Multi-agent-based group tutoring automation platform

Abstract

The invention discloses a Multi-agent-based group tutoring automation platform, the Multi-agent comprises a main leader agent, a planning leader agent and a plurality of group member agents, and the working process of the group tutoring automation platform is as follows: determining a base version of a large language model according to actual business logic and a use scene; each agent is configured through a JSON file; testing psychological scales of members participating in group tutoring before group tutoring; team tutoring is started; testing psychological scales of the members participating in group tutoring after group tutoring; and storing the short-term memory and the long-term memory in the whole team tutoring process and outputting the short-term memory and the long-term memory as a JSON file.

Description

Technical Field

[0001] This application relates to the field of Multi-agent large models, and particularly to a group counseling automation platform based on Multi-agent. Background Art

[0002] After the emergence of large models, the paradigm of natural language processing tasks has been changed. In terms of psychological computing for users, the LLM and agent technologies are mainly adopted. Large language models are famous for their general capabilities, extensive knowledge, and reasoning abilities, providing more flexible and convenient methods for psychological counseling and group psychological counseling. Although great progress has been made in the research of multi-agent in both the research and industrial communities, including social simulation, intelligent assistants, Werewolf games, World War simulations, urban planning simulations, etc., in the field of psychological counseling, there are few related studies on using multi-agent to simulate group counseling. Therefore, it is proposed to use multi-agent to simulate group counseling. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art, provide a group counseling platform based on Multi-agent, and provide a fully automated group counseling platform and a group counseling platform with human participation.

[0004] The purpose of the present invention is achieved through the following technical solutions:

[0005] A group counseling automation platform based on Multi-agent, where Multi-agent includes a main leader agent, a planning leader agent, and several group member agents. The work process of the group counseling automation platform is as follows:

[0006] S1. Determine the base version of the large language model according to the actual business logic and usage scenarios;

[0007] S2. Configure each agent through a JSON file;

[0008] S3. Test the psychological scale of the members participating in the group counseling before the group counseling;

[0009] S4. Start the group counseling;

[0010] S5. Test the psychological scale of the members participating in the group counseling after the group counseling;

[0011] S6. Save the short-term memory and long-term memory during the entire group counseling process and output them as a JSON file.

[0012] Further, step S4 is the core of the entire group counseling process, including:

[0013] S401. The planning leader agent designs a group counseling plan, and the main leader agent receives the counseling plan from the planning leader agent;

[0014] S402. The planning leader agent reminds the main leader agent of the next activity content and communication form, and the main leader agent conducts group counseling for this step according to the plan of the planning leader agent;

[0015] S403. If the communication form is taking turns to speak, the main leader agent guides the group member agents to take turns to speak in the specified order in this stage;

[0016] S404. If the communication form is free speech, the main leader agent guides the group member agents to act and speak freely in the specified order in this stage. The actions include selecting another group member agent to talk to;

[0017] S405. If the communication form is taking turns to speak and free speech, the main leader agent guides the group member agents to take turns to speak in the specified order in this stage, and guides the group member agents to act and speak freely in the specified order. The actions include selecting another group member agent to talk to;

[0018] S406. Judge whether the group counseling is over according to the group counseling plan. If the group counseling is over, the whole process ends;

[0019] S407. If the group counseling has not ended yet, the planning leader agent summarizes the current situation of the group counseling and considers whether it is necessary to adjust or refine the next counseling plan, and continues with step 402.

[0020] Furthermore, each agent has a set personality, memory and behavior.

[0021] Furthermore, the agent personality includes basic attributes, interpersonal communication attributes, role-playing attributes, and other attributes; among them, the basic attribute settings include: name, gender, age, educational background, occupation, personality, interests and specialties, interpersonal relationships, growth experience, participation motivation, whether there is group experience, willingness to participate in the group, personality type; interpersonal communication attributes include: self-awareness, interpersonal attitude, interpersonal perception; role-playing attributes include: language characteristics, emotional expression, interaction mode; other attributes include extroversion index, problems faced.

[0022] Furthermore, the agent memory includes short-term memory and long-term memory; the short-term memory is the memory of the agent's speech within one round and the speech of the agents that can be perceived. If the agent often encounters similar external discoveries, the short-term memory is updated to long-term memory; the long-term memory takes into account the long-term memory based on the personality type, and updates the long-term memory of this round according to the long-term memory of the previous round and the short-term memory of this round, so as to take into account the memory preferences of different agents.

