system

Abstract

The system according to this embodiment aims to detect and improve bias and compliance violations in AI agents. [Solution] The system according to the embodiment comprises a bias detection unit, a compliance check unit, an improvement suggestion unit, and a monitoring unit. The bias detection unit detects bias. The compliance check unit checks compliance based on the bias detected by the bias detection unit. The improvement suggestion unit makes improvement suggestions based on the results obtained by the compliance check unit. The monitoring unit monitors the agent in operation based on the improvement suggestions made by the improvement suggestion unit.

Description

Technical Field

[0001] The technology of the present disclosure relates to a system.

Background Art

[0002] Patent Document 1 discloses a method for controlling a persona chatbot, which is performed by at least one processor, and includes steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the prior art, there is a problem that comprehensive checking and improvement of the bias and compliance violations of AI agents have not been sufficiently carried out.

[0005] The system according to the embodiment aims to detect and improve the bias and compliance violations of AI agents.

Means for Solving the Problems

[0006] The system according to this embodiment comprises a bias detection unit, a compliance check unit, an improvement suggestion unit, and a monitoring unit. The bias detection unit detects bias. The compliance check unit checks compliance based on the bias detected by the bias detection unit. The improvement suggestion unit makes improvement suggestions based on the results obtained by the compliance check unit. The monitoring unit monitors the agent in operation based on the improvement suggestions made by the improvement suggestion unit. [Effects of the Invention]

[0007] The system according to this embodiment can detect and correct biases and compliance violations in AI agents. [Brief explanation of the drawing]

[0008] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Modes for carrying out the invention]

[0009] Hereinafter, an example of an embodiment of the system relating to the technology of this disclosure will be described with reference to the attached drawings.

[0010] First, let's explain the terminology used in the following explanation.

[0011] In the following embodiments, the signed processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Furthermore, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), or TPU (Tensor Processing Unit).

[0012] In the following embodiments, signed RAM (Random Access Memory) is a memory that temporarily stores information and is used as work memory by the processor.

[0013] In the following embodiments, the signed storage is one or more non-volatile storage devices that store various programs and various parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes.

[0014] In the following embodiments, the numbered communication I / F (Interface) is an interface including a communication processor, an antenna, and the like. The communication I / F manages communication between a plurality of computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0015] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it may be only A, only B, or a combination of A and B. Also, in this specification, when expressing three or more matters connected by "and / or", the same concept as "A and / or B" is applied.

[0016] [First Embodiment] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0017] As shown in FIG. 1, the data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0018] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. Also, the database 24 and the communication I / F 26 are connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0019] The smart device 14 comprises a computer 36, a receiving device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The receiving device 38, output device 40, and camera 42 are also connected to the bus 52.

[0020] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, and accepts user input. The touch panel 38A accepts user input via touch by detecting contact with an object (e.g., a pen or finger). The microphone 38B accepts user input via voice by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and microphone 38B to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 (see Figure 2) acquires the data indicating the user input.

[0021] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user by outputting the data in a form perceptible to the user (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0022] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0023] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0024] As shown in Figure 2, in the data processing device 12, a specific processing is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" related to the technology of this disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0025] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0026] In the smart device 14, specific processing is performed by the processor 46. The storage 50 stores a specific processing program 60. The specific processing program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 operating as a control unit 46A according to the specific processing program 60 executed on the RAM 48. The smart device 14 also has a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0027] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device having the data generation model 58. The data processing device 12 may also be a server device or a terminal device owned by a user (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of processing by the data processing system 10 according to the first embodiment will be described.

[0028] (Example of form 1) The AI ​​agent evaluation system according to an embodiment of the present invention is an advanced AI agent, "AgentInsight," for evaluating and improving the safety and ethical aspects of AI agents used by companies and organizations. This AI agent evaluation system comprehensively checks various aspects of AI agents, such as bias, discrimination, inappropriate responses, and privacy protection, to mitigate risks from the perspective of compliance and social responsibility. The AI ​​agent evaluation system provides evaluations and improvement suggestions tailored to each company's standards, enhancing the reliability and ethical aspects of AI agents. For example, the AI ​​agent evaluation system performs bias detection of AI agents. To mitigate the risk that AI agents may contain bias or discriminatory language towards specific groups, the AI ​​agent evaluation system automatically detects potential biases contained in datasets and agent responses and provides improvement suggestions tailored to each company. For example, it detects biases related to gender, age, and race and presents specific correction proposals. Next, the AI ​​agent evaluation system performs compliance checks. Based on legal regulations and privacy protection requirements, the AI ​​agent evaluation system evaluates whether the AI ​​agent's responses are appropriate and points out problems. For example, it addresses industry-specific legal regulations for financial institutions, medical institutions, and public institutions and identifies violation risks. Furthermore, the AI ​​agent evaluation system visualizes ethical aspects and safety. The AI ​​agent evaluation system clearly presents analysis results and areas for improvement, and provides customized advice to reduce ethical risks. For example, the AI ​​agent evaluation system automatically analyzes the responses of AI agents to check for bias and risks. It also utilizes generative AI for automated analysis of responses. Using generative AI, the AI ​​agent evaluation system analyzes whether agent responses contain bias or risks, and analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. Furthermore, the AI ​​agent evaluation system uses generative AI to conduct evaluations and improvements in accordance with the latest laws, regulations, and social values. Finally, the AI ​​agent evaluation system provides a continuous monitoring function.The AI ​​agent evaluation system regularly evaluates agents in operation and immediately notifies users of any new risks or biases detected. This ensures the continuous safety and ethical conduct of AI agents. As a result, the AI ​​agent evaluation system improves the reliability and ethical conduct of AI agents, providing an environment in which companies and organizations can use them with confidence. For example, in regulated and safety-conscious industries such as financial institutions, medical institutions, and public institutions, it can reduce bias and risks in AI agents and gain user trust. It can also enhance the reliability and ethical conduct of AI agents in commercial sectors such as technology companies, consumer goods and retail, and advertising and media. In short, the AI ​​agent evaluation system can evaluate and improve the safety and ethical conduct of AI agents used by companies and organizations.

[0029] The AI ​​agent evaluation system according to this embodiment comprises a bias detection unit, a compliance check unit, an improvement suggestion unit, and a monitoring unit. The bias detection unit detects bias. The bias detection unit automatically detects, for example, potential biases contained in the dataset or the agent's response. The bias detection unit can detect, for example, gender bias, racial bias, algorithmic bias, etc. The bias detection unit can detect, for example, unconscious bias or bias in the dataset. The compliance check unit checks compliance based on the bias detected by the bias detection unit. The compliance check unit evaluates whether the response is appropriate based, for example, on legal regulations and privacy protection requirements. The compliance check unit addresses, for example, industry-specific legal regulations such as those of financial institutions, medical institutions, and public institutions. The compliance check unit evaluates the response based on legal regulations such as GDPR and CCPA. The compliance check unit evaluates the response based on, for example, corporate policies and ethical standards. The improvement suggestion unit makes improvement suggestions based on the results obtained by the compliance check unit. The improvement suggestion unit makes correction suggestions based, for example, on the results of the bias detection unit and the compliance check unit. The Improvement Proposal Department proposes, for example, modifications to algorithms or datasets. The Improvement Proposal Department makes modification proposals based on, for example, the type of proposal and evaluation criteria. The Monitoring Department monitors the agents in operation based on the improvement proposals made by the Improvement Proposal Department. The Monitoring Department periodically evaluates the agents in operation and immediately notifies of any new risks or biases as soon as they are detected. The Monitoring Department evaluates the agents based on, for example, the frequency of monitoring and evaluation criteria. The Monitoring Department automatically analyzes the responses of the agents in operation to check for biases or risks. As a result, the AI ​​agent evaluation system according to this embodiment is capable of detecting bias in AI agents, performing compliance checks, proposing improvements, and monitoring.

[0030] The bias detection unit detects bias. For example, it automatically detects potential biases contained in datasets and agent responses. Specifically, the bias detection unit uses natural language processing techniques to analyze the agent's responses and detect biases and discriminatory expressions towards specific attributes. For example, in the case of gender bias, it checks whether the agent is making responses that contain inappropriate stereotypes towards a particular gender. In the case of racial bias, it checks whether discriminatory expressions or biases towards a particular race are included. In the case of algorithmic bias, it verifies whether the agent's learning algorithm is producing unfair results for a particular group. The bias detection unit utilizes statistical methods and machine learning models to detect unconscious biases and biases in datasets. For example, it analyzes the distribution of the dataset and detects biases in the data towards specific attributes. It also clusters the agent's responses and analyzes the trends in responses towards specific groups. As a result, the bias detection unit can detect potential biases contained in agent responses with high accuracy and ensure the fairness of the agents. Furthermore, the bias detection unit reports in detail the type and degree of the detected bias and provides specific guidelines for improvement. This means the bias detection unit plays a crucial role in early detection of agent bias and taking appropriate countermeasures.

