaborative work.

[0056] In this system 100, learners U1 and U2 learn to build conceptual understanding through explanatory activities about psychological concepts with human collaborators using concept maps, as exemplified by learning by doing. As shown in Figure 1, system 100, for example, features two educational conversational agents 120 with different roles. One is agent 121 (teacher agent 121) that provides vertical support by providing metacognitive facilitation to learners U1 and U2. The other agent 122 is a student agent 122 that generates a new concept map while studying the contents of the learners' concept maps.

[0057] In other words, learners U1 and U2 teach the student agent how to create concept maps by creating them themselves, and then they observe the generated examples (concept maps) and make comments about them as they learn (horizontal support).

[0058] <2.4 Details of the support system>

[0059] The assistance system 100 includes the server 110 and agent 120, as well as various modules 130, 140, 150, and 160. The assistance system also has the function of linking with a cognitive architecture (ACT-R). This system is built in C# and links with ACT-R, which is written in LISP, using socket communication.

[0060] FIG. 2 shows an example of the hardware configuration of the support system 100 and the client devices 210 and 220. As shown in FIG.

[0061] The support system 100 is configured by one or more computers. The computer that configures the support system 100 includes a processor 101 and a memory 102 connected to the processor 101. The computer that configures the support system 100 may include a communication device 103 for communicating with the client devices 210 and 220.

[0062] The processor 101 is, for example, a CPU. The memory 102 includes, for example, a primary storage device and a secondary storage device. The primary storage device is, for example, a RAM. The secondary storage device is, for example, a hard disk drive (HDD) or a solid state drive (SSD). The memory 102 includes a computer program 102A executed by the processor 101. The processor 101 reads and executes the computer program 102A stored in the memory 102. The computer program 102A has program code representing instructions for causing a computer to function as the assistance system 100 according to the embodiment.

[0063] The client devices 210 and 220 are configured by computers. The computers configuring the client devices 210 and 220 include a processor 101 and a memory 102 connected to the processor 101. The computers configuring the client devices 210 and 220 may include a communication device 203 for communicating with the support system 100.

[0064] The processor 201 is, for example, a CPU. The memory 202 includes, for example, a primary storage device and a secondary storage device. The primary storage device is, for example, a RAM. The secondary storage device is, for example, a hard disk drive (HDD) or a solid state drive (SSD). The memory 202 includes a computer program 202A executed by the processor 201. The processor 101 reads and executes the computer program 202A stored in the memory 102. The computer program 202A has program code representing instructions for causing a computer to function as the client devices 210, 220 (client module 230) according to the embodiment.

[0065] 3(A) and (B) are display examples of the screen interface 10 of the client devices 210 and 220 of learners U1 and U2. The screen interface 10 of the client devices 210 and 220 is a learning environment 10 (working environment 10) provided to the users, learners U1 and U2. FIG. 3(B) shows an example of an actual screen interface, and FIG. 3(A) explains the individual screen areas in FIG. 3(B).

[0066] The screen interface 10, which serves as the work environment 10, includes pre-task content presentation sections 11 and 12. In FIGS. 3(A) and 3(B), the pre-task content presentation sections 11 and 12 are located at the upper left and lower left of the screen interface 10. The pre-task content presentation sections 11 and 12 display the user's pre-task schedule. As an example, the pre-task content presentation sections 11 and 12 may display concept maps that the users, learners U1 and U2, individually created for the same assignment before (during the individual phase of) collaborative learning (collaborative work).

[0067] 3(A) and 3(B), the first pre-work content presentation unit 11 at the bottom left displays the pre-created concept map of one of the learner pair U1 and U2, and the second pre-work content presentation unit 12 at the top left displays the pre-created concept map of the other of the learner pair. For example, in the client device 210 of learner U1, the first pre-work content presentation unit 11 displays the pre-created concept map of learner U1, and the second pre-work content presentation unit 12 displays the pre-created concept map of learner U2. Meanwhile, in the client device 220 of learner U2, the first pre-work content presentation unit 11 displays the pre-created concept map of learner U2, and the second pre-work content presentation unit 12 displays the pre-created concept map of learner U1.

[0068] The pre-created concept maps are stored in the server 110 and are automatically loaded and displayed from the server 110. Specifically, the concept map information (sentences and display positions of nodes and links) created independently by each learner U1, U2 is sent from the client devices 210, 220 to the server 110, and when used in a collaborative situation, the server 110 is queried and the map is displayed in the appropriate location.