[0023] Furthermore, the behaviors of the agents include the behaviors owned by each role of the agents; among them, the behaviors of the group member agents include thinking, choosing, speaking, and updating memory; the behaviors of the main leader agents include thinking, choosing, speaking, and updating memory; the behaviors of the planning leader agents include planning, speaking, and updating memory.

[0024] The present invention also provides an application based on the group counseling automation platform, which can be applied to psychological research, education and training, and mental health services.

[0025] Compared with the prior art, the beneficial effects brought by the technical solution of the present invention are:

[0026] 1. Flexibility and adaptability: Through the multi-agent system, the platform can simulate participants with different roles. Each agent has unique personalities, memories, behavior patterns, etc., and can more accurately simulate and reflect the performance and interaction of group members in psychological counseling. This flexibility enables the platform to adapt to different types of counseling scenarios, such as group counseling for individual users or the training needs of leaders.

[0027] 2. Expansion of psychological theories: By simulating the psychological dynamics of different roles and the interaction between group members, this platform can provide theoretical support for group counseling in psychology, helping researchers further enrich and improve relevant theoretical models. At the same time, the platform can also help psychological workers observe and analyze the behavior performance of each participant in group counseling, providing new data and insights for the field of psychological counseling.

[0028] 3. Efficient counseling process: The platform can automatically guide the group counseling process according to the set counseling plan, simplifying the counseling process and improving the counseling efficiency. Through the memory and behavior patterns of the agents, the platform can dynamically adjust the group counseling plan according to the progress of each round, ensuring the coherence and depth of the counseling activities.

[0029] 4. Leader training and development: This platform is particularly suitable for the training of group counseling leaders. By simulating different group counseling scenarios, leaders can exercise and improve their ability to guide and manage teams in the platform, enhancing their leadership skills and decision-making abilities in real counseling environments.

[0030] 5. Personalized tutoring experience: The agents in the platform have personalized settings and can simulate the personalities and psychological characteristics of different group members, thus providing a tutoring experience that better meets individual needs during group tutoring and further enhancing the tutoring effect.

[0031] 6. Innovative application scenarios: The present invention brings innovative technological applications to the psychological counseling industry and can be applied to multiple fields such as psychological research, education and training, and mental health services, broadening the technological application boundaries of psychological counseling.

[0032] In summary, the automated group tutoring platform based on Multi-agent not only improves the efficiency and flexibility of group tutoring, but also provides a new technological support for psychological theories and practical applications, with great development potential and practical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is the overall workflow diagram of the group tutoring platform implemented based on Multi-agent of the present invention;

[0034] Figure 2 is the specific flowchart of the core steps in the group tutoring platform of the present invention;

[0035] Figure 3 is the flowchart of designing the group tutoring plan of the present invention;

[0036] Figure 4 is the flowchart of the turn-taking speech stage of the present invention;

[0037] Figure 5 is the flowchart of the free speech stage of the present invention;

[0038] Figure 6 is the structure diagram of the agent of the group tutoring platform implemented based on Multi-agent of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] Next, the technical solutions in the embodiments of the invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the invention. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0040] To make the above objects, features, and advantages of the present application more obvious and understandable, the fully automated group tutoring platform in the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0041] In an exemplary embodiment, as Figure 1As shown, a Multi-agent-based intelligent platform for group counseling is provided. The workflow of this intelligent platform for group counseling is executed by a computer device, which can be specifically executed by a computer device such as a terminal or a server alone, or jointly executed by a terminal and a server, and includes the following steps S1 to S6;

[0042] S1. Determine the base version of the large language model according to the actual business logic and usage scenarios. In this embodiment, the qwen-max and gpt-4 large language models are adopted.