[0031] The Compliance Check Department checks compliance based on biases detected by the Bias Detection Department. Specifically, the Compliance Check Department evaluates whether the agent's responses are appropriate based on legal regulations and privacy protection requirements. For example, in financial institutions, it verifies whether the agent is providing information that infringes on customer privacy, and in healthcare institutions, it verifies whether confidential patient information is being handled appropriately. In public institutions, it evaluates whether the agent is providing information fairly and transparently. The Compliance Check Department evaluates the agent's responses based on legal regulations such as GDPR and CCPA, and verifies whether legal requirements are being met. For example, based on GDPR, it checks whether the agent is properly managing user data and obtaining user consent. Based on CCPA, it evaluates whether the agent is respecting the consumer privacy rights of California. Furthermore, the Compliance Check Department evaluates the agent's responses based on corporate policies and ethical standards. For example, based on corporate ethical standards, it verifies whether the agent is adhering to guidelines to prevent discriminatory language. This allows the Compliance Check Department to rigorously evaluate whether the agent's responses conform to legal regulations and corporate policies, ensuring the agent's credibility and legal compliance.

[0032] The Improvement Proposal Department makes improvement proposals based on the results obtained by the Compliance Check Department. Specifically, the Improvement Proposal Department makes suggestions for modifying agent responses and algorithms based on the results of the Bias Detection Department and the Compliance Check Department. For example, if the Bias Detection Department detects a particular bias, the Improvement Proposal Department will present specific modification plans to eliminate that bias. If gender bias is detected, the department will review the agent's response content and propose modifications to eliminate gender stereotypes. If algorithmic bias is detected, the department will readjust the agent's learning algorithm and propose modifications to avoid unfair results for specific groups. The Improvement Proposal Department will present specific modification plans based on the type of proposal and evaluation criteria. For example, to correct bias in the dataset, the department will propose adding new data or resampling existing data. It will also review the response template to improve the agent's response content and propose using more neutral and fair language. This allows the Improvement Proposal Department to provide specific guidelines for eliminating agent bias and improving the quality and fairness of responses. Furthermore, the Improvement Proposal Department will monitor the implementation status of the proposals and make additional modification suggestions as needed. This allows the improvement suggestion department to support the continuous improvement of agents and enhance their reliability and quality.

[0033] The Monitoring Department monitors agents in operation based on improvement proposals submitted by the Improvement Proposal Department. Specifically, the Monitoring Department periodically evaluates agents in operation and immediately notifies of any new risks or biases detected. For example, it automatically analyzes agent responses to check for bias or risks. The Monitoring Department evaluates agents based on monitoring frequency and evaluation criteria to ensure that agent responses are always appropriate. The Monitoring Department analyzes agent responses in operation in real time and uses algorithms to detect abnormal patterns or inappropriate responses. For example, if an agent makes a biased response regarding a specific attribute, the Monitoring Department immediately detects the response and notifies the administrator. This allows the Monitoring Department to ensure that agent responses are always appropriate and maintain agent reliability. Furthermore, the Monitoring Department continuously evaluates agent performance and provides feedback for improvement. For example, it evaluates agent response speed and accuracy and proposes system adjustments or improvements as needed. This allows the Monitoring Department to optimize agent performance and improve agent quality and reliability.

[0034] The AI ​​agent evaluation system further includes an evaluation criteria update unit that uses generative AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The evaluation criteria update unit, for example, uses generative AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The evaluation criteria update unit, for example, updates evaluation criteria based on specific laws or social trends. The evaluation criteria update unit, for example, uses generative AI to automatically collect the latest laws, regulations, and social values ​​and reflect them in the evaluation criteria. The evaluation criteria update unit, for example, uses generative AI to collect relevant news articles and academic papers and reflect them in the evaluation criteria. This enables the evaluation criteria update unit to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. Some or all of the above processing in the evaluation criteria update unit may be performed using generative AI, for example, or without using generative AI. For example, the evaluation criteria update unit updates evaluation criteria based on information collected by generative AI. The evaluation criteria update unit can update evaluation criteria based on information collected by generative AI.

[0035] The AI ​​agent evaluation system further includes an automated response content analysis unit that uses a generative AI to analyze whether the agent's responses contain bias or risk. The automated response content analysis unit, for example, uses a generative AI to analyze whether the agent's responses contain bias or risk. The automated response content analysis unit, for example, uses a generative AI to automatically analyze the agent's responses and detect bias or risk. The automated response content analysis unit, for example, uses a generative AI to analyze the agent's responses and identify ethical or legal risks. The automated response content analysis unit, for example, uses a generative AI to analyze the agent's responses and identify frequently occurring inappropriate response patterns. This makes it possible for the automated response content analysis unit to analyze whether the agent's responses contain bias or risk. Some or all of the above processing in the automated response content analysis unit may be performed using a generative AI, for example, or without using a generative AI. For example, the automated response content analysis unit can use a generative AI to analyze the agent's responses and detect bias or risk. The automated response content analysis unit can use a generative AI to analyze the agent's responses and detect bias or risk.

[0036] The AI ​​agent evaluation system further includes a pattern recognition unit that analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The pattern recognition unit, for example, analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify frequently occurring inappropriate response patterns. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify ethical risks such as discriminatory remarks and privacy violations. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify frequently occurring bias patterns. This enables the pattern recognition unit to identify inappropriate response patterns and ethical risks. Some or all of the above-described processes in the pattern recognition unit may be performed using a generative AI, for example, or without a generative AI. For example, the pattern recognition unit can use a generative AI to analyze large amounts of data and identify frequently occurring inappropriate response patterns.

[0037] The bias detection unit can automatically detect potential biases contained in the dataset and agent responses. The bias detection unit automatically detects, for example, potential biases contained in the dataset and agent responses. The bias detection unit detects, for example, unconscious biases and biases in the dataset. The bias detection unit detects, for example, gender bias, racial bias, algorithmic bias, etc. This makes it possible for the bias detection unit to automatically detect potential biases. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit can detect potential biases by having the generative AI analyze the dataset and agent responses. The bias detection unit can detect potential biases by having the generative AI analyze the dataset and agent responses.

[0038] The compliance check department can evaluate whether a response is appropriate based on legal regulations and privacy protection requirements. The compliance check department evaluates whether a response is appropriate based on legal regulations and privacy protection requirements, for example. The compliance check department responds to industry-specific legal regulations, for example, those of financial institutions, medical institutions, and public institutions. The compliance check department evaluates responses based on legal regulations such as GDPR and CCPA, for example. The compliance check department evaluates responses based on corporate policies and ethical standards, for example. This makes it possible for the compliance check department to evaluate whether a response is appropriate based on legal regulations and privacy protection requirements. Some or all of the above processing in the compliance check department may be performed using, for example, generative AI, or not using generative AI. For example, the compliance check department may use generative AI to evaluate responses based on legal regulations and privacy protection requirements. The compliance check department may use generative AI to evaluate responses based on legal regulations and privacy protection requirements.

[0039] The improvement proposal unit can make correction suggestions based on the results of the bias detection unit and the compliance check unit. For example, the improvement proposal unit makes correction suggestions based on the results of the bias detection unit and the compliance check unit. For example, the improvement proposal unit proposes corrections to the algorithm or the dataset. For example, the improvement proposal unit makes correction suggestions based on the type of suggestion or evaluation criteria. This makes it possible for the improvement proposal unit to make correction suggestions based on the results of the bias detection unit and the compliance check unit. Some or all of the above processing in the improvement proposal unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement proposal unit may have a generative AI analyze the results of the bias detection unit and the compliance check unit and make correction suggestions. The improvement proposal unit may have a generative AI analyze the results of the bias detection unit and the compliance check unit and make correction suggestions.

[0040] The monitoring unit can periodically evaluate agents in operation and immediately notify if it detects any new risks or biases. For example, the monitoring unit periodically evaluates agents in operation and immediately notifies if it detects any new risks or biases. For example, the monitoring unit evaluates agents based on monitoring frequency and evaluation criteria. For example, the monitoring unit automatically analyzes the responses of agents in operation to check for biases or risks. This makes it possible for the monitoring unit to periodically evaluate agents in operation and immediately notify if it detects any new risks or biases. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit can use generative AI to analyze the responses of agents in operation and detect new risks or biases.

[0041] The bias detection unit can optimize its detection algorithm by taking into account biases specific to a particular industry or culture. For example, the bias detection unit optimizes its detection algorithm by taking into account biases specific to a particular industry or culture. For example, in the medical industry, the bias detection unit optimizes its algorithm to detect bias based on the patient's gender and age. For example, in the financial industry, the bias detection unit optimizes its algorithm to detect bias based on the customer's credit score. For example, in the education industry, the bias detection unit optimizes its algorithm to detect bias based on the student's academic performance and background. This makes it possible for the bias detection unit to optimize its detection algorithm by taking into account biases specific to a particular industry or culture. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit may use generative AI to optimize its detection algorithm by taking into account biases specific to a particular industry or culture. The bias detection unit can use generative AI to optimize its detection algorithm by taking into account biases specific to a particular industry or culture.

[0042] The bias detection unit can learn new bias patterns by referring to past bias detection history. For example, the bias detection unit learns new bias patterns by referring to past bias detection history. For example, the bias detection unit learns new bias patterns based on previously detected bias patterns and improves detection accuracy. For example, the bias detection unit analyzes past bias detection history to identify frequently occurring bias patterns and improve the detection algorithm. For example, the bias detection unit predicts new bias patterns by referring to past bias detection history and reflects them in the detection algorithm. This makes it possible for the bias detection unit to learn new bias patterns by referring to past bias detection history. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit learns new bias patterns by having the generative AI analyze past bias detection history. The bias detection unit can learn new bias patterns by having the generative AI analyze past bias detection history.

