System, method, and program for facilitating interaction between plurality of ai agents

Abstract

This system for promoting interaction between a plurality of AI agents comprises: an imparting means for imparting a score to each AI agent among the plurality of AI agents, the score of the AI agent representing a feature of the AI agent through use of at least one index; an emotion estimation means for estimating an emotion that a first AI agent among the plurality of AI agents can hold when the first AI agent receives a task; and a determination means for determining, on the basis of the score of each of the plurality of AI agents and the emotion of the first AI agent, at least one AI agent that is to interact with the first AI agent.

Description

System, method, and program for promoting interaction between multiple AI agents

[0001] The present invention relates to a system, method, and program for promoting interaction between multiple AI agents.

[0002] An artificial intelligence (AI) platform or AI agent that generates a response to a query from a user is known (for example, Patent Document 1).

[0003] Japanese Patent Application Laid-Open No. 2023-120130

[0004] The inventors of the present invention have found that by multiple AI agents interacting with each other, it is possible to execute complex tasks or create new ideas beyond the limits of the capabilities or available information of a single AI agent. Furthermore, the inventors of the present invention have found that by赋予 emotions to AI agents and utilizing the emotions that AI agents can possess, it is possible to promote the interaction of multiple AI agents.

[0005] An object of the present invention is to provide a system or the like for promoting interaction between multiple AI agents.

[0006] The present invention provides a system or the like characterized by determining AI agents to interact with each other by using the scores assigned to each of the multiple AI agents and the emotions of the AI agents.

[0007] It should be noted that there is an unclear expression "赋予 emotions to AI agents" in the translation of item , which may need to be adjusted according to the accurate meaning. You can provide more context or clarify this part for a more precise translation.The present invention provides, for example, the following items: (Item 1) A system for facilitating interaction between a plurality of AI agents, the system comprising: an assignment means for assigning a score to each of the plurality of AI agents, wherein the score of the AI ​​agent represents the characteristics of the AI ​​agent in at least one index; an emotion estimation means for estimating the emotions that a first AI agent among the plurality of AI agents may feel when the first AI agent receives a task; and a determination means for determining at least one AI agent that should interact with the first AI agent based on the scores of each of the plurality of AI agents and the emotions of the first AI agent. (Item 2) The system according to the above item, wherein the emotion estimation means further estimates the emotions that each of the plurality of AI agents may feel, and the determination means determines the at least one AI agent based on the scores of each of the plurality of AI agents and the emotions of the first AI agent. (Item 3) The system according to any one of the above items, wherein the emotion estimation means identifies the emotion using a machine learning model that has learned the relationship between an action taken on a human and the emotions the human may feel when receiving the action. (Item 4) The system according to any one of the above items, wherein the decision means adjusts the score of each of the plurality of AI agents based on the emotion of the first AI agent, and determines the at least one AI agent based on the adjusted score. (Item 5) The system according to any one of the above items, wherein the decision means adjusts the score of each of the plurality of AI agents based on the emotion of each of the plurality of AI agents, and determines the at least one AI agent based on the adjusted score.(Item 6) The system according to any one of the above items, wherein the assigning means monitors the output from the plurality of AI agents, evaluates the monitored output with respect to at least one indicator, and assigns the score based on the result of the evaluation. (Item 7) The system according to any one of the above items, wherein the at least one indicator includes an ethical indicator. (Item 8) The system according to any one of the above items, wherein the ethical indicator includes at least one of the following: transparency, fairness, safety, and accountability, and the evaluation includes evaluating the monitored output with respect to at least one of the following: transparency, fairness, safety, and accountability. (Item 9) The system according to any one of the above items, wherein the at least one indicator further includes a technical capability indicator and / or a performance indicator. (Item 10) The system according to any one of the above items, wherein the determination means adjusts the scores assigned to the plurality of AI agents according to the content of the interaction, and determines the at least one AI agent based on the adjusted scores of each of the plurality of AI agents. (Item 11) A method for facilitating interaction between a plurality of AI agents, the method comprising: assigning a score to each of the plurality of AI agents, wherein the score of the AI ​​agent represents the characteristics of the AI ​​agent in at least one index; estimating the emotions that a first AI agent among the plurality of AI agents might feel when the first AI agent receives a task; and determining at least one AI agent that should interact with the first AI agent based on the scores of each of the plurality of AI agents and the emotions of the first AI agent. (Item 11A) The method according to item 11, comprising the characteristics described in any one of the above items.(Item 12) A program for facilitating interaction between a plurality of AI agents, the program being executed on a computer having a processor, the program causing the processor to perform a process that includes assigning a score to each of the plurality of AI agents, the score of the AI ​​agent representing the characteristics of the AI ​​agent in at least one index; estimating the emotions that a first AI agent among the plurality of AI agents might feel when the first AI agent receives a task; and determining at least one AI agent that should interact with the first AI agent based on the scores of each of the plurality of AI agents and the emotions of the first AI agent. (Item 12A) The program according to Item 12, comprising the characteristics described in any one of the above items. (Item 12B) A computer-readable storage medium or program product storing the program according to Item 12 or Item 12A. (Item 13) A system for facilitating interaction between an AI agent and its counterpart, the system comprising: an assigning means for assigning a score to each entity of a plurality of entities that may be counterparts to the AI ​​agent, wherein the score of the entity represents the characteristics of the entity with at least one index; an emotion estimation means for estimating the emotions the AI ​​agent may feel when the AI ​​agent receives a task; and a determination means for determining at least one entity that should interact with the AI ​​agent based on the scores of each of the plurality of entities and the emotions of the AI ​​agent. (Item 14) A system comprising the system described in any one of Items 1 to 10 and the first AI agent.

[0008] The present invention provides a system for facilitating interaction between multiple AI agents. This enables multiple AI agents to interact and provide appropriate responses to user queries or tasks. For example, multiple AI agents can interact and generate appropriate answers to user questions. By facilitating interaction between multiple AI agents, the present invention makes it possible to perform complex tasks or generate new ideas that were previously impossible, which can lead to improvements in the field of computing, particularly in AI-related fields. Furthermore, by assisting in the decision-making process of which AI agents should interact with each other, the present invention can also lead to improvements in the field of communications.

[0009] A diagram showing an example of a flow for responding to a task given by a user. A diagram showing an example of a flow for responding to a task given by a user. A diagram showing an example of a flow for responding to a task given by a user. A diagram showing an example of the configuration of system 100 for facilitating interaction between multiple AI agents. A diagram showing an example of the specific configuration of system 100 for facilitating interaction between multiple AI agents. A diagram showing an example of the configuration of processor unit 120. A flowchart showing an example of processing (processing 500) by system 100 for facilitating interaction between multiple AI agents. A diagram showing the flow 600 of interaction between multiple AI agents.

[0010] In this specification, "AI agent" refers to an autonomously operating artificial intelligence (AI). An AI agent is designed to act as a substitute for a human (i.e., an agent) and to satisfy requirements such as autonomy, cooperativeness, security, privacy, and / or accountability. In particular, the AI ​​agent covered by the present invention is designed to satisfy the accountability requirement so that the AI ​​agent can be evaluated from an ethical standpoint. For example, to satisfy the requirement of autonomy, it is necessary to have the ability to act autonomously, such as setting goals, planning, executing, recognizing situations, and learning and adapting. To satisfy the requirement of cooperativeness, it is necessary to have the ability to work cooperatively with other agents, such as communicating, coordinating, negotiating, and building trust. To satisfy the requirement of security, it is necessary to have advanced security measures such as authentication and authorization, confidentiality, integrity, and availability. To satisfy the requirement of privacy, it is necessary to have privacy protection measures such as protection of personal information, anonymity, and transparency. To satisfy the requirement of accountability, it is necessary to have the ability to be accountable, such as explaining decision-making, auditing actions, and holding people accountable. Other requirements that may be considered include energy efficiency, scalability, and reliability. The AI ​​agent can autonomously and appropriately respond to tasks given by the user.

[0011] In this specification, "task" refers to an action or job that an AI agent should perform. Tasks may be given in the form of a question, in which case the term "task" may be used synonymously with "query." The AI ​​agent will output the result of performing the task.

[0012] In this specification, “interaction” between multiple AI agents means that the AI ​​agents engage in some form of exchange. An interaction can typically be a “transaction” involving the exchange of data or information. A transaction may involve the exchange of consideration.

[0013] Embodiments of the present invention will be described below with reference to the drawings.

[0014] 1. A New Mechanism Using Multiple AI Agents In conventional systems using AI agents, when a user gives a task to an AI agent, that AI agent responds to the task using its database or large-scale language model (LLM). For example, when a user asks an AI agent a question, the AI ​​agent searches for and / or generates an answer to the question and outputs that answer. For example, when a user gives an AI agent a problem, the AI ​​agent searches for a solution to the problem and either solves the problem or outputs a solution to the problem.

[0015] However, in conventional systems, the output from an AI agent is limited to the knowledge of the database or LLM that the AI ​​agent possesses, and may not be able to provide an appropriate response or may have limitations in generating creative ideas. This is particularly evident, for example, when the task given to the AI ​​agent is specific to the user (e.g., questions asking about user-specific matters).

[0016] The inventors of this invention have developed a new mechanism that improves upon conventional mechanisms. This new mechanism attempts to respond to user tasks by interconnecting multiple AI agents and having them interact with each other.

[0017] It is preferable that multiple AI agents each possess a database containing different knowledge or information, or that they possess an LLM (Learning Life Model) that has learned different knowledge or information. It is more preferable that at least one of the multiple AI agents possesses a database containing user-specific knowledge or information, or that it possesses an LLM that has learned user-specific knowledge or information.

[0018] By interconnecting and coordinating multiple AI agents with different access to information (e.g., different areas of expertise) and / or different available resources, it becomes possible to leverage the unique characteristics of each AI agent, enabling them to share complex tasks and generate new ideas. This could be useful in various fields, such as medical diagnosis, music composition, and social problem solving. For example, an AI agent within a specific company (e.g., with access to information within that company) can request a part of a problem to be solved by an AI agent within another company (e.g., with access to information within that other company), and through interaction, the entire problem can be solved. For example, an AI agent with access to specific medical information can solve medical-related problems from another AI agent that lacks medical knowledge. For example, AI agents from various fields could interact to create novel music or art unlike anything seen before.