[0069] During collaborative learning between learner pair U1 and U2, the concept maps displayed in pre-work content presentation units 11 and 12 are for reference only and are not edited. By displaying both pre-created concept maps of learner pair U1 and U2, each learner pair U1 and U2 can grasp their own and the other's thoughts and understanding during the individual phase, facilitating mutual understanding.

[0070] The screen interface 10, which serves as the work environment 10, includes a collaborative drawing area 13A. In FIGS. 3(A) and 3(B), the collaborative drawing area 13A is located at the bottom center of the screen interface 10. The collaborative drawing area 13A is an area where a pair of learners U1 and U2, who are users, can collaboratively draw together. As an example, the collaborative drawing area 13A is a collaborative concept map drawing area.

[0071] The learner pair U1 and U2 refer to each other's pre-created concept maps and converse in the chat area 13B (described later) to collaboratively create a concept map for the same topic as the pre-created concept map. In the collaborative drawing area 13A, each of the learner pair U1 and U2 can simultaneously create and edit a concept map in real time. The collaborative drawing areas 13A of the client devices 210 and 220 of the learner pair U1 and U2 are synchronized, and the same content is displayed.

[0072] The concept map drawing function in the collaborative drawing area 13A is provided by the concept map drawing module 150 shown in Figure 1. In the system 100, learners U1 and U2 use the chat area 13B (described later) to carry out learning activities by creating a concept map while collaboratively explaining things to each other.

[0073] The concept map drawing function allows you to create and delete concept map elements (nodes, links, link labels) in the drawing area 13A, and these contents are synchronized and displayed in real time on your and your collaborators' screens.

[0074] When creating nodes and links, learners U1 and U2 do not freely enter their label names, but select them in a pull-down menu from a pre-prepared list (e.g., for nodes, "internal," "external," "anxiety," etc.). The concept map drawing module 150 manages the data required for the generation of concept maps by the two learners U1 and U2 and the agent 120 (student agent 122), and transfers the information to module 230 of the client devices 210 and 220 via the server 110.

[0075] Two learners U1 and U2 collaboratively create a concept map in the collaborative drawing area 13A. Each learner pair, U1 and U2, works together to create a collaborative concept map, referring to their own pre-created concept maps displayed in the presentation areas 11 and 12. When one learner adds or modifies nodes and links, the concept map information (document and location information) is transmitted via the server to the other learner's collaborative concept map. The drawing content is displayed using server-client communication, allowing learners U1 and U2 to progress through the assignment while checking their additions and modifications in near real time. Furthermore, all events (such as adding concept maps or changing the concept map's location) are logged and saved on the server, allowing for analysis of learners U1 and U2's assignment history.

[0076] The screen interface 10, which serves as the work environment 10, includes a chat area 13B. In FIGS. 3(A) and 3(B), the chat area 13B is located at the bottom right of the screen interface 10. The chat area 13B is an area where users, learners U1 and U2, can have a conversation, for example, text-based. The chat area 13B displays the conversations between the learner pair U1 and U2. The chat area 13B is provided with an input section 13C for inputting one's own conversation text.

[0077] The chat area 13B also displays comments from the agent 120. The comments from the agent 120 may include, for example, advice from the teacher agent 121. The comments from the agent 120 may also include comments from the student agent 122 (such as filling in a concept map). The comments such as advice from the teacher agent 121 are generated by the facilitation presentation module 160. The comments from the student agent 122 are generated by the facilitation presentation module 140.

[0078] Regarding the facilitation presentation module 160, in this embodiment, the facilitation presentation module 160 of the present system 100 was newly developed based on the design of the educational conversational agent (ESPA (Explanation Support by Pedagogical Agent)) shown in "Hayashi Yugo, Inoue Tomoo, "Design of Collaborative Learning with Multiple Educational Conversational Agents," IEICE Transactions on Education, Information and Communication Engineers, Vol. J98-A, No. 1, pp. 76-84, 2015." and "Y. Hayashi, "Gaze awareness and metacognitive suggestions by a pedagogical conversational agent: an experimental investigation on interventions to support collaborative learning process and performance," International Journal of Computer-Supported Collaborative Learning, Vol. 15, pp. 469-498, Dec. 2020."

[0079] This facilitation presentation module 160 has a mechanism for presenting metacognitive facilitation prompts. The facilitation presentation module 160 extracts important keywords from the text in the chat area 13B obtained via the server 110, and generates metacognitive facilitation prompts to promote explanation activities according to the content. The generated facilitation presentation module 160 is displayed in the chat area 13B by the teacher agent 121 and presented to the learners U1 and U2.