[0043] The bias detection unit can prioritize the detection of region-specific biases by considering the user's geographical location information. For example, the bias detection unit prioritizes the detection of region-specific biases by considering the user's geographical location information. For example, if the user is in a specific region, the bias detection unit prioritizes the detection of region-specific biases in that region. For example, the bias detection unit optimizes the algorithm for detecting region-specific biases based on the user's geographical location information. For example, the bias detection unit quickly detects region-specific biases by considering the user's geographical location information. This makes it possible for the bias detection unit to prioritize the detection of region-specific biases by considering the user's geographical location information. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit can detect region-specific biases by having the generative AI analyze the user's geographical location information.

[0044] The bias detection unit can analyze a user's social media activity and detect relevant biases. For example, the bias detection unit analyzes a user's social media activity and detects relevant biases. For example, the bias detection unit analyzes a user's social media activity and detects posts containing specific biases. For example, the bias detection unit optimizes an algorithm for detecting relevant biases based on the user's social media activity. For example, the bias detection unit quickly detects relevant biases by referring to the user's social media activity. This enables the bias detection unit to analyze a user's social media activity and detect relevant biases. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit may use generative AI to analyze a user's social media activity and detect relevant biases.

[0045] The compliance check unit can optimize its check algorithms by taking into account the legal regulations of specific industries and regions. For example, the compliance check unit optimizes its check algorithms by taking into account the legal regulations of specific industries and regions. For example, in the healthcare industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of patient privacy. For example, in the financial industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of customer data. For example, in the education industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of student data. This makes it possible for the compliance check unit to optimize its check algorithms by taking into account the legal regulations of specific industries and regions. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may use generative AI to optimize its check algorithms by taking into account the legal regulations of specific industries and regions. The compliance check unit may use generative AI to optimize its check algorithms by taking into account the legal regulations of specific industries and regions.

[0046] The compliance check unit can learn new violation patterns by referring to past violation history. The compliance check unit learns new violation patterns by referring to past violation history, for example. The compliance check unit learns new violation patterns based on past violation history and improves check accuracy, for example. The compliance check unit analyzes past violation history, identifies frequently occurring violation patterns, and improves the check algorithm, for example. The compliance check unit predicts new violation patterns by referring to past violation history and reflects them in the check algorithm, for example. This makes it possible for the compliance check unit to learn new violation patterns by referring to past violation history. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit has generative AI analyze past violation history and learn new violation patterns. The compliance check unit can have generative AI analyze past violation history and learn new violation patterns.

[0047] The compliance check unit can prioritize checking region-specific laws and regulations by taking into account the user's geographical location information. For example, the compliance check unit prioritizes checking region-specific laws and regulations by taking into account the user's geographical location information. For example, if the user is in a specific region, the compliance check unit prioritizes checking region-specific laws and regulations for that region. For example, the compliance check unit optimizes the algorithm for checking region-specific laws and regulations based on the user's geographical location information. For example, the compliance check unit quickly checks region-specific laws and regulations by taking into account the user's geographical location information. This makes it possible for the compliance check unit to prioritize checking region-specific laws and regulations by taking into account the user's geographical location information. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may have generative AI analyze the user's geographical location information and check region-specific laws and regulations. The compliance check unit may have generative AI analyze the user's geographical location information and check region-specific laws and regulations.

[0048] The compliance check unit can analyze a user's social media activity and check for relevant legal and regulatory violations. For example, the compliance check unit analyzes a user's social media activity and checks for relevant legal and regulatory violations. For example, the compliance check unit analyzes a user's social media activity and checks for posts that contain specific legal and regulatory violations. For example, the compliance check unit optimizes an algorithm for checking for relevant legal and regulatory violations based on a user's social media activity. For example, the compliance check unit refers to a user's social media activity and quickly checks for relevant legal and regulatory violations. This makes it possible for the compliance check unit to analyze a user's social media activity and check for relevant legal and regulatory violations. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may use generative AI to analyze a user's social media activity and check for relevant legal and regulatory violations. The compliance check unit may use generative AI to analyze a user's social media activity and check for relevant legal and regulatory violations.

[0049] The improvement suggestion unit can optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. For example, the improvement suggestion unit can optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding patient privacy protection in the medical industry. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding customer data protection in the financial industry. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding student data protection in the education industry. This makes it possible for the improvement suggestion unit to optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, generative AI, or without generative AI. For example, the improvement suggestion unit can optimize its suggestion algorithm by having generative AI consider improvement suggestions specific to a particular industry or culture. The improvement suggestion unit can optimize its suggestion algorithm by having generative AI consider improvement suggestions specific to a particular industry or culture.

[0050] The improvement suggestion unit can learn new improvement ideas by referring to past improvement suggestion history. For example, the improvement suggestion unit learns new improvement ideas by referring to past improvement suggestion history. For example, the improvement suggestion unit learns new improvement ideas based on past improvement suggestion history and improves suggestion accuracy. For example, the improvement suggestion unit analyzes past improvement suggestion history, identifies frequently occurring improvement ideas, and improves the suggestion algorithm. For example, the improvement suggestion unit predicts new improvement ideas by referring to past improvement suggestion history and reflects them in the suggestion algorithm. This makes it possible for the improvement suggestion unit to learn new improvement ideas by referring to past improvement suggestion history. Some or all of the above processes in the improvement suggestion unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement suggestion unit may have a generative AI analyze past improvement suggestion history and learn new improvement ideas. The improvement suggestion unit may have a generative AI analyze past improvement suggestion history and learn new improvement ideas.

[0051] The improvement suggestion unit can prioritize suggesting region-specific improvement plans by considering the user's geographical location information. For example, the improvement suggestion unit prioritizes suggesting region-specific improvement plans by considering the user's geographical location information. For example, if the user is in a specific region, the improvement suggestion unit prioritizes suggesting region-specific improvement plans for that region. For example, the improvement suggestion unit optimizes an algorithm for suggesting region-specific improvement plans based on the user's geographical location information. For example, the improvement suggestion unit quickly suggests region-specific improvement plans by considering the user's geographical location information. This makes it possible for the improvement suggestion unit to prioritize suggesting region-specific improvement plans by considering the user's geographical location information. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement suggestion unit may have a generative AI analyze the user's geographical location information and suggest region-specific improvement plans. The improvement suggestion unit may have a generative AI analyze the user's geographical location information and suggest region-specific improvement plans.

[0052] The improvement suggestion unit can analyze a user's social media activity and propose relevant improvement suggestions. For example, the improvement suggestion unit can analyze a user's social media activity and propose relevant improvement suggestions. For example, the improvement suggestion unit can analyze a user's social media activity and propose posts containing specific improvement suggestions. For example, the improvement suggestion unit can optimize an algorithm for proposing relevant improvement suggestions based on a user's social media activity. For example, the improvement suggestion unit can refer to a user's social media activity and quickly propose relevant improvement suggestions. This enables the improvement suggestion unit to analyze a user's social media activity and propose relevant improvement suggestions. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, generative AI, or without generative AI. For example, the improvement suggestion unit can have a generative AI analyze a user's social media activity and propose relevant improvement suggestions.

[0053] The monitoring unit can optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. For example, the monitoring unit can optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. For example, in the healthcare industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting patient privacy. For example, in the financial industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting customer data. For example, in the education industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting student data. This makes it possible for the monitoring unit to optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or not using generative AI. For example, the monitoring unit can optimize the monitoring algorithm by having generative AI take into account risks specific to a particular industry or culture. The monitoring unit can optimize the monitoring algorithm by having generative AI take into account risks specific to a particular industry or culture.

[0054] The monitoring unit can learn new risk patterns by referring to past monitoring history. For example, the monitoring unit learns new risk patterns by referring to past monitoring history. For example, the monitoring unit learns new risk patterns based on past monitoring history and improves monitoring accuracy. For example, the monitoring unit analyzes past monitoring history to identify frequently occurring risk patterns and improve the monitoring algorithm. For example, the monitoring unit predicts new risk patterns by referring to past monitoring history and reflects them in the monitoring algorithm. This makes it possible for the monitoring unit to learn new risk patterns by referring to past monitoring history. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit may have generative AI analyze past monitoring history and learn new risk patterns. The monitoring unit may have generative AI analyze past monitoring history and learn new risk patterns.

[0055] The monitoring unit can prioritize monitoring region-specific risks by taking into account the user's geographical location information. For example, the monitoring unit prioritizes monitoring region-specific risks by taking into account the user's geographical location information. For example, if the user is in a specific region, the monitoring unit prioritizes monitoring region-specific risks for that region. For example, the monitoring unit optimizes an algorithm for monitoring region-specific risks based on the user's geographical location information. For example, the monitoring unit quickly monitors region-specific risks by taking into account the user's geographical location information. This makes it possible for the monitoring unit to prioritize monitoring region-specific risks by taking into account the user's geographical location information. Some or all of the above processing in the monitoring unit may be performed using, for example, a generative AI, or without a generative AI. For example, the monitoring unit may use a generative AI to analyze the user's geographical location information and monitor region-specific risks. The monitoring unit may use a generative AI to analyze the user's geographical location information and monitor region-specific risks.

[0056] The monitoring unit can analyze users' social media activity and monitor related risks. For example, the monitoring unit analyzes users' social media activity and monitors related risks. For example, the monitoring unit analyzes users' social media activity and monitors posts containing specific risks. For example, the monitoring unit optimizes algorithms for monitoring related risks based on users' social media activity. For example, the monitoring unit refers to users' social media activity and quickly monitors related risks. This makes it possible for the monitoring unit to analyze users' social media activity and monitor related risks. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit may have generative AI analyze users' social media activity and monitor related risks. The monitoring unit may have generative AI analyze users' social media activity and monitor related risks.