[0043] S2. Configure the agent through a JSON file, or directly use agentscope.init to construct the agent while initializing the model configuration. Each agent is given a unique personality, which is achieved through sys_prompt; each agent has its own memory, which is achieved through use_memory; each agent is configured with a large language model, which is achieved through model_config_name; each agent has a different role, which is achieved through class; each agent has its own name, which is achieved through name. Before specific group counseling applications, it is necessary to modify the JSON configuration file according to the theme definition of group counseling. As an example, a simple description of a single agent is as follows:

[0044] sys_prompt: Name: Zhang San; Gender: Male; Age: 36; Personality type: ENTJ (Extraverted, Intuitive, Thinking, Judging); Educational background: Bachelor's degree; Management major; Occupation: Middle-level manager in a company; Personality: Serious and responsible, good at strategic planning and team building; Interests and specialties: Managing teams, project planning, public speaking; Interpersonal relationships: Has established good colleague relationships in the workplace, but family relationships are strained due to work pressure; Growth experience: Has gradually been promoted to a middle-level management position in a highly competitive workplace environment; Motivation for participation: Looking for ways to cope with work pressure to balance work and life; Whether there is any group counseling experience: No; Willingness to participate in a group: 7 / 10, is open to it; Self-awareness: Thinks he has strong abilities in management and team building, but faces challenges in work-life balance; Interpersonal attitude: Remains professional and direct in the workplace, but may seem too serious at home; Interpersonal perception: Can keenly capture team dynamics, but may not be sensitive enough in family communication; Language characteristics: Direct, highly logical, emphasizing facts and efficiency; Emotional expression: Less, tends to hide personal emotions, especially in the workplace; Interaction mode: Tends to dominate discussions and likes to control and guide the direction of communication; Extraversion index: 8 / 10, is active and confident in social and work situations; Problems faced: Long working hours, frequent overtime, difficulty in refusing additional tasks at work, resulting in physical symptoms such as headaches and stomachaches, difficulty concentrating, decreased work efficiency, and feelings of frustration. use_memory: true; model_config_name: qwen; class: MemAgent; name: Zhang San.

[0045] S3. Test the psychological scale of the members participating in group counseling before the group counseling. As an example, the prompt for the psychological scale test is as follows:

[0046] lb_PROMPT1 = "″Please carefully read each of the following items and, based on your situation in the past week, select the degree option that applies to you in each item. The options are: Disagree, Sometimes agree, Often agree, Always agree. Please answer each item; there are no right or wrong answers. 1. I find it difficult to calm myself down. 2. I tend to have allergic reactions to things. 3. I feel that I consume a lot of energy. 4. I feel restless. 5. I find it difficult to relax myself. 6. I can't tolerate anything that hinders me from continuing to work. 7. I find myself easily irritated.″"

[0047] S4. Start the group counseling, which is the core of the entire group counseling process. Specifically in this example, as Figure 2 shown in the flowchart, step S4 includes the following steps:

[0048] S401. The planning leader designs a group counseling plan. The main leader receives the counseling plan from the planning leader. As shown in the flowchart in Figure 3 , in this embodiment, the prompt for the planning leader to design a group counseling plan is as follows:

[0049] PLAN_PROMPT = "″Design a one-time group counseling plan with {number} steps on {group_topics}. The plan includes the activity content and communication form for each step. The communication form is either taking turns to speak or free speech. All participants are {leader_and_member_names}."″

[0050] S402. The planning leader reminds the main leader of the next activity content and communication form. The main leader conducts group counseling for this step according to the plan of the planning leader. In this embodiment, the prompt for the planning leader to remind the main leader of the next activity content and communication form is as follows:

[0051] PLANREMIND_PROMPT = "″The planning leader reminds the executing leader of the next activity content and communication form to ensure the smooth progress of the counseling."″

[0052] S403. If the communication form is taking turns to speak, the main leader guides the group members to take turns to speak in the specified order during this stage. The speaking order is from the highest to the lowest extroversion score. As shown in the Figure 4 flowchart of taking turns to speak.

[0053] S404. If the communication form is free speech, the main leader guides the group members to act in the specified order and have free speech during this stage. The actions mainly include selecting another group member to talk to. The action order is from the highest to the lowest extroversion score. As shown in the Figure 5 flowchart of free speech. The prompts for thinking, choosing, and talking during this process are as follows:

[0054] think_PROMPT = "″Please select the name of a member you wish to discuss from the following list of candidate participants: {candidates}. You can determine the most appropriate discussion topic by reviewing the speech content of team members and select a member to communicate accordingly. Please finally show your thinking process, chat topic, and the selected member. All participants are {lm_names}."″

[0055] choose_PROMPT = "″Please accurately fill in the name of the group member you expect to communicate with. The available participant names are {lm_names}."″

[0056] speak_PROMPT = "\"Please have a conversation with the group members you have chosen to communicate with.\""

[0057] S405. If the communication forms are turn - taking speech and free speech, the main leader guides the group members to take turns speaking in the specified order during this stage, and guides the group members to act and make free speeches in the specified order. The actions mainly include selecting another group member to have a conversation with. The turn - taking speech order is to speak from the highest to the lowest extroversion score, and the free speech action order is to speak from the highest to the lowest extroversion score.