[0080] The facilitation prompts presented to learners are selected from five types based on the aforementioned previous research (Hayashi and Inoue, 2015). Based on findings from learning science and cognitive science, prompts are designed to promote self-regulated learning (SRL) and to encourage learners' metacognitive activities.

[0081] The teacher agent 121 also has a function of summarizing text using the API of GPT3.5, which is a type of LLM 170. The teacher agent 121 summarizes text from conversations between the learner pair U1 and U2 (and the student agent 122) in the chat area 13B using the LLM 170, and outputs the summarized content to the chat area 13B. For example, the teacher agent 121 collects speech data on the content of the conversations between the learners U1 and U2 at one-minute intervals, and sends a query to the LLM 170 to summarize the text.

[0082] The screen interface 10, which serves as the work environment 10, includes a presentation unit 15 for presenting content generated by a second agent 122, which is a student agent 122. In FIGS. 3(A) and (B), the presentation unit 15 is located at the top center of the screen interface 10. The second agent 122, which is a student agent 122, monitors the drawing content in the collaborative drawing area 13A, learns the drawing content in real time based on the drawing content, and generates a new concept map by inferring using the knowledge base that the student agent 122 possesses. The concept map generated by the student agent 122 is displayed on the presentation unit 15.

[0083] The student agent 122 plays the role of a third learner from the perspective of the learner pair U1 and U2. The student agent 122 plays the role of a member of collaborative learners together with the learner pair U1 and U2, and learns and recognizes the content of the collaborative concept map created collaboratively by the learner pair U1 and U2. The student agent 122 uses its knowledge base to generate a new concept map recognized by the student agent 122 from the content of the collaborative concept map.

[0084] By displaying the concept map generated by the student agent 122 on the presentation unit 15, the learner pair U1 and U2 can understand how the student agent 122 perceives the concept map they created. In other words, the student agent 122 provides support to the learner pair U1 and U2 by providing them with reflections on their collaborative work (horizontal support).

[0085] The concept map is generated by the student agent 122 using the knowledge generation module 130. In order to implement an AI system that generates knowledge similar to that of a human learner, the concept map is generated using a cognitive architecture that is considered useful for simulating human knowledge acquisition and knowledge generation on a computer. As an example of a cognitive architecture, ACT-R, a representative model of cognitive architecture, is used (JR Anderson, Hayashi Yugo (translation), Cognitive Modeling: Understanding the Mind Based on ACT-R Theory, Kyoritsu Shuppan, Tokyo, 2021).

[0086] In this embodiment, (Y. Hayashi and S. Shimojo, “Modeling perspective taking and knowledge use in collaborative explanation: Investigation by laboratory experiment and computer simulation using act-r,” Artificial Intelligence in Education, eds. by MM Rodrigo, N. Matsuda, AI Cristea, and V. Dimitrova, pp.647-652, Springer International Publishing, Cham, 2022, hereinafter referred to as “Hayashi(2022a)”) and (Y. Hayashi and S. Shimojo, “Relevant knowledge use during collaborative explanation activities: Investigation by laboratory experiment and computer simulation using act-r,” Collaboration Technologies and Social Computing, eds. by L.-H. Wong, Y. Hayashi, CA Collazos, C. Alvarez, G. Zurita, and N. Baloian,pp.52-66,Springer International We constructed a model for automatically generating concept maps as the knowledge generation module 130 using a computer simulation model for the generation of learners' concept maps, which was constructed based on Hayashi (2022b).

[0087] The support system 100 implements this concept map generation model on ACT-R, and the model performs inference while communicating with the server 110. Furthermore, the node and link information of the concept map collaboratively created by the learner pair U1 and U2 is transmitted to ACT-R via the server 110. The ACT-R infers and outputs knowledge about the nodes and links for generating the concept map. The output knowledge is transmitted to the client devices 210 and 220 via the server 110, where the concept map is then drawn. It should be noted that this model does not simply generate concept maps identical to those of learners U1 and U2. Rather, it generates concept maps based on the learners' typical knowledge obtained from prior knowledge content obtained in previous studies (Hayashi (2022a), Hayashi (2022b)), resulting in the generation of knowledge for the concept map with a certain degree of freedom.