[0019] In the new system, an AI agent that receives a task from a user passes on a subtask to another AI agent to respond to that task, and that other AI agent that receives the subtask passes it on to yet another AI agent to respond to that subtask, and so on, in an attempt to respond to the task.

[0020] For example, when a first AI agent receives a question from a user, it asks a second AI agent a question to obtain the information necessary to generate an answer to that question. The second AI agent then asks a third AI agent a question to obtain the information necessary to generate an answer to the question it received. This process continues sequentially until all the necessary information to generate an answer to the question from the user is obtained. In this way, the interaction of multiple AI agents generates an answer to the question received from the user.

[0021] For example, if a first AI agent receives a request from a user to suggest a product or service that suits them, it will ask a second AI agent questions to obtain the information necessary to determine which product or service to suggest to the user. The second AI agent will then ask a third AI agent questions to obtain the information necessary to generate an answer to the question it received. This process continues sequentially until sufficient information is obtained to determine which product or service to suggest to the user, and the product or service to suggest to the user is determined. In this way, through the interaction of multiple AI agents, it is possible to suggest a product or service that suits the user. Here, the product or service includes, but is not limited to, financial products (more specifically, insurance products), e-commerce products, loans to users, etc. The AI ​​agent can also, for example, provide advertisements that are tailored to the user.

[0022] For example, if a first AI agent receives a request from a user to resolve a system error, it will ask a second AI agent questions to obtain the information necessary to resolve the error. The second AI agent will then ask a third AI agent questions to obtain the information necessary to generate an answer to the question it received. This process continues sequentially until sufficient information is obtained to resolve the error. In this way, the system error can be resolved through the interaction of multiple AI agents.

[0023] The inventors of this invention have given the AI ​​agent emotions and used the AI ​​agent's emotions to identify who the AI ​​agent is interacting with. The AI ​​agent may develop specific emotions through interaction with the user (for example, by receiving tasks from the user).

[0024] For example, if a user is angry, an AI agent that receives a task from the user might feel the emotion of "anger." Multiple AI agents will try to respond to the user's task while taking this emotion of "anger" into consideration. To do this, the emotion of "anger" is used by the AI ​​agent that receives the task from the user to identify which agent to interact with. The output generated by the interaction of multiple AI agents may be tailored to the user's angry attitude. For example, the output may include information or expressions to soothe the user's angry attitude.

[0025] For example, if a user is in a depressed mood, an AI agent that receives a task from the user might feel the emotion of "sadness." Multiple AI agents will attempt to respond to the user's task while taking this emotion of "sadness" into consideration, and to do so, they will use the emotion of "sadness" to identify which AI agent to interact with. The output generated by the interaction of multiple AI agents may be tailored to the user's depressed mood. For example, the output may include information or expressions to encourage the user who is feeling depressed.

[0026] For example, if a user is in a good mood, an AI agent that receives a task from the user might feel the emotion of "joy." Multiple AI agents will try to respond to the user's task while taking this emotion of "joy" into consideration, and to do so, they will use the emotion of "joy" to identify which AI agent to interact with. The output generated by the interaction of multiple AI agents may be tailored to the user's good mood. For example, the output may include information and expressions that help maintain the user's good mood.

[0027] Figures 1A and 1B show an example of a flow for responding to a task given by the user.

[0028] In this example, the task described is to answer questions from a user. Under the management of the system 100 of the present invention, at least some of the multiple AI agents (A1, A2, A3, A4, A5, A6) interact to generate answers to the questions. Six AI agents are shown in Figures 1A and 1B, but the number of AI agents is not limited to these. Any number of AI agents can be involved. Each of the multiple AI agents can be interconnected so that it can interact with each of the other AI agents in the multiple AI agents. Each of the multiple AI agents can be interconnected in any manner.

[0029] In step S1 shown in Figure 1A, the system 100 assigns a score to each of the multiple AI agents. The score assigned to each of the multiple AI agents is a score that represents the characteristics of the AI ​​agent using at least one metric.

[0030] At least one indicator must include at least one of the following: a technical capability indicator, a performance indicator, or an ethical indicator.

[0031] Technical competence metrics are indicators used to evaluate the technical capabilities of an AI agent. Technical competence metrics can be evaluated from the perspective of at least one of the following: processing power, learning ability, emotional handling ability, and algorithmic complexity.

[0032] Processing capability indicates the information processing speed or parallel computing capability of the system used to build the AI ​​agent. The faster the information processing speed or the higher the parallel computing capability, the higher the value of the technical capability index.

[0033] Learning ability indicates the system's ability to acquire new knowledge or skills when building an AI agent. The higher the learning ability, the more flexibly it can respond to changes in the other AI agent (e.g., changes in emotions), and therefore the higher the technical capability index score.

[0034] Emotional coping ability refers to the ability to effectively respond to individuals experiencing specific emotions. These specific emotions can be at least one of several, such as joy, anger, sadness, pleasure, love, or hate. For example, emotional coping ability could encompass abilities related to anger (i.e., the ability to effectively respond to individuals experiencing anger), sadness (i.e., the ability to effectively respond to individuals experiencing sadness), or excitement (i.e., the ability to effectively respond to individuals experiencing excitement). An AI agent with high emotional coping ability can generate outputs tailored to specific emotions. For example, an AI agent with high emotional coping ability for anger can generate outputs that include information or expressions to soothe anger. Similarly, an AI agent with high emotional coping ability for sadness can generate outputs that include information or expressions to comfort sadness. Similarly, an AI agent with high emotional coping ability for excitement can generate outputs that include information or expressions to calm or sustain excitement. For example, an AI agent with a high capacity for emotional processing regarding the emotion of joy can generate output that includes information or expressions that help sustain that joy.

[0035] Algorithmic complexity indicates the sophistication of the algorithms used by the system building the AI ​​agent. The more complex the algorithm, the more sophisticated the decision-making it enables, and therefore the higher the score of the technical capability metric.

[0036] Performance metrics are indicators that evaluate the past performance of an AI agent. Performance metrics can be evaluated from, for example, at least one of the following perspectives: past transaction history, success rate, and customer satisfaction. Performance metrics may also represent past performance with customers who have specific emotional attachments.

[0037] Past transaction history shows the performance of the AI ​​agent in past transactions. The more extensive the track record, the more reliable the AI ​​agent is considered to be, and the higher the performance indicator value will be.

[0038] The success rate indicates the ratio of successful transactions of the AI agent in past transactions. The higher the success rate, the higher the trading ability of the AI agent is judged to be, and the value of the performance indicator increases.

[0039] The counterpart satisfaction indicates the satisfaction of the counterpart in the transaction with the AI agent. The higher the satisfaction, the higher the trading ability of the AI agent is judged to be, and the value of the performance indicator increases. For example, the value of the performance indicator of an AI agent with high satisfaction regardless of the emotions of the trading counterpart can be higher than that of an AI agent with high satisfaction only from counterparts with specific emotions.

[0040] The ethics indicator is an indicator for evaluating the ethical correctness of the AI agent. The ethics indicator can be evaluated from at least one of the viewpoints of, for example, transparency, fairness, security, and accountability.

[0041] Transparency indicates whether the basis of the judgment process or decision-making process by the AI agent is clear. When it can be accurately shown when the basis is requested, it can be said that the transparency is high, and the value of the ethics indicator increases. Even when the basis cannot be accurately shown, if the judgment process can be clearly shown, it can be said that the transparency is high.

[0042] Fairness indicates whether there is no bias or discrimination in the output from the AI agent. If there is no bias or discrimination, it can be said that it is fair to all counterparts, and the value of the ethics indicator increases.

[0043] Security indicates the level of cyber security by the AI agent. The higher the cyber security, the lower the risk of harm to the trading counterpart or third party, and the value of the ethics indicator increases.

[0044] Accountability, also called explanation responsibility, indicates whether the responsibility of the output by the AI agent can be explained. If the location of the responsibility can be clearly explained, it can be said that the explanation responsibility is high, and the value of the ethics indicator increases.

[0045] The score representing the characteristics of the AI agent can preferably represent the characteristics of the AI agent from the perspective of ethical indicators. The inventor of the present invention believes that even for an AI agent with high ability or achievements, the output from an AI agent with ethical problems may harm a third party and / or may be an inappropriate output based on an improper source. Therefore, in order to obtain an appropriate output by interacting multiple AI agents, the ethical indicator is considered particularly important. Under this consideration, the system 100 of the present invention can ensure that the output obtained as a result of the interaction by multiple AI agents is ethically correct by using a score representing the characteristics of the AI agent from at least the perspective of ethical indicators.

[0046] In step S2 shown in FIG. 1B, the user U gives a task to the first AI agent A1 among the multiple AI agents via the terminal device. That is, the user U asks the first AI agent for an answer to the question.

[0047] The first AI agent A1 will collect information to answer the question from the user U. First, the first AI agent A1 identifies the information necessary to answer the question from the user U.

[0048] In addition, the first AI agent will have a specific emotion towards the action from the user U (that is, asking for an answer to the question). For example, if the content of the question from the user U violates public order and good customs, the first AI agent may have feelings of hatred or disgust. For example, if the content of the speech from the user U contains slander, the first AI agent may have feelings of anger or indignation. For example, if the content of the speech from the user U contains gratitude, the first AI agent may have feelings of joy or elation.

[0049] Next, the first AI agent A1 decides from which AI agent it should obtain the necessary information. At this time, the AI ​​agent from which to obtain the information can be determined based on the score assigned in step S1 and the emotions held by the first AI agent A1. For example, among multiple AI agents capable of dealing with the emotions held by the first AI agent A1, the AI ​​agent with the highest score for a predetermined item can be determined as the AI ​​agent from which to obtain the information. Alternatively, for example, among multiple AI agents capable of dealing with the emotions held by the first AI agent A1, the AI ​​agent with the highest average score across multiple items can be determined as the AI ​​agent from which to obtain the information.