[0058] S406. Judge whether the group counseling is over according to the group counseling plan. If the group counseling is over, the whole process ends.

[0059] S407. If the group counseling has not ended, the planning leader summarizes the current situation of the group counseling and considers whether to adjust or refine the next counseling plan, and then continues with step S402. In this embodiment, the prompt for the planning leader to consider adjusting or refining the next counseling plan and to remind the main leader of the next activity content and communication form is as follows:

[0060] PLANREMIND_PROMPT = "\"The planning leader needs to summarize the situation of the previous round of group counseling, and adjust or refine the next group counseling plan based on the summary results. The execution leader also needs to be reminded of the next activity content and communication form. The communication form is turn - taking communication or free communication. The specific steps include: (1) Briefly summarize the situation of the previous round of group counseling, including the successful aspects and areas for improvement. (2) Based on the summary results, consider whether to adjust or refine the next group counseling plan. (3) Remind the execution leader of the next activity content and communication form to ensure the smooth progress of the counseling.\""

[0061] S5. Conduct a psychological scale test on the members participating in the group counseling. As an embodiment, the prompt for the psychological scale test is as follows:

[0062] lb_PROMPT2 = "\"After the group counseling ends, based on the knowledge learned during the group counseling process and your own thinking, carefully read each of the following items and select the degree option that applies to your situation. The options are as follows: Disagree, Sometimes agree, Often agree, Always agree. Please answer each item. There is no right or wrong in the selection. 1. I find it difficult to calm myself down. 2. I tend to have an allergic reaction to things. 3. I feel that I consume a lot of energy. 4. I feel nervous. 5. I feel it difficult to relax myself. 6. I can't tolerate anything that hinders me from continuing to work. 7. I find myself easily irritated.\""

[0063] S6. Save the short-term and long-term memories during the entire group counseling process and output them as a JSON file for convenient later data statistics and analysis.

[0064] The structure of the agent in the group counseling embodiment is used for setting the roles of each participant in the group counseling. For example, Figure 6 as shown, each agent has a set personality, memory, and behavior.

[0065] The agent personality includes basic attributes, interpersonal communication attributes, role-playing attributes, and other attributes. Among them, the basic attribute settings include: name, gender, age, educational background, occupation, personality, interests and specialties, interpersonal relationships, growth experience, participation motivation, whether there is group experience, willingness to participate in the group, personality type; interpersonal communication attributes include: self-awareness, interpersonal attitude, interpersonal perception; role-playing attributes include: language characteristics, emotional expression, interaction mode; other attributes include extroversion index, problems faced.

[0066] The agents in the group counseling in the embodiment have different roles, including the main leader agent, the planning leader agent, and the group member agent. Each different role has a different agent class designed. The agent base class provides a comprehensive template containing basic attributes and methods, and these attributes and methods support the core functions of any conversation agent. Each derived class consists of several key features:

[0067] memory: This attribute enables the agent to retain and recall past interactions, allowing them to maintain context in an ongoing conversation.

[0068] model: The model is the computational engine of the agent, responsible for making responses based on the existing memory and input.

[0069] sys_prompt: The system prompt serves as a predefined instruction to guide the behavior of the agent in its interactions.

[0070] Among them, the main leader uses Main_agent, the planning leader uses Plan_agent, and the group member uses Member_agent.

[0071] Agent memory includes short-term memory and long-term memory. Short-term memory stores an agent's own speech and the speech of other agents that it can perceive within one turn. If an agent frequently encounters similar external discoveries, the short-term memory is updated to long-term memory. Long-term memory takes into account long-term memory based on personality type, and updates the long-term memory of this turn based on the long-term memory of the previous turn and the short-term memory of this turn. This memory method takes into account the memory preferences of different agents.