[0057] The evaluation criteria update unit can optimize the update algorithm by taking into account criteria specific to a particular industry or culture. For example, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria specific to a particular industry or culture. For example, in the medical industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding patient privacy protection. For example, in the financial industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding customer data protection. For example, in the education industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding student data protection. This makes it possible for the evaluation criteria update unit to optimize the update algorithm by taking into account criteria specific to a particular industry or culture. Some or all of the above processing in the evaluation criteria update unit may be performed using, for example, generative AI, or without generative AI. For example, the evaluation criteria update unit may use generative AI to optimize the update algorithm by taking into account criteria specific to a particular industry or culture. The evaluation criteria update unit can use generative AI to optimize the update algorithm by taking into account criteria specific to a particular industry or culture.

[0058] The evaluation criteria update unit can prioritize updating region-specific criteria by taking into account the user's geographical location information. For example, the evaluation criteria update unit prioritizes updating region-specific criteria by taking into account the user's geographical location information. For example, if the user is in a specific region, the evaluation criteria update unit prioritizes updating the region-specific criteria for that region. For example, the evaluation criteria update unit optimizes the algorithm for updating region-specific criteria based on the user's geographical location information. For example, the evaluation criteria update unit quickly updates region-specific criteria by taking into account the user's geographical location information. This makes it possible for the evaluation criteria update unit to prioritize updating region-specific criteria by taking into account the user's geographical location information. Some or all of the above processing in the evaluation criteria update unit may be performed using, for example, a generative AI, or without a generative AI. For example, the evaluation criteria update unit may have a generative AI analyze the user's geographical location information and update region-specific criteria. The evaluation criteria update unit can have a generative AI analyze the user's geographical location information and update region-specific criteria.

[0059] The automated response analysis unit can optimize its analysis algorithm by considering response content specific to a particular industry or culture. For example, the automated response analysis unit optimizes its analysis algorithm by considering response content specific to a particular industry or culture. For example, in the medical industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting patient privacy. For example, in the financial industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting customer data. For example, in the education industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting student data. This makes it possible for the automated response analysis unit to optimize its analysis algorithm by considering response content specific to a particular industry or culture. Some or all of the above processing in the automated response analysis unit may be performed using, for example, generative AI, or without generative AI. For example, the automated response analysis unit may use generative AI to optimize its analysis algorithm by considering response content specific to a particular industry or culture. The automated response analysis unit may use generative AI to optimize its analysis algorithm by considering response content specific to a particular industry or culture.

[0060] The automated response analysis unit can prioritize the analysis of region-specific response content by considering the user's geographical location information. For example, the automated response analysis unit prioritizes the analysis of region-specific response content by considering the user's geographical location information. For example, if the user is in a specific region, the automated response analysis unit prioritizes the analysis of region-specific response content for that region. For example, the automated response analysis unit optimizes the algorithm for analyzing region-specific response content based on the user's geographical location information. For example, the automated response analysis unit quickly analyzes region-specific response content by considering the user's geographical location information. This makes it possible for the automated response analysis unit to prioritize the analysis of region-specific response content by considering the user's geographical location information. Some or all of the above processing in the automated response analysis unit may be performed using, for example, a generative AI, or without a generative AI. For example, the automated response analysis unit can use a generative AI to analyze the user's geographical location information and analyze region-specific response content.

[0061] The pattern recognition unit can optimize its recognition algorithm by considering patterns specific to a particular industry or culture. For example, the pattern recognition unit can optimize its recognition algorithm by considering patterns specific to a particular industry or culture. For example, in the medical industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to patient privacy protection. For example, in the financial industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to customer data protection. For example, in the education industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to student data protection. This makes it possible for the pattern recognition unit to optimize its recognition algorithm by considering patterns specific to a particular industry or culture. Some or all of the above processing in the pattern recognition unit may be performed using, for example, generative AI, or without generative AI. For example, the pattern recognition unit can optimize its recognition algorithm by having the generative AI consider patterns specific to a particular industry or culture.

[0062] The pattern recognition unit can prioritize the recognition of region-specific patterns by taking into account the user's geographical location information. For example, the pattern recognition unit prioritizes the recognition of region-specific patterns by taking into account the user's geographical location information. For example, if the user is in a specific region, the pattern recognition unit prioritizes the recognition of region-specific patterns for that region. For example, the pattern recognition unit optimizes the algorithm for recognizing region-specific patterns based on the user's geographical location information. For example, the pattern recognition unit quickly recognizes region-specific patterns by taking into account the user's geographical location information. This makes it possible for the pattern recognition unit to prioritize the recognition of region-specific patterns by taking into account the user's geographical location information. Some or all of the above processing in the pattern recognition unit may be performed using, for example, a generative AI, or without a generative AI. For example, the pattern recognition unit may use a generative AI to analyze the user's geographical location information and recognize region-specific patterns. The pattern recognition unit can use a generative AI to analyze the user's geographical location information and recognize region-specific patterns.

[0063] The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows.

[0064] The AI ​​agent evaluation system can take into account the user's geographical location and perform evaluations based on region-specific laws, regulations, and cultural backgrounds. For example, in a particular region where privacy protection regulations are strict, the evaluation will be based on those local laws. In another region, bias detection will be performed based on specific cultural backgrounds. Furthermore, evaluation criteria that reflect the social values ​​of each region can be set. This allows the AI ​​agent evaluation system to perform evaluations that are tailored to region-specific requirements.

[0065] The AI ​​agent evaluation system can analyze a user's social media activity and provide evaluations based on the user's interests and values. For example, it can focus on detecting biases related to topics that users frequently mention on social media. It can estimate specific values ​​and ethical perspectives from the user's posts and set evaluation criteria based on them. Furthermore, it can customize how evaluation results are presented using data obtained from the user's social media activity. This allows the AI ​​agent evaluation system to provide evaluations tailored to each user's individual interests and values.

[0066] The following briefly describes the processing flow for example form 1.

[0067] Step 1: The bias detection unit automatically detects potential biases contained in the dataset and agent responses. For example, it can detect gender bias, racial bias, algorithmic bias, etc. It can also detect unconscious biases and biases in the dataset. Step 2: The compliance check unit checks compliance based on biases detected by the bias detection unit. For example, it evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. It addresses industry-specific legal regulations for financial institutions, healthcare institutions, and public institutions, and evaluates the response based on legal regulations such as GDPR and CCPA. It also evaluates the response based on corporate policies and ethical standards. Step 3: The Improvement Proposal Department makes improvement proposals based on the results obtained by the Compliance Check Department. For example, it makes correction proposals based on the results of the Bias Detection Department and the Compliance Check Department, proposing modifications to the algorithm or dataset. Correction proposals are made based on the type of proposal and evaluation criteria. Step 4: The Monitoring Department monitors the agents in operation based on the improvement proposals submitted by the Improvement Proposal Department. For example, it periodically evaluates the agents in operation and immediately notifies them if any new risks or biases are detected. It evaluates the agents based on the monitoring frequency and evaluation criteria, and automatically analyzes the responses of the agents in operation to check for biases or risks.

[0068] (Example of form 2) The AI ​​agent evaluation system according to an embodiment of the present invention is an advanced AI agent, "AgentInsight," for evaluating and improving the safety and ethical aspects of AI agents used by companies and organizations. This AI agent evaluation system comprehensively checks various aspects of AI agents, such as bias, discrimination, inappropriate responses, and privacy protection, to mitigate risks from the perspective of compliance and social responsibility. The AI ​​agent evaluation system provides evaluations and improvement suggestions tailored to each company's standards, enhancing the reliability and ethical aspects of AI agents. For example, the AI ​​agent evaluation system performs bias detection of AI agents. To mitigate the risk that AI agents may contain bias or discriminatory language towards specific groups, the AI ​​agent evaluation system automatically detects potential biases contained in datasets and agent responses and provides improvement suggestions tailored to each company. For example, it detects biases related to gender, age, and race and presents specific correction proposals. Next, the AI ​​agent evaluation system performs compliance checks. Based on legal regulations and privacy protection requirements, the AI ​​agent evaluation system evaluates whether the AI ​​agent's responses are appropriate and points out problems. For example, it addresses industry-specific legal regulations for financial institutions, medical institutions, and public institutions and identifies violation risks. Furthermore, the AI ​​agent evaluation system visualizes ethical aspects and safety. The AI ​​agent evaluation system clearly presents analysis results and areas for improvement, and provides customized advice to reduce ethical risks. For example, the AI ​​agent evaluation system automatically analyzes the responses of AI agents to check for bias and risks. It also utilizes generative AI for automated analysis of responses. Using generative AI, the AI ​​agent evaluation system analyzes whether agent responses contain bias or risks, and analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. Furthermore, the AI ​​agent evaluation system uses generative AI to conduct evaluations and improvements in accordance with the latest laws, regulations, and social values. Finally, the AI ​​agent evaluation system provides a continuous monitoring function.The AI ​​agent evaluation system regularly evaluates agents in operation and immediately notifies users of any new risks or biases detected. This ensures the continuous safety and ethical conduct of AI agents. As a result, the AI ​​agent evaluation system improves the reliability and ethical conduct of AI agents, providing an environment in which companies and organizations can use them with confidence. For example, in regulated and safety-conscious industries such as financial institutions, medical institutions, and public institutions, it can reduce bias and risks in AI agents and gain user trust. It can also enhance the reliability and ethical conduct of AI agents in commercial sectors such as technology companies, consumer goods and retail, and advertising and media. In short, the AI ​​agent evaluation system can evaluate and improve the safety and ethical conduct of AI agents used by companies and organizations.