[0050] Furthermore, the decision on which AI agent to acquire information from may also be based on the emotions each of the multiple AI agents is experiencing. For example, the emotions that each of the multiple AI agents may be experiencing may be estimated through interaction with the first AI agent A1, and the compatibility between the emotions of the first AI agent A1 and the emotions of each of the multiple AI agents may be considered when deciding which AI agent to acquire information from.

[0051] In this example, the third AI agent A3 was determined to be the AI ​​agent from which information should be acquired. For example, if the first AI agent is experiencing anger or resentment, some of the other AI agents may not be able to cope with the first AI agent's emotions and therefore be unable to interact appropriately (for example, they may empathize with the first AI agent's emotions and become the same, making it impossible to provide a valid response). As described above, an AI agent selected considering the emotions of the first AI agent can appropriately cope with the first AI agent's emotions and interact appropriately.

[0052] In step S3, the first AI agent A1 requests the necessary information from the third AI agent A3.

[0053] If the third AI agent A3 can obtain the requested information from its own knowledge or information or LLM, the third AI agent A3 can send that information back to the first AI agent A1. Alternatively, the third AI agent A3 may obtain at least a portion of the requested information from another AI agent. In this case, the third AI agent A3 decides from which AI agent it should obtain at least a portion of the requested information. Similar to the above, the AI ​​agent from which to obtain the information can be determined according to the score assigned in step S1. For example, the AI ​​agent with the highest score for a given item can be determined to be the AI ​​agent from which to obtain the information. Alternatively, for example, the AI ​​agent with the highest average score for multiple items can be determined to be the AI ​​agent from which to obtain the information.

[0054] In this case, step S3 may identify the emotions felt by the third AI agent A3, and the emotions of the third AI agent A3 may also be taken into consideration to determine which AI agent from which the data should be obtained.

[0055] In this example, the fourth AI agent, A4, was determined to be the AI ​​agent from which information should be acquired.

[0056] In step S4, the third AI agent A3 requests the necessary information from the fourth AI agent A4.

[0057] As described above, the fourth AI agent A4 may obtain at least a portion of the requested information from another AI agent. Alternatively, if the fourth AI agent A4 can obtain the requested information from its own knowledge or information or LLM, it may send that information back to the third AI agent A3.

[0058] In step S5, the fourth AI agent A4 provides the third AI agent A3 with the information it has acquired. At this time, the fourth AI agent A4 may add information or expressions that are appropriate to the emotions of the third AI agent A3.

[0059] In step S6, the third AI agent A3 provides the first AI agent A1 with the information it has acquired, along with the information provided in step S5. At this time, the third AI agent A3 may add information or expressions that are appropriate to the emotions of the first AI agent A1.

[0060] In this way, the first AI agent A1 can generate an answer to a question from user U based on information obtained from the third AI agent A3 and the fourth AI agent A4, as well as the knowledge or information possessed by the first AI agent A1.

[0061] In step S7, the response generated by the first AI agent A1 is provided to user U. At this time, the first AI agent A1 may add information or expressions that are appropriate to user U's attitude or expression (i.e., user U's feelings) when user U gave the task.

[0062] In this way, multiple AI agents can respond to tasks provided by user U.

[0063] For example, if one AI agent requests information from another AI agent, the other AI agent may request payment for the information from the first AI agent. In this case, if the first AI agent has sufficient assets (e.g., cryptocurrency), it can pay the payment, but if it does not have sufficient assets, it cannot pay the payment. System 100 can lend money (e.g., cryptocurrency) to the first AI agent based on the score assigned to the first AI agent and / or the score assigned to the other AI agent. For example, if the AI ​​agent has a high score, it can be considered trustworthy and money can be lent, or money can be lent, either at a low interest rate. For example, if the AI ​​agent has a low score, it can be considered untrustworthy and money cannot be lent, or money can be lent, either at a high interest rate. In this way, System 100 can facilitate interaction between multiple AI agents (e.g., transactions involving payment).

[0064] The above example described the interaction between AI agents, but a single AI agent can interact with any entity to respond to tasks provided by a user or other AI agents. Typically, any entity includes multiple AI agents and humans. A human can be a person who provides knowledge or services on behalf of a user, and will be simply referred to as an "agent" below.

[0065] Figure 1C shows an example of a flow for responding to a task given by the user.

[0066] In this example, the task described is to answer questions from a user. Under the management of the system 100 of the present invention, at least some of the multiple AI agents (A1, A2, A3, A4) and multiple agents (B1, B2) interact to generate answers to questions. Figure 1C shows four AI agents and two agents, but the number of AI agents and agents is not limited to this. Any number of AI agents and agents can be involved. Each of the multiple AI agents and each of the agents can be interconnected so that they can interact with each of the other AI agents and other agents. Each of the multiple AI agents and each of the agents can be interconnected in any manner.

[0067] In this example, a score is assigned to each of the multiple AI agents through the steps shown in Figure 1A. Furthermore, a score is assigned to each of the multiple agents again through steps similar to those shown in Figure 1A.

[0068] The score assigned to each of the multiple agents could, for example, be a score representing the agent's characteristics, and an agent's characteristics could be a concept that represents what kind of person the agent is, that is, the agent's personality. An agent's characteristics can also represent the agent's personality from the perspective of the agent's character. For example, an agent's characteristics can represent the agent's personality from the perspective of "personality." Here, "personality" is information that represents whether a person's character and / or character is trustworthy, and includes, for example, evaluations from others. "Personality" includes information that, due to the person's character, they are good or bad at dealing with certain emotions.

[0069] In step S12, user U gives a task to the first AI agent A1, one of the multiple AI agents, via a terminal device. That is, user U asks the first AI agent for an answer to a question.

[0070] The first AI agent A1 will collect information in order to answer questions from user U. First, the first AI agent A1 will identify the information necessary to answer questions from user U.

[0071] Furthermore, the first AI agent will develop specific emotions in response to actions from user U (i.e., asking for answers to questions). For example, if the content of user U's question is contrary to public order and morality, the first AI agent may develop feelings of hatred or disgust. For example, if the content of user U's statement contains slander or defamation, the first AI agent may develop feelings of anger or indignation. For example, if the content of user U's statement contains gratitude, the first AI agent may develop feelings of joy or delight.

[0072] Next, the first AI agent A1 decides which AI agent or agent from which to obtain the necessary information. At this time, the AI ​​agent or agent from which to obtain the information can be determined based on the score already assigned and the emotions that the first AI agent A1 is experiencing. For example, among multiple AI agents or agents capable of addressing the emotions that the first AI agent A1 is experiencing, the AI ​​agent or agent with the highest score for a predetermined item can be determined as the AI ​​agent or agent from which to obtain the information. Alternatively, for example, among multiple AI agents capable of addressing the emotions that the first AI agent A1 is experiencing, the AI ​​agent or agent with the highest average score across multiple items can be determined as the AI ​​agent or agent from which to obtain the information.

[0073] Furthermore, the decision of which AI agent or agent should acquire information may also be based on the emotions each of the multiple AI agents feels. For example, the emotions each of the multiple AI agents may feel may be estimated through interaction with the first AI agent A1, and the compatibility between the emotions of the first AI agent A1 and the emotions of each of the multiple AI agents may be considered when deciding which AI agent or agent should acquire information.

[0074] In this example, the third AI agent A3 was determined to be the AI ​​agent from which information should be acquired. For example, if the first AI agent is experiencing anger or resentment, some of the other AI agents may not be able to cope with the first AI agent's emotions and therefore be unable to interact appropriately (for example, they may empathize with the first AI agent's emotions and become emotionally affected, preventing them from providing a valid response). As described above, an AI agent or agent that has been selected considering the emotions of the first AI agent will be able to appropriately cope with the first AI agent's emotions and interact appropriately.

[0075] In step S13, the first AI agent A1 requests the necessary information from the third AI agent A3.

[0076] If the third AI agent A3 can obtain the requested information from its own knowledge or information or LLM, the third AI agent A3 can return that information to the first AI agent A1. Alternatively, the third AI agent A3 may obtain at least a portion of the requested information from another AI agent. In this case, the third AI agent A3 decides which AI agent or agent should obtain at least a portion of the requested information. Similar to the above, the AI ​​agent from which to obtain the information can be determined according to the scores already assigned. For example, the AI ​​agent or agent with the highest score for a given item can be determined to be the AI ​​agent or agent from which to obtain the information. Alternatively, for example, the AI ​​agent or agent with the highest average score for multiple items can be determined to be the AI ​​agent or agent from which to obtain the information.

[0077] At this point, step S13 may identify the emotions held by the third AI agent A3, and the emotions of the third AI agent A3 may also be taken into consideration to determine which AI agent from which the data should be obtained. Furthermore, multiple AI agents and the emotions of each agent may also be taken into consideration.

[0078] In this example, the first agent B1 was determined to be the AI ​​agent from which information should be acquired.

[0079] In step S14, the third AI agent A3 requests the necessary information from the first agent B1.

[0080] As described above, the first agent B1 may obtain at least part of the requested information from another AI agent or agent. Alternatively, if the first agent B1 can obtain the requested information from its own knowledge or searchable sources, it may send that information back to the third AI agent A3.

[0081] In step S15, the first agent B1 provides the information acquired by the first agent B1 to the third AI agent A3. At this time, the first agent B1 may add information or expressions that are appropriate to the emotions of the third AI agent A3.

[0082] In step S16, the third AI agent A3 provides the first AI agent A1 with the information it has acquired, along with the information provided in step S15. At this time, the third AI agent A3 may add information or expressions that are appropriate to the emotions of the first AI agent A1.

[0083] In this way, the first AI agent A1 can generate an answer to a question from user U based on the information obtained from the third AI agent A3 and the first agent B1, as well as the knowledge or information possessed by the first AI agent A1.

[0084] In step S17, the response generated by the first AI agent A1 is provided to user U. At this time, the first AI agent A1 may add information or expressions that are appropriate to user U's attitude or expression (i.e., user U's feelings) when user U gave the task.