[0088] Below, we provide a supplementary explanation of the mechanism and operating procedures of the concept map generation model of the embodiment. First, the knowledge of the model described here is loaded as a knowledge base with the learner knowledge obtained from experiments in previous research (Hayashi (2022a), Hayashi (2022b)). In other words, the knowledge of good, bad, or average learners from experiments in previous research can be described in the model, making it possible to prepare AI agents with knowledge of specific learner demographics.

[0089] The model is designed to process in four stages: (1) referencing map information (links and nodes), (2) searching its own knowledge base based on the referenced information, (3) retrieving and saving (learning) the knowledge to be used, and (4) drawing a concept map. The design and description of this stage was done by writing a unique program using the LISP language within the ACT-R system.

[0090] In the above "(1) See Map Information (Links and Nodes)," the ACT-R model processes information obtained from the second agent 122, the student agent 122. The server 110 sends the link and node information of the concept map to the student agent 122, which handles the transmission and reception to and from ACT-R.

[0091] The student agent 122 converts the link and node information into a format that ACT-R can read and sends it to the ACT-R board (interface). The ACT-R model perceives the contents of the node and link information written on the board and reads that content. The model performs actions similar to human perceptual processing, such as (a) paying attention to the board, (b) reading any characters if any are there, and (c) searching for others.

[0092] Next, in "(2) Searching own knowledge base based on referenced information," the obtained node and link information is judged to be knowledge that exists in the knowledge (chunks) described in the model. Note that the knowledge described in the model (knowledge base) here is loaded with the learner's knowledge obtained in the past experiments mentioned above.

[0093] In "(3) Recalling and Saving Knowledge to Use (Learning)," the process differs depending on whether the knowledge examined in (2) above exists. If the knowledge exists, the process proceeds to (4), which uses that knowledge to proceed to drawing. If the knowledge does not exist, the agent determines whether to use the new knowledge or its own knowledge. This decision is described in the form of a different production rule, and which production rule is used is determined by manipulating the agent's degree of egocentrism. Analysis of this egocentrism was conducted in Hayashi (2022a), Hayashi (2022b), showing that learners with high egocentrism only refer to their own knowledge, while learners with low egocentrism prioritize using new knowledge received from collaborators.

[0094] Experimental results have shown that the latter approach allows for smoother communication and mutual understanding during collaborative learning. The ACT-R model of this embodiment reproduces this point and is designed to be flexibly configured to operate in either (i) a case of high egocentrism or (ii) a case of low egocentrism. In other words, there are cases where knowledge from the student agent 122 is actively used (low egocentrism) and cases where knowledge from the student agent 122 is not actively used (high egocentrism), and the knowledge adopted by the model changes depending on the mode. While the production rules were developed independently, the algorithm used to ignite knowledge is implemented using the utility functions within ACT-R.

[0095] Then, in the final stage of "(4) Drawing a Concept Map," the knowledge determined in (3) above is written on a board used to transmit it to the student agent 122.

[0096] The screen interface 10, which serves as the work environment 10, includes avatar display units 17A and 17B. In FIGS. 3A and 3B, the avatar display units 17A and 17B are provided in the upper right corner of the screen interface 10. In the upper right corner, the lower part is the avatar display unit 17A of the first agent 121, which is the teacher agent 121, and the upper part is the avatar display unit 17B of the second agent 122, which is the student agent 122. The avatar display units 17A and 17B display the faces of the teacher agent 121 and the student agent 122. The faces of the agents 121 and 122 are configured to move when the agents 121 and 122 make a statement in the chat area 13B or when an entry is made in the concept map presentation unit 15. This allows the learners U1 and U2 to distinguish and recognize which agent is acting. It is preferable that the agents 121 and 122 are displayed in a distinguishable manner, but they do not have to be distinguished. For example, only one agent 120 that plays the roles of both teacher and student may be displayed.

[0097] <2.5 Examples of collaborative learning>

[0098] 4 to 14 show examples of joint learning (collaboration) between learners U1 and U2 using the support system 100 of the embodiment.

[0099] FIG. 4 shows the screen interface 10 in its initial state (step S1), i.e., at the start of collaborative learning. In the initial state, the presentation units 11 and 12 display the pre-created concept maps of the learners U1 and U2. Nothing is drawn in the collaborative drawing area 13A, and the concept map displayed on the presentation unit 15 by the student agent 122 is also blank. There is no conversation in the chat area 13B. In the examples of FIGS. 4 to 14, only one agent 120, who plays the roles of both teacher and student, is displayed on the avatar display unit 17.