[0069] The AI ​​agent evaluation system according to this embodiment comprises a bias detection unit, a compliance check unit, an improvement suggestion unit, and a monitoring unit. The bias detection unit detects bias. The bias detection unit automatically detects, for example, potential biases contained in the dataset or the agent's response. The bias detection unit can detect, for example, gender bias, racial bias, algorithmic bias, etc. The bias detection unit can detect, for example, unconscious bias or bias in the dataset. The compliance check unit checks compliance based on the bias detected by the bias detection unit. The compliance check unit evaluates whether the response is appropriate based, for example, on legal regulations and privacy protection requirements. The compliance check unit addresses, for example, industry-specific legal regulations such as those of financial institutions, medical institutions, and public institutions. The compliance check unit evaluates the response based on legal regulations such as GDPR and CCPA. The compliance check unit evaluates the response based on, for example, corporate policies and ethical standards. The improvement suggestion unit makes improvement suggestions based on the results obtained by the compliance check unit. The improvement suggestion unit makes correction suggestions based, for example, on the results of the bias detection unit and the compliance check unit. The Improvement Proposal Department proposes, for example, modifications to algorithms or datasets. The Improvement Proposal Department makes modification proposals based on, for example, the type of proposal and evaluation criteria. The Monitoring Department monitors the agents in operation based on the improvement proposals made by the Improvement Proposal Department. The Monitoring Department periodically evaluates the agents in operation and immediately notifies of any new risks or biases as soon as they are detected. The Monitoring Department evaluates the agents based on, for example, the frequency of monitoring and evaluation criteria. The Monitoring Department automatically analyzes the responses of the agents in operation to check for biases or risks. As a result, the AI ​​agent evaluation system according to this embodiment is capable of detecting bias in AI agents, performing compliance checks, proposing improvements, and monitoring.

[0070] The bias detection unit detects bias. For example, it automatically detects potential biases contained in datasets and agent responses. Specifically, the bias detection unit uses natural language processing techniques to analyze the agent's responses and detect biases and discriminatory expressions towards specific attributes. For example, in the case of gender bias, it checks whether the agent is making responses that contain inappropriate stereotypes towards a particular gender. In the case of racial bias, it checks whether discriminatory expressions or biases towards a particular race are included. In the case of algorithmic bias, it verifies whether the agent's learning algorithm is producing unfair results for a particular group. The bias detection unit utilizes statistical methods and machine learning models to detect unconscious biases and biases in datasets. For example, it analyzes the distribution of the dataset and detects biases in the data towards specific attributes. It also clusters the agent's responses and analyzes the trends in responses towards specific groups. As a result, the bias detection unit can detect potential biases contained in agent responses with high accuracy and ensure the fairness of the agents. Furthermore, the bias detection unit reports in detail the type and degree of the detected bias and provides specific guidelines for improvement. This means the bias detection unit plays a crucial role in early detection of agent bias and taking appropriate countermeasures.

[0071] The Compliance Check Department checks compliance based on biases detected by the Bias Detection Department. Specifically, the Compliance Check Department evaluates whether the agent's responses are appropriate based on legal regulations and privacy protection requirements. For example, in financial institutions, it verifies whether the agent is providing information that infringes on customer privacy, and in healthcare institutions, it verifies whether confidential patient information is being handled appropriately. In public institutions, it evaluates whether the agent is providing information fairly and transparently. The Compliance Check Department evaluates the agent's responses based on legal regulations such as GDPR and CCPA, and verifies whether legal requirements are being met. For example, based on GDPR, it checks whether the agent is properly managing user data and obtaining user consent. Based on CCPA, it evaluates whether the agent is respecting the consumer privacy rights of California. Furthermore, the Compliance Check Department evaluates the agent's responses based on corporate policies and ethical standards. For example, based on corporate ethical standards, it verifies whether the agent is adhering to guidelines to prevent discriminatory language. This allows the Compliance Check Department to rigorously evaluate whether the agent's responses conform to legal regulations and corporate policies, ensuring the agent's credibility and legal compliance.

[0072] The Improvement Proposal Department makes improvement proposals based on the results obtained by the Compliance Check Department. Specifically, the Improvement Proposal Department makes suggestions for modifying agent responses and algorithms based on the results of the Bias Detection Department and the Compliance Check Department. For example, if the Bias Detection Department detects a particular bias, the Improvement Proposal Department will present specific modification plans to eliminate that bias. If gender bias is detected, the department will review the agent's response content and propose modifications to eliminate gender stereotypes. If algorithmic bias is detected, the department will readjust the agent's learning algorithm and propose modifications to avoid unfair results for specific groups. The Improvement Proposal Department will present specific modification plans based on the type of proposal and evaluation criteria. For example, to correct bias in the dataset, the department will propose adding new data or resampling existing data. It will also review the response template to improve the agent's response content and propose using more neutral and fair language. This allows the Improvement Proposal Department to provide specific guidelines for eliminating agent bias and improving the quality and fairness of responses. Furthermore, the Improvement Proposal Department will monitor the implementation status of the proposals and make additional modification suggestions as needed. This allows the improvement suggestion department to support the continuous improvement of agents and enhance their reliability and quality.

[0073] The Monitoring Department monitors agents in operation based on improvement proposals submitted by the Improvement Proposal Department. Specifically, the Monitoring Department periodically evaluates agents in operation and immediately notifies of any new risks or biases detected. For example, it automatically analyzes agent responses to check for bias or risks. The Monitoring Department evaluates agents based on monitoring frequency and evaluation criteria to ensure that agent responses are always appropriate. The Monitoring Department analyzes agent responses in operation in real time and uses algorithms to detect abnormal patterns or inappropriate responses. For example, if an agent makes a biased response regarding a specific attribute, the Monitoring Department immediately detects the response and notifies the administrator. This allows the Monitoring Department to ensure that agent responses are always appropriate and maintain agent reliability. Furthermore, the Monitoring Department continuously evaluates agent performance and provides feedback for improvement. For example, it evaluates agent response speed and accuracy and proposes system adjustments or improvements as needed. This allows the Monitoring Department to optimize agent performance and improve agent quality and reliability.

[0074] The AI ​​agent evaluation system further includes an evaluation criteria update unit that uses generative AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The evaluation criteria update unit, for example, uses generative AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The evaluation criteria update unit, for example, updates evaluation criteria based on specific laws or social trends. The evaluation criteria update unit, for example, uses generative AI to automatically collect the latest laws, regulations, and social values ​​and reflect them in the evaluation criteria. The evaluation criteria update unit, for example, uses generative AI to collect relevant news articles and academic papers and reflect them in the evaluation criteria. This enables the evaluation criteria update unit to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. Some or all of the above processing in the evaluation criteria update unit may be performed using generative AI, for example, or without using generative AI. For example, the evaluation criteria update unit updates evaluation criteria based on information collected by generative AI. The evaluation criteria update unit can update evaluation criteria based on information collected by generative AI.

[0075] The AI ​​agent evaluation system further includes an automated response content analysis unit that uses a generative AI to analyze whether the agent's responses contain bias or risk. The automated response content analysis unit, for example, uses a generative AI to analyze whether the agent's responses contain bias or risk. The automated response content analysis unit, for example, uses a generative AI to automatically analyze the agent's responses and detect bias or risk. The automated response content analysis unit, for example, uses a generative AI to analyze the agent's responses and identify ethical or legal risks. The automated response content analysis unit, for example, uses a generative AI to analyze the agent's responses and identify frequently occurring inappropriate response patterns. This makes it possible for the automated response content analysis unit to analyze whether the agent's responses contain bias or risk. Some or all of the above processing in the automated response content analysis unit may be performed using a generative AI, for example, or without using a generative AI. For example, the automated response content analysis unit can use a generative AI to analyze the agent's responses and detect bias or risk. The automated response content analysis unit can use a generative AI to analyze the agent's responses and detect bias or risk.

[0076] The AI ​​agent evaluation system further includes a pattern recognition unit that analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The pattern recognition unit, for example, analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify frequently occurring inappropriate response patterns. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify ethical risks such as discriminatory remarks and privacy violations. The pattern recognition unit, for example, uses a generative AI to analyze large amounts of data and identify frequently occurring bias patterns. This enables the pattern recognition unit to identify inappropriate response patterns and ethical risks. Some or all of the above-described processes in the pattern recognition unit may be performed using a generative AI, for example, or without a generative AI. For example, the pattern recognition unit can use a generative AI to analyze large amounts of data and identify frequently occurring inappropriate response patterns.

[0077] The bias detection unit can automatically detect potential biases contained in the dataset and agent responses. The bias detection unit automatically detects, for example, potential biases contained in the dataset and agent responses. The bias detection unit detects, for example, unconscious biases and biases in the dataset. The bias detection unit detects, for example, gender bias, racial bias, algorithmic bias, etc. This makes it possible for the bias detection unit to automatically detect potential biases. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit can detect potential biases by having the generative AI analyze the dataset and agent responses. The bias detection unit can detect potential biases by having the generative AI analyze the dataset and agent responses.