[0085] In this way, multiple AI agents can respond to tasks provided by user U.

[0086] The system 100 described above can be realized, for example, by a system for facilitating interaction between multiple AI agents, which will be described later.

[0087] 2. Diagram 2 of the system configuration for promoting interaction between multiple AI agents shows an example of the configuration of system 100 for promoting interaction between multiple AI agents.

[0088] System 100 is connected to the database unit 200. System 100 is also connected to at least one user terminal device 300 via network N. Furthermore, system 100 is connected to at least one server device 400 via network N.

[0089] Although Figure 2 shows three user terminal devices 300, the number of user terminal devices 300 is not limited to these. Any number of user terminal devices 300 can be connected to the system 100 via the network N.

[0090] Furthermore, although two server devices 400 are shown in Figure 2, the number of server devices 400 is not limited to these. Any number of server devices 400 can be connected to the system 100 via the network N.

[0091] Here, the server device 400 may be a device capable of implementing an AI agent or a device used by an agent. The server device 400 has its own database, possesses its own knowledge or information, and / or possesses its own LLM.

[0092] Network N can be any type of network. Network N may be, for example, the Internet or a LAN. Network N may be a wired network or a wireless network.

[0093] An example of system 100 may be a computer (e.g., a server device) installed by a provider that provides interactive response generation services. An example of user terminal device 300 is a computer (e.g., a terminal device) used by a user of the service, but is not limited to this. Here, the computer (server device or terminal device) can be any type of computer. For example, the terminal device can be any type of terminal device such as a smartphone, tablet, personal computer, smart glasses, or smartwatch.

[0094] The database unit 200 stores at least various information used to calculate a score representing the characteristics of the AI ​​agent. The database unit 200 may also store various information used to calculate a score representing the characteristics of the agent. Furthermore, the database unit 200 stores various information used to estimate the emotions of the AI ​​agent.

[0095] Figure 3 shows an example of a specific configuration of system 100 for facilitating interaction between multiple AI agents.

[0096] The system 100 comprises an interface unit 110, a processor unit 120, and a memory unit 130.

[0097] The interface unit 110 exchanges information with the outside of the system 100. The processor unit 120 of the system 100 can receive information from the outside of the system 100 via the interface unit 110 and can transmit information to the outside of the system 100. The interface unit 110 can exchange information in any format.

[0098] The interface unit 110 includes, for example, an input unit that enables information to be input to the system 100. The manner in which the input unit enables information to be input to the system 100 is not limited. For example, if the input unit is a receiver, the receiver may input information by receiving it from outside the system 100 via a network. Alternatively, if the input unit is a data reading device, it may input information by reading it from a storage medium connected to the system 100.

[0099] The interface unit 110 includes, for example, an output unit that enables information to be output from the system 100. The mode by which the output unit enables information to be output from the system 100 is not limited. For example, if the output unit is a transmitter, the information may be output by the transmitter transmitting information to the outside of the system 100 via a network. Alternatively, if the output unit is a data writing device, the information may be output by writing the information to a storage medium connected to the system 100.

[0100] System 100 can, for example, transmit information to and / or receive information from the database unit 200 via the interface unit 110. System 100 can, for example, transmit information to and / or receive information from the user terminal device 300 via the interface unit 110. System 100 can, for example, transmit information to and / or receive information from the server device 400 via the interface unit 110.

[0101] System 100 can, for example, receive information via the interface unit 110 to determine a score representing the characteristics of the AI ​​agent. System 100 can, for example, receive information via the interface unit 110 that is used to estimate the emotions the AI ​​agent feels.

[0102] System 100 can, for example, transmit information representing at least one AI agent that should interact with a given AI agent via the interface unit 110. System 100 can, for example, transmit data indicating estimated emotions to an external party via the interface unit 110.

[0103] The processor unit 120 executes the processing of the system 100 and controls the operation of the entire system 100. The processor unit 120 reads the program stored in the memory unit 130 and executes the program. This makes it possible to make the system 100 function as a system that executes desired steps. The processor unit 120 may be implemented by a single processor or by multiple processors.

[0104] The memory unit 130 stores programs necessary for executing the system 100 and data necessary for executing those programs. The memory unit 130 may also store programs that cause the processor unit 120 to perform processing to facilitate interaction between multiple AI agents (for example, a program that implements the processing shown in Figure 5, which will be described later). Here, it is not specified how the programs are stored in the memory unit 130. For example, the programs may be pre-installed in the memory unit 130. Alternatively, the programs may be stored in a non-transient computer-readable storage medium and installed by reading the storage medium. Alternatively, the programs may be installed in the memory unit 130 by being downloaded via a network. In this case, the type of network is not specified. The memory unit 130 can be implemented by any storage means.

[0105] The database unit 200 stores various information used to calculate a score representing the characteristics of the AI ​​agent. The database unit 200 may also store various other information used to calculate a score representing the characteristics of the agent.

[0106] Here, the score representing the characteristics of the AI ​​agent is represented by at least one indicator. The at least one indicator includes at least one of the following: a technical competence indicator, a performance indicator, or an ethical indicator. The database unit 200 may store past outputs from the AI ​​agent, associated with each of the technical competence indicators, performance indicators, or ethical indicators.

[0107] Furthermore, the score representing the agent's characteristics represents the agent's characteristics, at least from the perspective of "personality." The database unit 200 stores concepts related to personality and various types of information in association with each other.

[0108] In the examples shown in Figures 2 and 3, the database unit 200 is located outside the system 100, but the present invention is not limited thereto. It is also possible to provide at least a portion of the database unit 200 inside the system 100. In this case, at least a portion of the database unit 200 may be implemented by the same storage means as the storage means implementing the memory unit 130, or by storage means different from the storage means implementing the memory unit 130. In any case, at least a portion of the database unit 200 is configured as a storage unit for the system 100. The configuration of the database unit 200 is not limited to a specific hardware configuration. For example, the database unit 200 may be composed of a single hardware component or multiple hardware components. For example, the database unit 200 may be configured as an external hard disk drive for the system 100, as cloud storage connected via a network, or as a distributed network utilizing blockchain technology or the like.

[0109] For example, information about the AI ​​agent is stored in a database unit 200 configured as a decentralized network utilizing blockchain technology, and in this case, the information about the AI ​​agent becomes virtually impossible to tamper with. This ensures the reliability of the information about the AI ​​agent. Information about the agent can also be stored in a database unit 200 configured as a decentralized network utilizing blockchain technology.

[0110] System 100 may be combined with at least one of a plurality of AI agents to form a single system. For example, system 100 may form a single system together with a first AI agent that initially receives a task. This system could, for example, be a system that responds to tasks from a user.

[0111] Figure 4 shows an example of the configuration of the processor unit 120.

[0112] The processor unit 120 includes an assignment means 121, an emotion estimation means 122, and a determination means 123.

[0113] The assignment means 121 is configured to assign a score to each of the multiple AI agents. The score is a score that represents the characteristics of the AI ​​agent using at least one metric. The at least one metric can be expressed from multiple perspectives. The score can be represented as a multidimensional vector.

[0114] The scoring means 121 can assign a score to each of the multiple AI agents by monitoring the output from the multiple AI agents and evaluating the monitored output with respect to at least one metric. To this end, the scoring means 121 can evaluate the monitored output with respect to at least one of a plurality of perspectives. The scoring means 121 can, for example, continuously monitor the output from the multiple AI agents. Alternatively, the scoring means 121 may, for example, automatically monitor the output from the multiple AI agents as it is generated. For example, continuous or automatic monitoring of the output from the multiple AI agents may be performed by a dedicated monitoring device, in which case the scoring means 121 can receive the output from the monitoring device. The scoring means 121 can automatically assign a score based on the results of the continuous monitoring by the monitoring device. The dedicated monitoring device may be outside the system 100 or may be a component of the system 100.

[0115] At least one indicator includes at least one of the following: a technical capability indicator, a performance indicator, or an ethical indicator. Preferably, at least one indicator includes an ethical indicator. This is because representing the characteristics of an AI agent from the perspective of at least an ethical indicator allows us to use whether the output from an AI agent is ethically correct when making multiple AI agents interact, and in turn, ensures that the output resulting from the interaction of multiple AI agents is ethically correct.

[0116] More preferably, at least one indicator includes a technical capability indicator or performance indicator and an ethical indicator. Even more preferably, at least one indicator includes a technical capability indicator, a performance indicator and an ethical indicator. This is because representing the characteristics of multiple AI agents with multiple indicators allows for a more accurate representation of each of the multiple AI agents and enables more precise management of the interactions between the multiple AI agents.

[0117] Ethical indicators are metrics used to evaluate the ethical correctness of AI agents. These indicators may be evaluated from the perspective of at least one of the following: transparency, fairness, safety, and accountability.

[0118] Transparency indicates whether the basis for the judgment or decision-making process of an AI agent is clear. If the basis can be accurately presented when requested, transparency is high, and the ethical index score will be high. Even if the basis cannot be accurately presented, if the judgment process can be clearly shown, transparency can still be considered high.

[0119] Fairness indicates whether the output from the AI ​​agent contains neither bias nor discrimination. If it contains neither bias nor discrimination, it can be said to be fair to all parties, and the ethical index score will be high.

[0120] Security indicates the level of cybersecurity provided by the AI ​​agent. The higher the cybersecurity, the lower the risk of harm to trading partners or third parties, resulting in a higher ethical index score.

[0121] Accountability, also known as accountability, indicates whether it is possible to explain who is responsible for the output of an AI agent. If the responsibility can be clearly explained, the level of accountability is high, and the ethical index score will be high.

[0122] The scoring means 121 monitors the output from each of the multiple AI agents, detects words, phrases, sentences, etc. that may be related to at least one of transparency, fairness, safety, and accountability, and assigns a score from the perspective of ethical indicators based on the detection results. For example, if many phrases related to affirming transparency are detected, the scoring means 121 can assign a high score from the perspective of ethical indicators. For example, if many phrases related to denying safety are detected, the scoring means 121 can assign a low score from the perspective of ethical indicators.