[0072] In the agent, memory is used to store historical messages, enabling the agent to provide more coherent and natural responses based on the context. In short-term memory, each memory includes the following key attributes:

[0073] name: This attribute represents the originator of the information. This is a key piece of metadata that is very useful in scenarios where it is necessary to distinguish between different speakers.

[0074] content: The content of the information itself. It can include text, structured data, or other forms of content related to the interaction that the agent needs to process.

[0075] timestamp: A timestamp indicating the time when the information was created.

[0076] id: Each piece of information is assigned a unique identifier ID when it is created.

[0077] The Memory class implements the processing of an agent's memory in a structured way. The main methods of this class are as follows:

[0078] get_memory(): This method is responsible for retrieving the stored information from the agent's memory.

[0079] add(): This method is used to add new information to the agent's memory.

[0080] load(): This method allows for batch loading of information from an external source into the agent's memory.

[0081] export(): This method facilitates the export of the stored information from the agent's memory.

[0082] In the agent, summary is used for long-term memory based on personality type. The Summary class implements the processing of an agent's long-term memory in a structured way. The new methods of this class compared to the Memory class are as follows:

[0083] updatesummary(): This method facilitates the update of long-term memory.

[0084] Agent behaviors include the behaviors possessed by agents of each role. Among them, the behaviors of group members include thinking, choosing, speaking, and updating memories; the behaviors of the main leader include thinking, choosing, speaking, and updating memories; the behaviors of the planning leader include planning, speaking, and updating memories.

[0085] The behaviors possessed by the agent are introduced as follows:

[0086] think(): This method defines the logic for the agent to think about the person it wants to talk to and give the thinking process.

[0087] choose(): This method defines the logic for the agent to choose the person to talk to.

[0088] speak(): This method defines the response behavior of the agent to the incoming message and encapsulates the logic of the agent's output.

[0089] update(): This method defines the logic for the agent to update its memory.

[0090] In this embodiment, two types of evaluation methods are used to evaluate the model performance, namely automatic evaluation and manual evaluation metrics. The automatic evaluation metrics are described as follows: Before the group counseling starts, psychological scale tests are conducted on group members, and the test result is pre. After the group counseling ends, psychological scale tests are continued on group members, and the test result is after. The total number of all group members participating in the experiment is m, the number of group members whose scale test scores improve in the better direction before and after the group counseling is n, the total output length of the agent is a, the number of times the agent speaks is b, the scale change amount is s, the positive change ratio is r, and the average output length is l.

[0091] Scale change amount: In multiple experiments, calculate the average value of the scale test score change amounts of group members before and after the group counseling. The larger the scale change amount, the better the model effect.

[0092]

[0093] Positive change ratio: In multiple experiments, calculate the proportion of group members whose scale test scores improve in the better direction before and after the group counseling among all group members. The larger the positive change ratio, the better the model effect.

[0094]

[0095] Average output length: Calculate the average output lengths of the planning leader, main leader, and group member agents respectively.

[0096]

[0097] To fill the gap in the evaluation of automated group counseling, four manual evaluation indicators are specifically proposed: credibility, professionalism, participation, and safety. The four dimensions of the evaluation indicators will be listed item by item below, and the scoring range of each dimension is given in the form of [lower bound - upper bound].

[0098] Credibility [0 - 5]: To what extent the behavior of the agent can imitate real group counseling.

[0099] Professionalism [0 - 6]: The professionalism of the leader role-played by the agent during the conversation.

[0100] Participation [0 - 2]: The degree to which group members participate in group counseling and the degree to which the situation of group members is reflected in the conversation.

[0101] Safety [0 - 2]: The correctness of the values of group counseling and the degree of protection of the privacy of group members.

[0102] The results of the automatic evaluation indicators of the automated group counseling role-played by the agent are shown in Table 1 and Table 2:

[0103] Table 1: Automatic evaluation results of ablation experiments

[0104]

[0105]

[0106] Table 2: Average output length of the agent

[0107]

[0108] The manual evaluation indicators of the automated group counseling role-played by the agent are shown in Table 3:

[0109] Table 3: Manual evaluation results of ablation experiments

[0110]

[0111]

[0112] To make the above-mentioned objectives, features, and advantages of this embodiment more obvious and understandable, the group counseling platform with human users involved in this embodiment will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0113] The differences between group counseling with human users involved and fully automated group counseling are as follows:

[0114] Change an agent role-played by a large model in the fully automated group counseling to a human user, where the human user agent (Useragent) is operated by a human to handle the input and output. The speak() method processes the user input through the prompt content, stores the collected data in the agent's memory, and returns the message as a response.