[0078] The compliance check department can evaluate whether a response is appropriate based on legal regulations and privacy protection requirements. The compliance check department evaluates whether a response is appropriate based on legal regulations and privacy protection requirements, for example. The compliance check department responds to industry-specific legal regulations, for example, those of financial institutions, medical institutions, and public institutions. The compliance check department evaluates responses based on legal regulations such as GDPR and CCPA, for example. The compliance check department evaluates responses based on corporate policies and ethical standards, for example. This makes it possible for the compliance check department to evaluate whether a response is appropriate based on legal regulations and privacy protection requirements. Some or all of the above processing in the compliance check department may be performed using, for example, generative AI, or not using generative AI. For example, the compliance check department may use generative AI to evaluate responses based on legal regulations and privacy protection requirements. The compliance check department may use generative AI to evaluate responses based on legal regulations and privacy protection requirements.

[0079] The improvement proposal unit can make correction suggestions based on the results of the bias detection unit and the compliance check unit. For example, the improvement proposal unit makes correction suggestions based on the results of the bias detection unit and the compliance check unit. For example, the improvement proposal unit proposes corrections to the algorithm or the dataset. For example, the improvement proposal unit makes correction suggestions based on the type of suggestion or evaluation criteria. This makes it possible for the improvement proposal unit to make correction suggestions based on the results of the bias detection unit and the compliance check unit. Some or all of the above processing in the improvement proposal unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement proposal unit may have a generative AI analyze the results of the bias detection unit and the compliance check unit and make correction suggestions. The improvement proposal unit may have a generative AI analyze the results of the bias detection unit and the compliance check unit and make correction suggestions.

[0080] The monitoring unit can periodically evaluate agents in operation and immediately notify if it detects any new risks or biases. For example, the monitoring unit periodically evaluates agents in operation and immediately notifies if it detects any new risks or biases. For example, the monitoring unit evaluates agents based on monitoring frequency and evaluation criteria. For example, the monitoring unit automatically analyzes the responses of agents in operation to check for biases or risks. This makes it possible for the monitoring unit to periodically evaluate agents in operation and immediately notify if it detects any new risks or biases. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit can use generative AI to analyze the responses of agents in operation and detect new risks or biases.

[0081] The bias detection unit can estimate the user's emotions and adjust the accuracy of bias detection based on the estimated user emotions. For example, the bias detection unit estimates the user's emotions and adjusts the accuracy of bias detection based on the estimated user emotions. For example, if the user is angry, the bias detection unit increases the accuracy of bias detection and performs a more rigorous detection. For example, if the user is relaxed, the bias detection unit sets the accuracy of bias detection to a normal level and performs a standard detection. For example, if the user is anxious, the bias detection unit sets the accuracy of bias detection to a moderate level and performs a moderate detection. This makes it possible for the bias detection unit to adjust the accuracy of bias detection based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the bias detection unit may be performed using a generative AI, for example, or without a generative AI. For example, the bias detection unit uses a generating AI to analyze the user's emotions and adjust the accuracy of bias detection.

[0082] The bias detection unit can optimize its detection algorithm by taking into account biases specific to a particular industry or culture. For example, the bias detection unit optimizes its detection algorithm by taking into account biases specific to a particular industry or culture. For example, in the medical industry, the bias detection unit optimizes its algorithm to detect bias based on the patient's gender and age. For example, in the financial industry, the bias detection unit optimizes its algorithm to detect bias based on the customer's credit score. For example, in the education industry, the bias detection unit optimizes its algorithm to detect bias based on the student's academic performance and background. This makes it possible for the bias detection unit to optimize its detection algorithm by taking into account biases specific to a particular industry or culture. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit may use generative AI to optimize its detection algorithm by taking into account biases specific to a particular industry or culture. The bias detection unit can use generative AI to optimize its detection algorithm by taking into account biases specific to a particular industry or culture.

[0083] The bias detection unit can learn new bias patterns by referring to past bias detection history. For example, the bias detection unit learns new bias patterns by referring to past bias detection history. For example, the bias detection unit learns new bias patterns based on previously detected bias patterns and improves detection accuracy. For example, the bias detection unit analyzes past bias detection history to identify frequently occurring bias patterns and improve the detection algorithm. For example, the bias detection unit predicts new bias patterns by referring to past bias detection history and reflects them in the detection algorithm. This makes it possible for the bias detection unit to learn new bias patterns by referring to past bias detection history. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit learns new bias patterns by having the generative AI analyze past bias detection history. The bias detection unit can learn new bias patterns by having the generative AI analyze past bias detection history.

[0084] The bias detection unit can estimate the user's emotions and determine the priority of bias detection based on the estimated user emotions. For example, the bias detection unit estimates the user's emotions and determines the priority of bias detection based on the estimated user emotions. For example, if the user is angry, the bias detection unit sets the priority of bias detection to high and performs detection quickly. For example, if the user is relaxed, the bias detection unit sets the priority of bias detection to a normal level and performs standard detection. For example, if the user is anxious, the bias detection unit sets the priority of bias detection to a medium level and performs appropriate detection. This makes it possible for the bias detection unit to determine the priority of bias detection based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the bias detection unit may be performed using, for example, a generative AI, or without a generative AI. For example, the bias detection unit uses a generating AI to analyze the user's emotions and determine the priority for bias detection.

[0085] The bias detection unit can prioritize the detection of region-specific biases by considering the user's geographical location information. For example, the bias detection unit prioritizes the detection of region-specific biases by considering the user's geographical location information. For example, if the user is in a specific region, the bias detection unit prioritizes the detection of region-specific biases in that region. For example, the bias detection unit optimizes the algorithm for detecting region-specific biases based on the user's geographical location information. For example, the bias detection unit quickly detects region-specific biases by considering the user's geographical location information. This makes it possible for the bias detection unit to prioritize the detection of region-specific biases by considering the user's geographical location information. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit can detect region-specific biases by having the generative AI analyze the user's geographical location information.

[0086] The bias detection unit can analyze a user's social media activity and detect relevant biases. For example, the bias detection unit analyzes a user's social media activity and detects relevant biases. For example, the bias detection unit analyzes a user's social media activity and detects posts containing specific biases. For example, the bias detection unit optimizes an algorithm for detecting relevant biases based on the user's social media activity. For example, the bias detection unit quickly detects relevant biases by referring to the user's social media activity. This enables the bias detection unit to analyze a user's social media activity and detect relevant biases. Some or all of the above processing in the bias detection unit may be performed using, for example, generative AI, or without generative AI. For example, the bias detection unit may use generative AI to analyze a user's social media activity and detect relevant biases.

[0087] The compliance check unit can estimate the user's emotions and adjust the compliance check criteria based on the estimated user emotions. For example, if the user is angry, the compliance check unit may set the compliance check criteria strictly and perform a rigorous check. For example, if the user is relaxed, the compliance check unit may set the compliance check criteria to a normal level and perform a standard check. For example, if the user is anxious, the compliance check unit may set the compliance check criteria to a moderate level and perform a moderate check. This makes it possible for the compliance check unit to adjust the compliance check criteria based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit uses a generation AI to analyze the user's emotions and adjust the compliance check criteria.

[0088] The compliance check unit can optimize its check algorithms by taking into account the legal regulations of specific industries and regions. For example, the compliance check unit optimizes its check algorithms by taking into account the legal regulations of specific industries and regions. For example, in the healthcare industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of patient privacy. For example, in the financial industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of customer data. For example, in the education industry, the compliance check unit optimizes its check algorithms by taking into account the legal regulations concerning the protection of student data. This makes it possible for the compliance check unit to optimize its check algorithms by taking into account the legal regulations of specific industries and regions. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may use generative AI to optimize its check algorithms by taking into account the legal regulations of specific industries and regions. The compliance check unit may use generative AI to optimize its check algorithms by taking into account the legal regulations of specific industries and regions.

[0089] The compliance check unit can learn new violation patterns by referring to past violation history. The compliance check unit learns new violation patterns by referring to past violation history, for example. The compliance check unit learns new violation patterns based on past violation history and improves check accuracy, for example. The compliance check unit analyzes past violation history, identifies frequently occurring violation patterns, and improves the check algorithm, for example. The compliance check unit predicts new violation patterns by referring to past violation history and reflects them in the check algorithm, for example. This makes it possible for the compliance check unit to learn new violation patterns by referring to past violation history. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit has generative AI analyze past violation history and learn new violation patterns. The compliance check unit can have generative AI analyze past violation history and learn new violation patterns.

[0090] The compliance check unit can estimate the user's emotions and determine the priority of compliance checks based on the estimated user emotions. For example, if the user is angry, the compliance check unit will set a high priority for compliance checks and perform the checks quickly. If the user is relaxed, the compliance check unit will set the priority of compliance checks to a normal level and perform standard checks. If the user is anxious, the compliance check unit will set the priority of compliance checks to a medium level and perform moderate checks. This makes it possible for the compliance check unit to determine the priority of compliance checks based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit uses a generation AI to analyze user emotions and determine the priority of compliance checks.

[0091] The compliance check unit can prioritize checking region-specific laws and regulations by taking into account the user's geographical location information. For example, the compliance check unit prioritizes checking region-specific laws and regulations by taking into account the user's geographical location information. For example, if the user is in a specific region, the compliance check unit prioritizes checking region-specific laws and regulations for that region. For example, the compliance check unit optimizes the algorithm for checking region-specific laws and regulations based on the user's geographical location information. For example, the compliance check unit quickly checks region-specific laws and regulations by taking into account the user's geographical location information. This makes it possible for the compliance check unit to prioritize checking region-specific laws and regulations by taking into account the user's geographical location information. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may have generative AI analyze the user's geographical location information and check region-specific laws and regulations. The compliance check unit may have generative AI analyze the user's geographical location information and check region-specific laws and regulations.