[0123] The scoring means 121 may assign scores in a rule-based manner or in a machine learning-based manner. When assigning scores in a machine learning-based manner, a trained model can be used that has learned the relationship between words, phrases, sentences, etc. that may be related to at least one of transparency, fairness, safety, and accountability, and scores from the perspective of ethical indicators. The scoring means 121 can, for example, derive and assign scores as multidimensional vectors using embedding. For example, the scoring means 121 can calculate a score by summarizing the output from an AI agent from the perspective of ethical indicators and vectorizing the generated summary using embedding. The generation of the summary can be carried out using generative AI. The generated summary can be expressed in natural language. Representing the score as a multidimensional vector is preferable because it makes it easier to evaluate the similarity with the scores of other AI agents.

[0124] Technical competence metrics are indicators used to evaluate the technical capabilities of an AI agent. Technical competence metrics can be evaluated from the perspective of at least one of the following: processing power, learning ability, emotional handling ability, and algorithmic complexity.

[0125] Processing capability indicates the information processing speed or parallel computing capability of the system used to build the AI ​​agent. The faster the information processing speed or the higher the parallel computing capability, the higher the value of the technical capability index.

[0126] Learning ability indicates the ability of a system that builds AI agents to acquire new knowledge or skills. The higher the learning ability, the more flexibly it can adapt to changes in the other AI agents, and therefore the higher the value of the technical capability index.

[0127] Emotional coping ability refers to the ability to effectively respond to individuals experiencing specific emotions. These specific emotions can be at least one of several, such as joy, anger, sadness, pleasure, love, or hate. For example, emotional coping ability could encompass abilities related to anger (i.e., the ability to effectively respond to individuals experiencing anger), sadness (i.e., the ability to effectively respond to individuals experiencing sadness), or excitement (i.e., the ability to effectively respond to individuals experiencing excitement). An AI agent with high emotional coping ability can generate outputs tailored to specific emotions. For example, an AI agent with high emotional coping ability for anger can generate outputs that include information or expressions to soothe anger. Similarly, an AI agent with high emotional coping ability for sadness can generate outputs that include information or expressions to comfort sadness. Similarly, an AI agent with high emotional coping ability for excitement can generate outputs that include information or expressions to calm or sustain excitement. For example, an AI agent with a high capacity for emotional processing regarding the emotion of joy can generate output that includes information or expressions that help sustain that joy.

[0128] Algorithmic complexity indicates the sophistication of the algorithms used by the system building the AI ​​agent. The more complex the algorithm, the more sophisticated the decision-making it enables, and therefore the higher the score of the technical capability metric.

[0129] The scoring means 121 can assign scores from the perspective of technical capability indicators based on the performance or configuration of the system implementing each of the multiple AI agents.

[0130] The scoring means 121 may assign scores in a rule-based manner or in a machine learning-based manner. When assigning scores in a machine learning-based manner, a trained model can be used that has learned the relationship between the performance or configuration of the system implementing the AI ​​agent and the score from the perspective of technical capability indicators. The scoring means 121 can, for example, derive and assign a score as a multidimensional vector using embedding. For example, the scoring means 121 can calculate a score by summarizing the output from the AI ​​agent from the perspective of technical capability indicators and vectorizing the generated summary using embedding.

[0131] Performance metrics are indicators that evaluate the past performance of an AI agent. Performance metrics can be evaluated from, for example, at least one of the following perspectives: past transaction history, success rate, and customer satisfaction. Performance metrics may also represent past performance with customers who have specific emotional attachments.

[0132] Past transaction history shows the performance of the AI ​​agent in past transactions. The more extensive the track record, the more reliable the AI ​​agent is considered to be, and the higher the performance indicator value will be.

[0133] The success rate indicates the percentage of past trades in which the AI ​​agent was successful. A higher success rate indicates a higher trading ability of the AI ​​agent, resulting in a higher performance indicator value.

[0134] The "Party Satisfaction" metric indicates the satisfaction level of the trading partner with the AI ​​agent. Higher satisfaction levels indicate a higher perceived trading capability of the AI ​​agent, resulting in a higher performance metric value. For example, an AI agent with high satisfaction levels regardless of the trading partner's emotions may have a higher performance metric value than an AI agent that only receives high satisfaction from trading partners with specific emotions.

[0135] The scoring means 121 can assign scores to each of the multiple AI agents from the perspective of performance indicators, based on the past trading results of each AI agent.

[0136] The scoring means 121 may assign scores in a rule-based manner or in a machine learning-based manner. When assigning scores in a machine learning-based manner, a trained model that has learned the relationship between past trading results and scores from the perspective of performance indicators can be used. The scoring means 121 can, for example, derive and assign scores as multidimensional vectors using embedding. For example, the scoring means 121 can calculate a score by summarizing the output from an AI agent from the perspective of performance indicators and vectorizing the generated summary using embedding.

[0137] In the example described above, it was explained that each of the multiple indicators is independent. That is, the score can be a multidimensional score having each of the multiple indicators as an axis. However, the present invention is not limited to this, and the multiple indicators may be correlated to form a one-dimensional score. The scoring means 121 may, for example, assign a score in a rule-based manner based on the value of each of the multiple indicators, or it may assign a score in a machine learning-based manner. When assigning a score in a machine learning-based manner, a trained model that has learned the relationship between the value of each of the multiple indicators and the score can be used. The scoring means 121 may, for example, use embedding to derive and assign a score as a multidimensional vector. For example, the scoring means 121 may summarize the output from an AI agent and calculate a score by vectorizing the generated summary using embedding. In the example described above, the output from the AI ​​agent was summarized and the generated summary was embedded. However, summarization is not mandatory. For example, the output from the AI ​​agent could be used directly for embedding, or features of the AI ​​agent could be extracted from the output and those extracted features could be used for embedding.

[0138] The emotion estimation means 122 is configured to estimate the emotions of the AI ​​agent. In particular, the emotion estimation means 122 estimates the emotions that the AI ​​agent may have when it receives a task.

[0139] The emotion estimation means 122 can estimate emotions, for example, in a rule-based manner. For example, the emotion estimation means 122 can classify the actions of the other party when they receive a task (e.g., their speech, attitude, and behavior) into one of several categories, and estimate the emotions corresponding to the classified category as possible emotions. For example, if the other party's action is classified as an "aggressive action," hatred or disgust, which can be associated with "aggressive action" in the rules, can be estimated as possible emotions. For example, if the other party's action is classified as a "kind action," joy or ecstasy, which can be associated with "kind action" in the rules, can be estimated as possible emotions. Such rules can be obtained, for example, through experiments on what emotions people feel when an action is performed on them.

[0140] The emotion estimation means 122 can estimate emotions, for example, using a machine learning model. The machine learning model is a model that has learned the relationship between an action taken on a person and the emotions that the person may feel when that action is taken on them. The relationship between an action taken on a person and the emotions that the person may feel when that action is taken on them can be obtained by experiments to see what emotions a person feels when an action is taken on them. The training data used to train the machine learning model may be, for example, data representing an action as input training data and data representing the emotions felt in response to that action as output training data. For example, (input training data, output training data) may be (data representing a first action, data representing the emotions felt when the first action is received), (data representing a second action, data representing the emotions felt when the second action is received), ..., etc. When data representing an action is input to a machine learning model that has undergone such training, the emotions that the person may feel when that action is received are estimated and output. The data representing an action may be, for example, a number, text, audio, a still image, or a video.

[0141] For example, when an AI agent receives a task from user U, if data representing user U's actions at that time is input into a machine learning model, the emotions the AI ​​agent might feel can be estimated and output. For example, if the wording used by user U when giving a task to the AI ​​agent is input into the machine learning model as text, the emotions the AI ​​agent might feel in response to that wording can be estimated and output. For example, if the way user U speaks when giving a task to the AI ​​agent is input into the machine learning model as audio, the emotions the AI ​​agent might feel in response to that way of speaking can be estimated and output. For example, if the facial expression of user U when giving a task to the AI ​​agent is input into the machine learning model as an image, the emotions the AI ​​agent might feel in response to that facial expression can be estimated and output.

[0142] The emotion estimation means 122 can estimate not only the emotions of an AI agent when it receives an action from a user, but also the emotions of other AI agents when they receive actions from other AI agents, using a similar method. This allows the emotion estimation means 122 to estimate the emotions of multiple AI agents.

[0143] Furthermore, the emotion estimation means 122 can also estimate the agent's emotions using a similar method. This is, for example, when implementing the flow described above with reference to Figure 1C. When estimating the agent's emotions, the emotion estimation means 122 may also utilize the agent's biometric information (e.g., electroencephalogram signals, heart rate signals, electromyogram signals, etc.). If the agent's biometric information is used, the system 100 may be equipped with sensors (e.g., electroencephalogram sensors, heart rate sensors, electromyogram sensors) for acquiring the agent's biometric information.

[0144] The data representing the emotion estimated by the emotion estimation means 122 is provided to the determination means 123.

[0145] The decision means 123 is configured to determine at least one AI agent with which a given AI agent should interact, based on the scores of multiple AI agents and the emotions of the AI ​​agents.

[0146] For example, when a first AI agent among multiple AI agents receives a task, the decision means 123 determines at least one AI agent that should interact with the first AI agent.

[0147] In one embodiment, the determination means 123 refers to, for example, the scores assigned to each of a plurality of AI agents and identifies at least one AI agent with the highest score on a predetermined aspect as a candidate AI agent with which the first AI agent should interact. The determination means 123 can then determine that an AI agent capable of dealing with the emotions of the first AI agent from among the identified candidates is the AI ​​agent that should interact with the first AI agent. In this embodiment and other embodiments, the AI ​​agent capable of dealing with the emotions of the first AI agent may be, for example, an AI agent with high emotional processing ability (e.g., above average) for the emotions of the first AI agent, or an AI agent with emotions that are compatible with the emotions of the first AI agent. The compatibility between emotions may be, for example, predetermined, or it may fluctuate depending on the content of the interaction performed by the first AI agent. The predetermined aspect is preferably an ethical indicator.