[0100] Figure 5 shows an enlarged view of the pre-created concept maps in the presentation units 11 and 12 in Figure 4. The pre-created concept maps show concepts that learners U1 and U2 have created in advance on the theme of "Anxiety: I'm worried about the new semester." Note that in Figure 6 and subsequent figures, the presentation units 11 and 12 are sometimes omitted (due to space constraints), but in reality, the presentation units 11 and 12 continue to be displayed so that learners U1 and U2 can refer to them during the collaborative work process.

[0101] FIG. 6 shows the collaborative work process A (step S2) after the initial state. To create a collaborative concept map, first, an initial node 51 is created. As shown in FIG. 6, the label of node 51 is selected from a pull-down menu 19A. The creation of node 51 is performed, for example, by learner U1 (user1). The options in pull-down menu 19A are generated, for example, from the labels of nodes included in a pre-created concept map.

[0102] 7 shows collaboration process B (step S3) after collaboration process A. In step S3, the label “Anxiety: “I’m worried about the new school year”” selected in step S2 is displayed in node 51.

[0103] 8 shows collaborative work process C (step S4) after collaborative work process B. Between steps S3 and S4, learner U1 (user1) and learner U2 (user2) are having a conversation in chat area 13B while additionally creating nodes 52 and 53. Node 52 was created by learner U1 (user1), and node 53 was created by learner U2 (user2).

[0104] FIG. 9 shows collaborative work process D (step S5) after collaborative work process C. In step S5, learner U1 adds link 54 connecting node 51 and node 53 that he or she created, and is trying to determine label 55 for link 54. The label for link 54 is also selected from pull-down menu 19B. Options in pull-down menu 19B include "internal," "external," and "anxiety," among others.

[0105] 10 shows collaborative work process E (step S6) after collaborative work process D. As shown in FIGS. 4 to 9, the collaborative concept map created up to step S5 does not display the concept map generated by agent 120. This is because the collaborative concept map is still incomplete and agent 120 has not yet reached the point of inferring the knowledge that should be presented to the user.

[0106] In response to this, when the link 54 and its label 55 are created in step S5, the concept map generated by the agent 120 is displayed on the presentation unit 15 as shown in FIG.

[0107] The concept map displayed on the presentation unit 15 is drawn by the agent 120, who searches its knowledge base based on the contents of the collaborative concept map created in the collaborative drawing area 13A, determines the knowledge to present to the learners U1 and U2, and then draws it based on the determined knowledge. The concept map displayed on the presentation unit 15 is not the same as the concept map in the collaborative drawing area 13A, but rather shows the content that the agent 120 has learned and recognized based on the knowledge that the agent 120 possesses. Therefore, by referring to the concept map created by the agent 120, the learners U1 and U2 can understand how the agent 120 understands the collaborative concept map that they are creating. In other words, they can reflect on the collaborative concept map (horizontal support by the agent 120).

[0108] 11 shows collaborative work process F (step S7) after collaborative work process E. From step S6 to step S7, the content of the collaborative concept map remains unchanged, but knowledge inference by agent 120 progresses and the concept map generated by agent 120 becomes more complete. In other words, inference progresses and the amount of knowledge presented to the user by agent 120 increases. This allows learners U1 and U2 to further reflect on their work (horizontal support by agent 120).

[0109] 12 to 14 show collaborative work process G (step S8) after collaborative work process F. As learners U1 and U2 converse in chat area 13B and proceed with drawing a collaborative concept map in collaborative drawing area 13A, agent 120 generates and updates the concept map in real time and presents it to learners U1 and U2 on presentation unit 15. FIG. 13 shows an enlarged view of presentation unit 15 and collaborative drawing area 13A in FIG. 12.

[0110] FIG. 14 shows an enlarged view of the chat area 13B in FIG. 12. As shown in FIG. 14, the chat area 13B contains not only comments by learner U1 (user1) and learner U2 (user2), but also comments (facilitation) such as advice from the agent 120. In FIG. 14, examples of prompts are displayed, such as, "[Agent] Please refer to the information in the examples and learning text and rephrase the content in your own words.", "[Agent] Please summarize the information in the examples and learning text.", and "[Agent] Please add the information in the examples and learning text to the concept map." This allows learners U1 and U2 to proceed with their work while receiving suggestions and advice from the agent 120 (vertical support by the agent 120).