[0092] The compliance check unit can analyze a user's social media activity and check for relevant legal and regulatory violations. For example, the compliance check unit analyzes a user's social media activity and checks for relevant legal and regulatory violations. For example, the compliance check unit analyzes a user's social media activity and checks for posts that contain specific legal and regulatory violations. For example, the compliance check unit optimizes an algorithm for checking for relevant legal and regulatory violations based on a user's social media activity. For example, the compliance check unit refers to a user's social media activity and quickly checks for relevant legal and regulatory violations. This makes it possible for the compliance check unit to analyze a user's social media activity and check for relevant legal and regulatory violations. Some or all of the above processing in the compliance check unit may be performed using, for example, generative AI, or without generative AI. For example, the compliance check unit may use generative AI to analyze a user's social media activity and check for relevant legal and regulatory violations. The compliance check unit may use generative AI to analyze a user's social media activity and check for relevant legal and regulatory violations.

[0093] The improvement suggestion unit can estimate the user's emotions and adjust the content of the improvement suggestion based on the estimated user emotions. For example, the improvement suggestion unit estimates the user's emotions and adjusts the content of the improvement suggestion based on the estimated user emotions. For example, if the user is angry, the improvement suggestion unit will make quick and specific improvement suggestions. For example, if the user is relaxed, the improvement suggestion unit will make detailed improvement suggestions. For example, if the user is anxious, the improvement suggestion unit will make reassuring improvement suggestions. This makes it possible for the improvement suggestion unit to adjust the content of the improvement suggestion based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the improvement suggestion unit may be performed using a generative AI, for example, or without a generative AI. For example, the improvement suggestion unit may have a generative AI analyze the user's emotions and adjust the content of the improvement suggestion. The improvement suggestion department uses a generation AI to analyze user emotions and adjust the content of improvement suggestions accordingly.

[0094] The improvement suggestion unit can optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. For example, the improvement suggestion unit can optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding patient privacy protection in the medical industry. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding customer data protection in the financial industry. For example, the improvement suggestion unit can optimize its algorithm by suggesting improvement suggestions regarding student data protection in the education industry. This makes it possible for the improvement suggestion unit to optimize its suggestion algorithm by considering improvement suggestions specific to a particular industry or culture. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, generative AI, or without generative AI. For example, the improvement suggestion unit can optimize its suggestion algorithm by having generative AI consider improvement suggestions specific to a particular industry or culture. The improvement suggestion unit can optimize its suggestion algorithm by having generative AI consider improvement suggestions specific to a particular industry or culture.

[0095] The improvement suggestion unit can learn new improvement ideas by referring to past improvement suggestion history. For example, the improvement suggestion unit learns new improvement ideas by referring to past improvement suggestion history. For example, the improvement suggestion unit learns new improvement ideas based on past improvement suggestion history and improves suggestion accuracy. For example, the improvement suggestion unit analyzes past improvement suggestion history, identifies frequently occurring improvement ideas, and improves the suggestion algorithm. For example, the improvement suggestion unit predicts new improvement ideas by referring to past improvement suggestion history and reflects them in the suggestion algorithm. This makes it possible for the improvement suggestion unit to learn new improvement ideas by referring to past improvement suggestion history. Some or all of the above processes in the improvement suggestion unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement suggestion unit may have a generative AI analyze past improvement suggestion history and learn new improvement ideas. The improvement suggestion unit may have a generative AI analyze past improvement suggestion history and learn new improvement ideas.

[0096] The improvement suggestion unit can estimate the user's emotions and determine the priority of improvement suggestions based on the estimated emotions. For example, if the user is angry, the improvement suggestion unit will set the priority of improvement suggestions high and make suggestions quickly. If the user is relaxed, the improvement suggestion unit will set the priority of improvement suggestions to a normal level and make standard suggestions. If the user is anxious, the improvement suggestion unit will set the priority of improvement suggestions to a medium level and make appropriate suggestions. This makes it possible for the improvement suggestion unit to determine the priority of improvement suggestions based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, generative AI or not using generative AI. For example, the improvement suggestion department uses a generation AI to analyze user emotions and determine the priority of improvement suggestions.

[0097] The improvement suggestion unit can prioritize suggesting region-specific improvement plans by considering the user's geographical location information. For example, the improvement suggestion unit prioritizes suggesting region-specific improvement plans by considering the user's geographical location information. For example, if the user is in a specific region, the improvement suggestion unit prioritizes suggesting region-specific improvement plans for that region. For example, the improvement suggestion unit optimizes an algorithm for suggesting region-specific improvement plans based on the user's geographical location information. For example, the improvement suggestion unit quickly suggests region-specific improvement plans by considering the user's geographical location information. This makes it possible for the improvement suggestion unit to prioritize suggesting region-specific improvement plans by considering the user's geographical location information. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, a generative AI, or without a generative AI. For example, the improvement suggestion unit may have a generative AI analyze the user's geographical location information and suggest region-specific improvement plans. The improvement suggestion unit may have a generative AI analyze the user's geographical location information and suggest region-specific improvement plans.

[0098] The improvement suggestion unit can analyze a user's social media activity and propose relevant improvement suggestions. For example, the improvement suggestion unit can analyze a user's social media activity and propose relevant improvement suggestions. For example, the improvement suggestion unit can analyze a user's social media activity and propose posts containing specific improvement suggestions. For example, the improvement suggestion unit can optimize an algorithm for proposing relevant improvement suggestions based on a user's social media activity. For example, the improvement suggestion unit can refer to a user's social media activity and quickly propose relevant improvement suggestions. This enables the improvement suggestion unit to analyze a user's social media activity and propose relevant improvement suggestions. Some or all of the above processing in the improvement suggestion unit may be performed using, for example, generative AI, or without generative AI. For example, the improvement suggestion unit can have a generative AI analyze a user's social media activity and propose relevant improvement suggestions.

[0099] The monitoring unit can estimate the user's emotions and adjust the monitoring frequency based on the estimated emotions. For example, the monitoring unit estimates the user's emotions and adjusts the monitoring frequency based on the estimated emotions. For example, if the user is angry, the monitoring unit sets the monitoring frequency high and checks frequently. For example, if the user is relaxed, the monitoring unit sets the monitoring frequency to a normal level and performs standard checks. For example, if the user is anxious, the monitoring unit sets the monitoring frequency to a moderate level and performs appropriate checks. This makes it possible for the monitoring unit to adjust the monitoring frequency based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the monitoring unit may be performed using, for example, a generative AI, or without a generative AI. For example, the monitoring unit uses a generative AI to analyze the user's emotions and adjusts the monitoring frequency. The monitoring unit uses a generating AI to analyze the user's emotions and adjust the monitoring frequency accordingly.

[0100] The monitoring unit can optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. For example, the monitoring unit can optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. For example, in the healthcare industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting patient privacy. For example, in the financial industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting customer data. For example, in the education industry, the monitoring unit can optimize the monitoring algorithm by taking into account risks related to protecting student data. This makes it possible for the monitoring unit to optimize the monitoring algorithm by taking into account risks specific to a particular industry or culture. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or not using generative AI. For example, the monitoring unit can optimize the monitoring algorithm by having generative AI take into account risks specific to a particular industry or culture. The monitoring unit can optimize the monitoring algorithm by having generative AI take into account risks specific to a particular industry or culture.

[0101] The monitoring unit can learn new risk patterns by referring to past monitoring history. For example, the monitoring unit learns new risk patterns by referring to past monitoring history. For example, the monitoring unit learns new risk patterns based on past monitoring history and improves monitoring accuracy. For example, the monitoring unit analyzes past monitoring history to identify frequently occurring risk patterns and improve the monitoring algorithm. For example, the monitoring unit predicts new risk patterns by referring to past monitoring history and reflects them in the monitoring algorithm. This makes it possible for the monitoring unit to learn new risk patterns by referring to past monitoring history. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit may have generative AI analyze past monitoring history and learn new risk patterns. The monitoring unit may have generative AI analyze past monitoring history and learn new risk patterns.

[0102] The monitoring unit can estimate the user's emotions and determine monitoring priorities based on the estimated emotions. For example, if the user is angry, the monitoring unit sets a high monitoring priority and performs a quick check. If the user is relaxed, the monitoring unit sets a normal monitoring priority and performs a standard check. If the user is anxious, the monitoring unit sets a medium monitoring priority and performs a moderate check. This allows the monitoring unit to determine monitoring priorities based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the monitoring unit may be performed using a generative AI, for example, or without a generative AI. For example, the monitoring unit may use a generative AI to analyze the user's emotions and determine monitoring priorities. The monitoring unit uses a generation AI to analyze user emotions and determine monitoring priorities.

[0103] The monitoring unit can prioritize monitoring region-specific risks by taking into account the user's geographical location information. For example, the monitoring unit prioritizes monitoring region-specific risks by taking into account the user's geographical location information. For example, if the user is in a specific region, the monitoring unit prioritizes monitoring region-specific risks for that region. For example, the monitoring unit optimizes an algorithm for monitoring region-specific risks based on the user's geographical location information. For example, the monitoring unit quickly monitors region-specific risks by taking into account the user's geographical location information. This makes it possible for the monitoring unit to prioritize monitoring region-specific risks by taking into account the user's geographical location information. Some or all of the above processing in the monitoring unit may be performed using, for example, a generative AI, or without a generative AI. For example, the monitoring unit may use a generative AI to analyze the user's geographical location information and monitor region-specific risks. The monitoring unit may use a generative AI to analyze the user's geographical location information and monitor region-specific risks.