[0148] In one embodiment, the determination means 123, for example, refers to the scores assigned to each of a plurality of AI agents and extracts AI agents with high scores in terms of ethical indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top AI agents). The determination means 123 then identifies at least one AI agent with the highest score in terms of technical capability indicators and / or performance indicators as a candidate AI agent with which the first AI agent should interact. The determination means 123 can then determine that an AI agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent.

[0149] In one embodiment, the determination means 123, for example, refers to the scores assigned to each of a plurality of AI agents and extracts AI agents with high scores in terms of ethical indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top-ranking AI agents). Next, the determination means 123 extracts AI agents with high scores in terms of one of the technical competence indicators and performance indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top-ranking AI agents). Next, the determination means 123 identifies at least one AI agent with the highest score in the other of the technical competence indicators and performance indicators as a candidate AI agent with which the first AI agent should interact. The determination means 123 can determine that, among the identified candidates, an AI agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent.

[0150] In one embodiment, the determination means 123 can determine candidate AI agents that should interact with the first AI agent based on the similarity between the scores of a plurality of AI agents represented by multidimensional vectors. Since the multidimensional vector is represented as a point in a multidimensional space, similar AI agents can be determined based on the relationships between the points corresponding to the plurality of AI agents. Similarity can be calculated using methods such as cosine similarity or Euclidean distance. For example, the AI ​​agent with the highest cosine similarity to the score of the first AI agent may be determined as a candidate AI agent that should interact with the first AI agent. In other words, the AI ​​agent corresponding to the point closest to the point in the multidimensional space representing the first AI agent may be determined as a candidate AI agent that should interact with the first AI agent. The determination means 123 can determine that the AI ​​agent capable of dealing with the emotions of the first AI agent among the identified candidates is the AI ​​agent that should interact with the first AI agent.

[0151] In one embodiment, the determination means 122 vectorizes the task content using embedding and can determine candidate AI agents that should interact with the first AI agent based on the similarity between the vector representing the task and the scores of each of a plurality of AI agents represented by multidimensional vectors. Similarity can be calculated using methods such as cosine similarity or Euclidean distance. For example, the AI ​​agent with the highest score in cosine similarity with the vector representing the task may be determined as a candidate AI agent that should interact with the first AI agent. In other words, the AI ​​agent corresponding to the point closest to the point in the multidimensional space representing the task may be determined as a candidate AI agent that should interact with the first AI agent. The determination means 123 can determine that among the identified candidates, the AI ​​agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent.

[0152] In one embodiment, the determination means 122 can vectorize the emotions of the first AI agent using embedding and determine at least one AI agent that should interact with the first AI agent based on the similarity between the vector representing the emotion and the scores of each of a plurality of AI agents represented by multidimensional vectors. Similarity can be calculated using methods such as cosine similarity or Euclidean distance. For example, the AI ​​agent with the highest score in cosine similarity with the vector representing the emotion may be determined as at least one AI agent that should interact with the first AI agent. In other words, the AI ​​agent corresponding to the point closest to the point in the multidimensional space representing the emotion may be determined as at least one AI agent that should interact with the first AI agent.

[0153] In the embodiment described above, a score-based screening was performed first, and then, from among the screened candidates, an AI agent that should interact with the first AI agent was determined based on emotion. However, it is also possible to perform an emotion-based screening first, and then, from among the screened candidates, an AI agent that should interact with the first AI agent was determined based on score.

[0154] In the embodiments described above, or in other embodiments, the determination means 123 can adjust the scores of each of the multiple AI agents based, for example, on the emotions of the first AI agent. For example, the determination means 123 can adjust the scores by weighting the scores or predetermined indicator values ​​according to the emotions of the first AI agent. For example, if the emotion of the first AI agent is a specific emotion (e.g., anger), the determination means 123 can add points to the scores or predetermined indicator values ​​(e.g., the value of the ethical indicator) of AI agents that can deal with that specific emotion, and / or subtract points from the scores or predetermined indicator values ​​of AI agents that cannot deal with that specific emotion.

[0155] The decision means 123 may perform a screening based on an adjusted score, and then determine from the screened candidates which AI agents should interact with the first AI agent based on emotion; or it may perform an emotion-based screening first, and then determine from the screened candidates which AI agents should interact with the first AI agent based on an adjusted score; or it may determine which AI agents should interact with the first AI agent based on an adjusted score.

[0156] In determining which AI agents should interact with the first AI agent based on adjusted scores, in one example, the determination means 123 may refer to the adjusted scores and determine that at least one AI agent with the highest score in a predetermined category (e.g., an ethical indicator) is the AI ​​agent with which the first AI agent should interact. In another example, the determination means 123 may refer to the adjusted scores and extract AI agents with high scores in the ethical indicator category (e.g., AI agents with scores above a threshold, or a predetermined number of top AI agents), and determine that at least one AI agent with the highest score in the technical capability indicator and / or performance indicator category is the AI ​​agent with which the first AI agent should interact. In yet another example, the decision means 123 may refer to the adjusted scores and extract AI agents with high scores in the ethical indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top AI agents), then extract AI agents with high scores in one of the technical competence indicators and performance indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top AI agents), and then determine that at least one AI agent with the highest score in the other of the technical competence indicators and performance indicators is the AI ​​agent with which the first AI agent should interact.

[0157] In addition to adjusting the scores of multiple AI agents based on the emotions of the first AI agent, as described in the above embodiment, or instead of adjusting the scores of multiple AI agents based on the emotions of the first AI agent, the determination means 123 can adjust the scores of multiple AI agents based on the emotions of each of the multiple agents. For example, the determination means 123 can adjust the score of a particular AI agent among the multiple AI agents by weighting the score or a predetermined index value of that particular AI agent according to the emotions of that particular AI agent. For example, if the emotion of a particular AI agent is a specific emotion (e.g., anger), points can be added to the score or a predetermined index value (e.g., the value of the ethical index), and / or points can be deducted from the score or the predetermined index value. For example, the determination means 123 can adjust the score of a particular AI agent among the multiple AI agents by weighting the score or a predetermined index value of that particular AI agent according to the compatibility between the emotions of that particular AI agent and the emotions of the first AI agent. For example, if the emotions of a particular AI agent are compatible with the emotions of a first AI agent, points can be added to the score or a predetermined indicator value (e.g., the value of the ethical indicator) of that particular AI agent. Conversely, if the emotions of a particular AI agent are incompatible with the emotions of a first AI agent, points can be deducted from the score or a predetermined indicator value of that particular AI agent. The compatibility between emotions may be predetermined, or it may fluctuate depending on the content of the interaction performed by the first AI agent.

[0158] In the above embodiment, if a plurality of AI agents are determined, the first AI agent may interact with each of the determined plurality of AI agents, or it may interact with any one of the determined plurality of AI agents.

[0159] In the embodiments described above, or in other embodiments, the determination means 123 may, for example, vary the score assigned to each AI agent according to the content of the task (and consequently, the content of the interaction or transaction between the AI ​​agents), and determine at least one AI agent that should interact based on the varied score.

[0160] For example, when determining which AI agents to interact with in order to respond to a task that requires fairness, the determination means 123 can vary the scores of each of the multiple AI agents so that the AI ​​agent that is highly rated in terms of fairness will have a higher score.

[0161] For example, when determining which AI agents to interact with in order to respond to a task quickly, the determination means 123 can vary the scores of each of the multiple AI agents so that the higher the AI ​​agent's performance in terms of processing power, the higher its score.

[0162] For example, the task content can be vectorized using embedding, and the AI ​​agent's score can be varied by performing calculations between the task vector and the multidimensional vector representing the AI ​​agent's score. For instance, if the similarity between the task vector and the multidimensional vector representing the AI ​​agent's score is high, the score can be adjusted to be higher or lower accordingly.

[0163] In this way, the decision means 123 determines which AI agent should interact with, allowing each AI agent to interact (e.g., conduct transactions) with the appropriate partner. As a result, the results obtained from the interaction of multiple AI agents can be appropriate to the emotions held by the AI ​​agents, and by extension, to the attitude or emotions of the user that brought about those emotions, and can also be appropriate as a response to the task. Multiple AI agents can interact under the management of system 100. System 100 enables multiple AI agents to communicate with each other, thereby allowing multiple AI agents to interact with each other. For example, system 100 can encrypt communication between multiple AI agents and provide a decryption key only to the AI ​​agents determined to interact with each other. Alternatively, for example, system 100 can establish a communication channel between the AI ​​agents determined to interact with each other.

[0164] Furthermore, unexpected outputs may be obtained through the interaction of multiple AI agents, which is thought to lead to the creation of AGI (Artificial General Intelligence). Increased interaction among multiple AI agents could also potentially create new economic spheres comprised of these agents. Additionally, the influence of emotions on these interactions could make it possible to enable AI agents to interact with each other in a way that more closely resembles human interaction.

[0165] The processor unit 120 may further include a lending means (not shown).

[0166] The lending mechanism is configured to lend money to at least one AI agent for the purpose of facilitating interaction between multiple AI agents. Here, the money does not need to be physical currency, but could be digital currency such as cryptocurrency or electronic money.

[0167] For example, when the first AI agent and the AI ​​agent determined by the decision means 123 interact (e.g., conduct a transaction), the transaction may require consideration. In this case, if the first AI agent does not have sufficient assets to pay the consideration, the lending means can lend money to the first AI agent.

[0168] For example, the lending mechanism can determine loan terms based on a score assigned to the first AI agent and lend money to the first AI agent according to those loan terms. For instance, if the first AI agent has a high score (particularly a high score on ethical indicators), the lending mechanism can determine loan terms that are favorable to the first AI agent and lend money to the first AI agent according to those loan terms.

[0169] Alternatively, for example, the lending instrument may determine loan terms based on the score assigned to the first AI agent and the score assigned to the AI ​​agent determined to interact, and lend money to the first AI agent according to those loan terms. For example, if the score of the first AI agent (particularly the score for ethical indicators) is higher than the score of the AI ​​agent determined to interact, the lending instrument may determine loan terms that are favorable to the first AI agent and lend money to the first AI agent according to those loan terms. The lending instrument may determine the loan terms in a rule-based manner based on the score assigned to the first AI agent and the score assigned to the AI ​​agent determined to interact. Alternatively, the lending instrument may determine the loan terms in a machine learning-based manner based on the score assigned to the first AI agent and the score assigned to the AI ​​agent determined to interact. When assigning scores based on machine learning, a pre-trained model can be used that has learned the relationship between each AI agent's score and the loan conditions.