[0111] When the agent 120 changes the content of the presentation unit 15, the agent 120 can make a statement to that effect and / or the details of the change in the chat area 13B. The agent 120 can also summarize the conversation in the chat area 13B and display the summary in the chat area 13B. The summary may include not only the statements of the learners U1 and U2 but also the statements of the agents 120 playing the roles of teacher and student. The agent 120 acting as a teacher may also summarize the statements of the learners U1 and U2 and the student agent 122 and display the summary in the chat area 13B.

[0112] The present invention is not limited to the above-described embodiment, and various modifications are possible. For example, the support system 100 can be used not only for collaborative work such as collaborative learning by multiple users, but also for work such as learning by a single user. In other words, it can be used even if there is no other person to study with.

[0113] Furthermore, the support system 100 is not limited to use in learning situations, but may also be used in intellectual activities other than learning, such as discussions or brainstorming.

[0114] Furthermore, the diagram drawn in the collaborative drawing area 13A is not limited to a concept map, but may be any other type of diagram that expresses knowledge. [Explanation of symbols]

[0115] 10: Work environment (screen interface) 11: 1st preliminary work content presentation section 12: 2nd preliminary work content presentation section 13A:Drawing area (collaborative drawing area) 13B: Chat area 13C: Input section 15: Presentation part 17: Avatar display area 17A: Avatar display section 17B: Avatar display section 19A: Pull-down menu 19B: Pull-down menu 51: Node 52: Node 53: Node 54: Link 55: Label 100: Support System 101: Processor 102: Memory 102A: Computer Programs 103: Communication equipment 110: Server 111: Database 121: First Agent 122: Second Agent 130: Knowledge generation module 140: Facilitation Presentation Module 150: Concept map drawing module 160: Facilitation Presentation Module 201: Processor 202: Memory 202A: Computer Programs 203: Communication equipment 210: Client device 220: Client device 230: Client module A: Collaborative work process B: Collaborative work process C: Collaborative work process D: Collaborative work process E: Collaborative work process F: Collaborative work process G: Collaborative work process U1: Learner U2: Learner

Claims

1. A support system that provides a work environment for a user and supports the user's work, an agent that provides vertical support to the user and horizontal support to the user; For the vertical assistance, the agent is configured to output suggestions related to the user's work; For the horizontal support, the agent is configured to search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to be presented to the user. Support system.

2. The working environment is a drawing area where multiple users can work together, including collaborative drawing; a chat area in which the plurality of users can chat; Including, The assistance system according to claim 1 .

3. the work environment includes a chat area where chatting is possible; For the vertical assistance, the agent is configured to output the suggestions to the chat area. The assistance system according to claim 1 .

4. the work environment includes a drawing area where work can be performed, including drawing; For the horizontal support, the agent is configured to search a knowledge base held by the agent based on the content of the drawing in the drawing area and determine knowledge to be presented to the user. The assistance system according to claim 1 .

5. the work environment includes a drawing area where work can be performed, including drawing a concept map; For the horizontal support, the agent is configured to search a knowledge base held by the agent based on the concept map in the drawing area, determine knowledge to be presented to the user, and present the concept map drawn based on the determined knowledge to the user. The assistance system according to claim 1 .

6. the work environment includes a chat area in which multiple users can chat; For the vertical assistance, the agent is configured to present a summary of the content of the chat to the user. The assistance system according to claim 1 .

7. The system further includes a pre-work content presenting unit that presents the pre-work content of the user to the user. The assistance system according to claim 1 .

8. The user's task is collaborative learning by multiple users. The assistance system according to claim 1 .

9. The agent: a first agent for the vertical support; a second agent for the horizontal support; Including, A first avatar representing the first agent and a second avatar representing the second agent are presented to the user. The assistance system according to claim 1 .

10. 1. A computer-implemented method executed by a computer to assist a user in a task, comprising: providing a work environment for said user; An agent provides vertical support to the user and also provides horizontal support to the user. Prepare for this. For the vertical assistance, the agent outputs suggestions related to the user's work; For the horizontal support, the agent searches a knowledge base that the agent has based on the work content in the work environment and determines knowledge to present to the user. Computer-implemented methods.

11. A computer program for operating a computer as a support system that provides a work environment for a user and supports the user's work, comprising: the support system includes an agent that provides vertical support to the user and horizontal support to the user; For the vertical assistance, the agent is configured to output suggestions related to the user's work; For the horizontal support, the agent is configured to search a knowledge base held by the agent based on the work content in the work environment and determine knowledge to be presented to the user. Computer program.

Images