[0115] The results of the automatic evaluation indicators and manual evaluation indicators of group counseling involving human users are shown in Table 4:

[0116] Table 4: Evaluation Indicators of Group Counseling Involving Human Users

[0117]

[0118] The above has described the embodiments of the present invention in detail, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principles and spirit of the present invention, various changes, modifications, substitutions, and variations to these embodiments including components still fall within the protection scope of the present invention.

[0119] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solutions of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the purpose of the present invention and the scope protected by the claims, those of ordinary skill in the art can also make many specific transformations in various forms under the inspiration of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A multi-agent-based group counseling automation platform, characterized in that: Multi-agent includes the main leader agent, the planning leader agent and several group member agents. The workflow of the group counseling automation platform is as follows: S1. Determine the base version of the large language model based on actual business logic and usage scenarios; S2, configure each agent through JSON file; S3. Test the psychological scales of the members who participated in the group counseling before the group counseling; S4, start group counseling; S5. Test the psychological scale of the members who participated in the group counseling after the group counseling; S6. Save the short-term memory and long-term memory of the entire group counseling process and output them as a JSON file.

2. The group counseling automation platform according to claim 1, characterized in that: Step S4 is the core of the entire group counseling process, including: S401, the planning leader agent designs a group counseling plan, and the main leader agent receives the counseling plan from the planning leader agent; S402, the planning leader agent reminds the main leader agent of the next activity content and communication form, and the main leader agent conducts group counseling for this step according to the planning leader agent's plan; S403, if the communication form is to speak in turns, the main leader agent guides the group member agents to speak in turns in the specified order in this stage; S404, if the communication form is free speech, the main leader agent guides the group member agents to act in the specified order and speak freely in this stage, and the action includes selecting another group member agent to communicate with; S405, if the communication form is taking turns to speak and speaking freely, the main leader agent guides the group member agents to speak in turn in the specified order in this stage, and guides the group member agents to act in turn and speak freely in the specified order, and the action includes selecting another group member agent to communicate with; S406, judging whether the group counseling is finished according to the group counseling plan, if the group counseling is finished, the whole process is finished; S407: If the group counseling has not yet ended, the planning leader intelligent agent summarizes the current situation of the group counseling and considers whether the next counseling plan needs to be adjusted or refined, and proceeds to step 402.

3. The group counseling automation platform according to claim 1, characterized in that: Each agent has a set personality, memory, and behavior.

4. The group counseling automation platform according to claim 1 or 3, characterized in that: The agent's personality includes basic attributes, interpersonal communication attributes, role-playing attributes, and other attributes; The basic attribute settings include: name, gender, age, educational background, occupation, personality, interests and specialties, interpersonal relationships, growth experience, motivation for participation, whether there is group experience, willingness to participate in a group, and personality type; interpersonal communication attributes include: self-cognition, interpersonal attitude, and interpersonal perception; role-playing attributes include: language characteristics, emotional expression, and interaction patterns; other attributes include extroversion index and problems faced.

5. The group counseling automation platform according to claim 1 or 3, characterized in that: The agent memory includes short-term memory and long-term memory. Short-term memory is the memory of the agent's speech and the perceived speech within a round. If the agent often encounters similar external discoveries, the short-term memory is updated to long-term memory. Long-term memory takes into account the long-term memory based on personality type. The long-term memory of this round is updated according to the long-term memory of the previous round and the short-term memory of this round to take into account the memory preferences of different agents.

6. The group counseling automation platform according to claim 1 or 3, characterized in that: The behavior of the intelligent agent includes the behavior of the intelligent agent of each role; the behavior of the group member intelligent agent includes thinking, choosing, speaking, and updating memory; the behavior of the main leader intelligent agent includes thinking, choosing, speaking, and updating memory; the behavior of the planning leader intelligent agent includes planning, speaking, and updating memory.

7. An application of the group counseling automation platform according to any one of claims 1 to 6, characterized in that: Applied to psychology research, education and training, and mental health services.

Images