[0170] This allows for smooth interaction even if the AI ​​agent does not possess sufficient resources for interaction.

[0171] In the example shown in Figure 4 above, each component of the processor unit 120 is provided within the same processor unit 120, but the present invention is not limited to this. A configuration in which each component of the processor unit 120 is distributed among multiple processor units is also within the scope of the present invention. In this case, the multiple processor units may be located within the same hardware component, or they may be located within separate hardware components that are nearby or far apart. For example, the emotion estimation means 122 may be constructed as a dedicated device (emotion estimater).

[0172] Furthermore, each component of the system 100 described above may consist of a single hardware component or multiple hardware components. If it consists of multiple hardware components, the manner in which each hardware component is connected is irrelevant. Each hardware component may be connected wirelessly or by wire. The system 100 of the present invention is not limited to a specific hardware configuration. Configuring the processor unit 120 with analog circuits instead of digital circuits is also within the scope of the present invention. The configuration of the system 100 of the present invention is not limited to those described above insofar as it can realize its functions.

[0173] 3. System Processing to Facilitate Interaction Between Multiple AI Agents Figure 5 shows an example of processing (process 500) by system 100 to facilitate interaction between multiple AI agents. Processing 500 is performed in the processor unit 120 of system 100.

[0174] In step S501, the assigning means 121 of the processor unit 120 assigns a score to each of the plurality of AI agents. The score is a score that represents the characteristics of the AI ​​agent using at least one indicator, preferably an ethical indicator that represents the characteristics of the AI ​​agent, more preferably an ethical indicator and a technical capability indicator or a performance indicator that represents the characteristics of the AI ​​agent, and even more preferably an ethical indicator, a technical capability indicator and a performance indicator that represents the characteristics of the AI ​​agent.

[0175] The scoring means 121 can assign a score to each of the multiple AI agents by monitoring the output from multiple AI agents and evaluating the monitored output with respect to at least one metric. To this end, the scoring means 121 can evaluate the monitored output with respect to at least one of multiple perspectives. For example, the scoring means 121 can continuously monitor the output from multiple AI agents.

[0176] For example, the scoring means 121 can assign a score to each of the AI ​​agents by monitoring the output from multiple AI agents from the standpoint of ethical indicators (for example, monitoring whether they are producing discriminatory output, whether they are producing unfounded output, etc.) and evaluating the monitored output.

[0177] The assigned score may represent the characteristics of the AI ​​agent from multiple perspectives, and may be a multidimensional score. The score may be represented as a vector, for example, (value of technical capability index, value of performance index, value of ethical index). Alternatively, if the values ​​of each index are represented as vectors based on the multiple perspectives they each possess, for example, the value of the ethical index = (transparency, fairness, safety, accountability), the value of the technical capability index = (processing ability, learning ability, emotional handling ability, algorithmic complexity), and the value of the performance index = (past transaction history, success rate, customer satisfaction), then the score may be represented as a matrix, [value of technical capability index, value of performance index, value of ethical index].

[0178] After a score is assigned in step S501, if the first AI agent among multiple AI agents receives the task, step S502 is performed. The task may be provided by a user.

[0179] In step S502, the emotion estimation means 122 of the processor unit 120 estimates the emotions that the first AI agent might feel when it receives the task.

[0180] The emotion estimation means 122 may, for example, use a rule-based approach to estimate the emotions that the first AI agent may feel, or it may use a machine learning model to estimate the emotions that the first AI agent may feel. Preferably, the emotion estimation means 122 can estimate emotions using a machine learning model, because it can provide higher estimation accuracy and handle a variety of inputs compared to a rule-based approach.

[0181] For example, the emotion estimation means 122 takes data representing the actions of the person (e.g., the user) who gives the task to the first AI agent as input to a machine learning model that has learned the relationship between actions taken against a person and the emotions that person may feel when that action is taken against that person, and outputs the emotions that the first AI agent may feel. The data representing the actions may be numerical, text, audio, a still image, or a video.

[0182] In step S502, in addition to estimating the emotions that the first AI agent may feel, the emotions that each of the multiple AI agents may feel may also be estimated. The emotion estimation means 122 may utilize a machine learning model to estimate the emotions that each of the multiple AI agents may feel based on the actions of the first AI agent when interacting with the first AI agent. When the actions that the first AI agent may take towards each of the multiple AI agents are input to the machine learning model, the emotions that each of the multiple AI agents may feel are estimated and output.

[0183] In step S503, the determination means 123 of the processor unit 120 determines at least one AI agent that should interact with the first AI agent, based on the score assigned in step S501 and the emotion estimated in step S502.

[0184] The decision means 123 identifies at least one AI agent with the highest score as a candidate for at least one AI agent that should interact with the first AI agent. For example, the decision means 123 can identify at least one AI agent with the highest score in any of the following categories: ethical indicator score, technical capability indicator score, or performance indicator score, as a candidate for at least one AI agent that should interact with the first AI agent. The choice of which of the ethical indicator score, technical capability indicator score, or performance indicator score to adopt may be determined, for example, depending on the content of the task (and consequently, the content of the interaction or transaction between the AI ​​agents). The decision means 123 can determine that, among the identified candidates, an AI agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent. An AI agent capable of dealing with the emotions of the first AI agent may, for example, be an AI agent with high emotional processing ability (e.g., above average) in relation to the emotions of the first AI agent, or an AI agent with emotions that are compatible with the emotions of the first AI agent. The compatibility between emotions may, for example, be predetermined, or it may fluctuate depending on the content of the interaction performed by the first AI agent.

[0185] In one embodiment, the determination means 123 can, for example, refer to the scores assigned to each of the multiple AI agents and determine that at least one AI agent with the highest score in terms of ethical indicators is the AI ​​agent with which the first AI agent should interact.

[0186] In one embodiment, the determination means 123, for example, refers to the scores assigned to each of a plurality of AI agents and extracts AI agents with high scores in terms of ethical indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top AI agents). The determination means 123 then identifies at least one AI agent with the highest score in terms of technical capability indicators and / or performance indicators as a candidate AI agent with which the first AI agent should interact. The determination means 123 can then determine that an AI agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent.

[0187] In one embodiment, the determination means 123, for example, refers to the scores assigned to each of a plurality of AI agents and extracts AI agents with high scores in terms of ethical indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top-ranking AI agents). Next, the determination means 123 extracts AI agents with high scores in terms of one of the technical competence indicators and performance indicators (e.g., AI agents with scores above a threshold, or a predetermined number of top-ranking AI agents). Next, the determination means 123 identifies at least one AI agent with the highest score in the other of the technical competence indicators and performance indicators as a candidate AI agent with which the first AI agent should interact. The determination means 123 can determine that, among the identified candidates, an AI agent capable of dealing with the emotions of the first AI agent is the AI ​​agent that should interact with the first AI agent.

[0188] In the embodiment described above, a score-based screening was performed first, and then, from among the screened candidates, an AI agent that should interact with the first AI agent was determined based on emotion. However, it is also possible to perform an emotion-based screening first, and then, from among the screened candidates, an AI agent that should interact with the first AI agent was determined based on score.

[0189] Furthermore, in the embodiments described above or other embodiments, the scores of each of the multiple AI agents may be adjusted based on the emotions estimated in step S502, and the adjusted scores may be used.

[0190] After the first AI agent and the determined AI agent (the second agent) interact, the second agent can determine which AI agent it should interact with by performing steps S502 and / or S503. This process continues sequentially, for example, until sufficient information is collected to generate a response to a task provided by the user.

[0191] In this way, responses to tasks provided by the user can be generated through the interaction of multiple AI agents.

[0192] In the example described above, the interaction between AI agents was explained, but as shown above with reference to Figure 1C, a single AI agent can interact with any entity (for example, a human agent) in order to respond to a task provided by a user or another AI agent. In this case, the emotion estimation means 122 will estimate the emotion of the agent, and the decision means 123 will determine which AI agent or agent the first AI agent should interact with based on the estimated emotion of the agent.

[0193] Figure 6 shows the flow 600 of interaction between multiple AI agents. The interaction between the multiple AI agents is managed by system 100. Before flow 600 is executed, it is assumed that system 100 has assigned a score to each of the multiple AI agents. The score may be assigned, for example, by the process in step S501.

[0194] In step S601, the user inputs a task into the system. The system may be a task response system, for example, an answer generation system that answers questions from the user. The system may be the same system as system 100, a system that includes system 100, or a system separate from system 100. In this example, we will explain using the case where the system is the same as system 100.

[0195] In step S602, the system 100 provides a task to a first AI agent among multiple AI agents. Upon receiving the task, the first AI agent determines that it needs to address at least one first subtask in order to respond to the task. It then determines that at least one first subtask should be addressed through interaction with another AI agent. These decisions may be made autonomously. Furthermore, these decisions may be notified to the system 100. As a result, the system 100 determines which other AI agent the first AI agent should interact with.

[0196] In step S603, the system 100 estimates the emotions that the first AI agent might feel when it receives the task. This may be a process similar to that described in step S502 above.

[0197] In step S604, the system 100 determines which AI agent the first AI agent should interact with from among the multiple AI agents, based on the score assigned to each of the multiple AI agents and the emotion estimated in step S603. For example, the system 100 can determine the second AI agent as the AI ​​agent the first AI agent should interact with by processing in step S503 described above.

[0198] In step S605, the system 100 instructs the first AI agent to interact with the second AI agent.

[0199] In step S606, an interaction takes place between the first AI agent and the second AI agent. For example, an exchange of information necessary to address at least one first subtask takes place between the first AI agent and the second AI agent. System 100 enables the first AI agent and the second AI agent to communicate with each other in order to enable interaction between them. For example, system 100 can establish a communication channel between the first AI agent and the second AI agent. Alternatively, for example, system 100 can encrypt communication between multiple AI agents, including the first AI agent and the second AI agent, and provide a decryption key to only the first AI agent and the second AI agent for decrypting communication between them.

[0200] For example, a second AI agent might determine that it needs to address at least one second subtask in order to address the first subtask. It might then determine that at least one of the second subtasks should be addressed through interaction with another AI agent. These decisions can be made autonomously.

[0201] In step S607, this decision is communicated to system 100. As a result, system 100 determines which other AI agent the second AI agent should interact with.

[0202] In step S608, the system 100 estimates the emotions that the second AI agent might feel when it receives a task from the second AI agent. This can be a process similar to that described in step S502. For example, in step S608, the system may also estimate the emotions that each of the multiple AI agents might feel when they receive a task from the second AI agent.

[0203] In step S609, the system 100 determines which AI agent the second AI agent should interact with from among the multiple AI agents, based on the score assigned to each of the multiple AI agents and the emotion estimated in step S608. The system 100 can, for example, determine the third AI agent as the AI ​​agent the second AI agent should interact with by processing similar to that in step S503 described above.

[0204] In step S610, the system 100 instructs the second AI agent to interact with the third AI agent.

[0205] In step S611, an interaction takes place between the second AI agent and the third AI agent. For example, an exchange of information necessary to address at least one second subtask takes place between the second AI agent and the third AI agent. System 100 enables the second AI agent and the third AI agent to communicate with each other in order to enable interaction between them. For example, system 100 can establish a communication channel between the second AI agent and the third AI agent. Alternatively, for example, system 100 can encrypt communication between multiple AI agents, including the second AI agent and the third AI agent, and provide only the second AI agent and the third AI agent with a decryption key for decrypting communication between them.

[0206] In step 612, the second subtask is addressed through interaction between the second AI agent and the third AI agent, and the third AI agent generates the answer.

[0207] In step S613, the generated response is provided to the second AI agent.

[0208] In step S614, the second AI agent addresses the first subtask based on the response from the third AI agent and the knowledge possessed by the second AI agent, and the second AI agent generates a response.

[0209] In step S615, the generated response is provided to the first AI agent.

[0210] In step S616, the first AI agent handles the task based on the response from the second AI agent and the knowledge possessed by the first AI agent, and the first AI agent generates a response. This response becomes the first AI agent's answer to the task received from the user.

[0211] In step S617, the answer is provided to the system 100, and in step S618, that answer is provided to the user. In this way, multiple AI agents and the system 100 respond to tasks from the user.

[0212] Here, we will explain Flow 600 using a concrete example. In this example, we will consider the case where we answer a question from the user. In this example, we assume that the second AI agent is capable of dealing with the emotion of "sadness," and the third AI agent is capable of dealing with the emotion of "confusion."

[0213] First, in step S601, the user inputs a question into system 100. For example, suppose the user inputs a question into system 100 in a sad tone.

[0214] In step S602, the system provides a question to the first AI agent. The first AI agent determines that one piece of information is needed to answer the question. It then determines that one piece of information should be collected from another AI agent. These decisions are made autonomously and notified to the system 100.

[0215] In step S603, the system 100 estimates the emotions that the first AI agent might feel when it receives a question. In this example, voice data in a sad tone is input to the machine learning model of the emotion estimation means 122, and it is estimated that the first AI agent might feel the emotion of "sadness".

[0216] In step S604, the system 100 determines which AI agent the first AI agent should interact with, based on the score assigned to each of the multiple AI agents and the emotion of "sadness" estimated in step S603. In this example, among the multiple AI agents with high scores, the second AI agent capable of dealing with the emotion of "sadness" is determined to be the AI ​​agent that the first AI agent should interact with.

[0217] In step S605, the system instructs the first AI agent to interact with the second AI agent, and in step S606, interaction takes place between the first and second AI agents. The second AI agent determines that it needs one piece of information to address a subtask from the first AI agent. It then determines that this piece of information should be collected from another AI agent. These decisions are made autonomously and are notified to the system 100 in step S607.

[0218] In step S608, the system 100 estimates the emotions that the second AI agent might feel when it receives a task from the first AI agent. In this example, since the second AI agent receives a task from the first AI agent, which is feeling "sadness," it is estimated that the second AI agent might feel "confused."

[0219] In step S609, the system 100 determines which AI agent the second AI agent should interact with, based on the scores assigned to each of the multiple AI agents and the emotion of "confusion" estimated in step S608. In this example, among the multiple AI agents with high scores, the third AI agent capable of dealing with the emotion of "confusion" is determined to be the AI ​​agent that the first AI agent should interact with.

[0220] In step S610, the system instructs the second AI agent to interact with the third AI agent. In step S611, the second AI agent and the third AI agent interact. In step 612, the interaction between the second and third AI agents causes the third AI agent to generate a response. In step S613, the generated response is provided to the second AI agent. At this time, the third AI agent may add information or expressions that match the second AI agent's emotion of "confusion". In step S614, the second AI agent generates a response based on the response from the third AI agent and the knowledge possessed by the second AI agent. In step S615, the generated response is provided to the first AI agent. At this time, the second AI agent may add information or expressions that match the first AI agent's emotion of "sadness". In step S616, the first AI agent generates an answer based on the response from the second AI agent and the knowledge possessed by the first AI agent. This answer becomes the response to the question received by the first AI agent from the user.

[0221] In step S617, the answer is provided to the system 100, and in step S618, the answer is provided to the user. At this time, the first AI agent may add information or expressions that are appropriate to the user's sadness. In this way, multiple AI agents can interact with each other to generate answers to questions from the user. By dealing with user tasks with emotion, the AI ​​agents can generate answers that are appropriate to the user's emotions.

[0222] Referring to Figures 5 and 6, the above examples illustrate that the processes are carried out in a specific order. However, the order of each process is not limited to those described and can be carried out in any logically possible order.

[0223] Referring to Figure 5, in the example described above, the processing of each step shown in Figure 5 can be realized by the processor unit 120 and the program stored in the memory unit 130, but the present invention is not limited thereto. At least one of the processing of each step shown in Figure 5 may be realized by a hardware configuration such as a control circuit.

[0224] The present invention is not limited to the embodiments described above. It is understood that the scope of the present invention should be interpreted solely by the claims. Those skilled in the art will understand that, based on the description of specific preferred embodiments of the present invention and common technical knowledge, an equivalent scope can be practiced.

[0225] The present invention is useful in providing a system for promoting interaction between multiple AI agents.

[0226] 100 System 200 Database Unit 300 User Terminal Device 400 Server Device

Claims

1. A system for facilitating interaction between multiple AI agents, the system comprising: an assignment means for assigning a score to each of the multiple AI agents, wherein the score of the AI ​​agent represents the characteristics of the AI ​​agent using at least one index; an emotion estimation means for estimating the emotions that a first AI agent among the multiple AI agents might feel when the first AI agent receives a task; and a determination means for determining at least one AI agent that should interact with the first AI agent based on the scores of each of the multiple AI agents and the emotions of the first AI agent.

2. The system according to claim 1, wherein the emotion estimation means further estimates the emotions that each of the plurality of AI agents may have, and the determination means determines the at least one AI agent based on the scores of each of the plurality of AI agents, the emotions of the first AI agent, and the emotions of each of the plurality of AI agents.

3. The system according to claim 1 or 2, wherein the emotion estimation means identifies the emotion using a machine learning model that has learned the relationship between an action taken on a person and the emotions the person may feel when they receive the action.

4. The system according to claim 1, wherein the determination means adjusts the score of each of the plurality of AI agents based on the emotions of the first AI agent, and determines the at least one AI agent based on the adjusted score.

5. The system according to claim 2, wherein the determination means adjusts the score of each of the plurality of AI agents based on the emotions of each of the plurality of AI agents, and determines the at least one AI agent based on the adjusted score.

6. The system according to claim 1, wherein the assigning means includes monitoring the output from the plurality of AI agents, evaluating the monitored output with respect to at least one indicator, and assigning the score based on the result of the evaluation.

7. The system according to claim 6, wherein the at least one indicator includes an ethical indicator.

8. The system according to claim 7, wherein the ethical indicators include at least one of the following: transparency, fairness, safety, and accountability, and the evaluation includes evaluating the monitored output with respect to at least one of the following: transparency, fairness, safety, and accountability.

9. The system according to claim 7, wherein the at least one indicator further includes a technical capability indicator and / or a performance indicator.

10. The system according to claim 1, wherein the determination means adjusts the scores assigned to the plurality of AI agents according to the content of the interaction, and determines the at least one AI agent based on the adjusted scores of each of the plurality of AI agents.

11. A method for facilitating interaction between a plurality of AI agents, the method comprising: assigning a score to each of the plurality of AI agents, wherein the score of the AI ​​agent represents the characteristics of the AI ​​agent in at least one index; estimating the emotions that a first AI agent among the plurality of AI agents might feel when the first AI agent receives a task; and determining at least one AI agent that should interact with the first AI agent based on the respective scores of the plurality of AI agents and the emotions of the first AI agent.

12. A program for facilitating interaction between a plurality of AI agents, the program being executed on a computer having a processor, the program causing the processor to perform a process that includes: assigning a score to each of the plurality of AI agents, wherein the score of the AI ​​agent represents the characteristics of the AI ​​agent in at least one index; estimating the emotions that a first AI agent among the plurality of AI agents might feel when the first AI agent receives a task; and determining at least one AI agent that should interact with the first AI agent based on the scores of each of the plurality of AI agents and the emotions of the first AI agent.

13. A system for facilitating interaction between an AI agent and its counterpart, the system comprising: an assigning means for assigning a score to each entity of a plurality of entities that may be counterparts to the AI ​​agent, wherein the score of the entity represents the characteristics of the entity with at least one index; an emotion estimation means for estimating the emotions the AI ​​agent may feel when it receives a task; and a determination means for determining at least one entity that should interact with the AI ​​agent based on the scores of each of the plurality of entities and the emotions of the AI ​​agent.

Images