[0104] The monitoring unit can analyze users' social media activity and monitor related risks. For example, the monitoring unit analyzes users' social media activity and monitors related risks. For example, the monitoring unit analyzes users' social media activity and monitors posts containing specific risks. For example, the monitoring unit optimizes algorithms for monitoring related risks based on users' social media activity. For example, the monitoring unit refers to users' social media activity and quickly monitors related risks. This makes it possible for the monitoring unit to analyze users' social media activity and monitor related risks. Some or all of the above processing in the monitoring unit may be performed using, for example, generative AI, or without generative AI. For example, the monitoring unit may have generative AI analyze users' social media activity and monitor related risks. The monitoring unit may have generative AI analyze users' social media activity and monitor related risks.

[0105] The evaluation criteria update unit can estimate the user's emotions and adjust the frequency of evaluation criterion updates based on the estimated user emotions. For example, if the user is angry, the evaluation criteria update unit sets the frequency of evaluation criterion updates to a high level and updates frequently. If the user is relaxed, the evaluation criteria update unit sets the frequency of evaluation criterion updates to a normal level and performs standard updates. If the user is anxious, the evaluation criteria update unit sets the frequency of evaluation criterion updates to a moderate level and performs moderate updates. This makes it possible for the evaluation criteria update unit to adjust the frequency of evaluation criterion updates based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the evaluation criteria update unit may be performed using a generative AI, for example, or without a generative AI. For example, the evaluation criteria update unit uses a generating AI to analyze the user's emotions and adjust the frequency of evaluation criteria updates.

[0106] The evaluation criteria update unit can optimize the update algorithm by taking into account criteria specific to a particular industry or culture. For example, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria specific to a particular industry or culture. For example, in the medical industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding patient privacy protection. For example, in the financial industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding customer data protection. For example, in the education industry, the evaluation criteria update unit optimizes the update algorithm by taking into account criteria regarding student data protection. This makes it possible for the evaluation criteria update unit to optimize the update algorithm by taking into account criteria specific to a particular industry or culture. Some or all of the above processing in the evaluation criteria update unit may be performed using, for example, generative AI, or without generative AI. For example, the evaluation criteria update unit may use generative AI to optimize the update algorithm by taking into account criteria specific to a particular industry or culture. The evaluation criteria update unit can use generative AI to optimize the update algorithm by taking into account criteria specific to a particular industry or culture.

[0107] The evaluation criteria update unit can estimate the user's emotions and determine the priority of evaluation criteria based on the estimated user emotions. For example, if the user is angry, the evaluation criteria update unit sets the priority of the evaluation criteria high and updates quickly. If the user is relaxed, the evaluation criteria update unit sets the priority of the evaluation criteria to a normal level and performs a standard update. If the user is anxious, the evaluation criteria update unit sets the priority of the evaluation criteria to a medium level and performs a moderate update. This makes it possible for the evaluation criteria update unit to determine the priority of evaluation criteria based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the evaluation criteria update unit may be performed using a generative AI, for example, or without a generative AI. For example, the evaluation criteria update unit uses a generating AI to analyze the user's emotions and determine the priority of the evaluation criteria.

[0108] The evaluation criteria update unit can prioritize updating region-specific criteria by taking into account the user's geographical location information. For example, the evaluation criteria update unit prioritizes updating region-specific criteria by taking into account the user's geographical location information. For example, if the user is in a specific region, the evaluation criteria update unit prioritizes updating the region-specific criteria for that region. For example, the evaluation criteria update unit optimizes the algorithm for updating region-specific criteria based on the user's geographical location information. For example, the evaluation criteria update unit quickly updates region-specific criteria by taking into account the user's geographical location information. This makes it possible for the evaluation criteria update unit to prioritize updating region-specific criteria by taking into account the user's geographical location information. Some or all of the above processing in the evaluation criteria update unit may be performed using, for example, a generative AI, or without a generative AI. For example, the evaluation criteria update unit may have a generative AI analyze the user's geographical location information and update region-specific criteria. The evaluation criteria update unit can have a generative AI analyze the user's geographical location information and update region-specific criteria.

[0109] The automated response analysis unit can estimate the user's emotions and adjust the accuracy of the response analysis based on the estimated emotions. For example, if the user is angry, the automated response analysis unit will increase the accuracy of the response analysis and perform a more rigorous analysis. If the user is relaxed, the automated response analysis unit will set the accuracy of the response analysis to a normal level and perform a standard analysis. If the user is anxious, the automated response analysis unit will set the accuracy of the response analysis to a moderate level and perform a moderate analysis. This makes it possible for the automated response analysis unit to adjust the accuracy of the response analysis based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above-described processing in the automated response content analysis unit may be performed using, for example, a generating AI, or without using a generating AI. For example, the automated response content analysis unit may use a generating AI to analyze the user's emotions and adjust the accuracy of the response content analysis.

[0110] The automated response analysis unit can optimize its analysis algorithm by considering response content specific to a particular industry or culture. For example, the automated response analysis unit optimizes its analysis algorithm by considering response content specific to a particular industry or culture. For example, in the medical industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting patient privacy. For example, in the financial industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting customer data. For example, in the education industry, the automated response analysis unit optimizes its analysis algorithm by considering response content related to protecting student data. This makes it possible for the automated response analysis unit to optimize its analysis algorithm by considering response content specific to a particular industry or culture. Some or all of the above processing in the automated response analysis unit may be performed using, for example, generative AI, or without generative AI. For example, the automated response analysis unit may use generative AI to optimize its analysis algorithm by considering response content specific to a particular industry or culture. The automated response analysis unit may use generative AI to optimize its analysis algorithm by considering response content specific to a particular industry or culture.

[0111] The automated response analysis unit can estimate the user's emotions and determine the priority of response analysis based on the estimated emotions. For example, if the user is angry, the automated response analysis unit sets a high priority for response analysis and performs a rapid analysis. If the user is relaxed, the automated response analysis unit sets a normal level for response analysis and performs a standard analysis. If the user is anxious, the automated response analysis unit sets a medium level for response analysis and performs a moderate analysis. This allows the automated response analysis unit to determine the priority of response analysis based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, with an emotion engine or generative AI. Generative AI includes, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above-described processing in the automated response content analysis unit may be performed using, for example, a generating AI, or without using a generating AI. For example, the automated response content analysis unit may use a generating AI to analyze the user's emotions and determine the priority order for analyzing the response content. The automated response content analysis unit may use a generating AI to analyze the user's emotions and determine the priority order for analyzing the response content.

[0112] The automated response analysis unit can prioritize the analysis of region-specific response content by considering the user's geographical location information. For example, the automated response analysis unit prioritizes the analysis of region-specific response content by considering the user's geographical location information. For example, if the user is in a specific region, the automated response analysis unit prioritizes the analysis of region-specific response content for that region. For example, the automated response analysis unit optimizes the algorithm for analyzing region-specific response content based on the user's geographical location information. For example, the automated response analysis unit quickly analyzes region-specific response content by considering the user's geographical location information. This makes it possible for the automated response analysis unit to prioritize the analysis of region-specific response content by considering the user's geographical location information. Some or all of the above processing in the automated response analysis unit may be performed using, for example, a generative AI, or without a generative AI. For example, the automated response analysis unit can use a generative AI to analyze the user's geographical location information and analyze region-specific response content.

[0113] The pattern recognition unit can estimate the user's emotions and adjust the accuracy of pattern recognition based on the estimated emotions. For example, the pattern recognition unit estimates the user's emotions and adjusts the accuracy of pattern recognition based on the estimated emotions. For example, if the user is angry, the pattern recognition unit increases the accuracy of pattern recognition to perform a more precise recognition. For example, if the user is relaxed, the pattern recognition unit sets the accuracy of pattern recognition to a normal level and performs a standard recognition. For example, if the user is anxious, the pattern recognition unit sets the accuracy of pattern recognition to a moderate level and performs a moderate recognition. This makes it possible for the pattern recognition unit to adjust the accuracy of pattern recognition based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or a generative AI. The generative AI is a text generation AI (e.g., LLM) or a multimodal generation AI, but is not limited to such examples. Some or all of the above processing in the pattern recognition unit may be performed using a generative AI, for example, or without a generative AI. For example, the pattern recognition unit uses a generative AI to analyze the user's emotions and adjust the accuracy of pattern recognition.

[0114] The pattern recognition unit can optimize its recognition algorithm by considering patterns specific to a particular industry or culture. For example, the pattern recognition unit can optimize its recognition algorithm by considering patterns specific to a particular industry or culture. For example, in the medical industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to patient privacy protection. For example, in the financial industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to customer data protection. For example, in the education industry, the pattern recognition unit can optimize its recognition algorithm by considering patterns related to student data protection. This makes it possible for the pattern recognition unit to optimize its recognition algorithm by considering patterns specific to a particular industry or culture. Some or all of the above processing in the pattern recognition unit may be performed using, for example, generative AI, or without generative AI. For example, the pattern recognition unit can optimize its recognition algorithm by having the generative AI consider patterns specific to a particular industry or culture.

[0115] The pattern recognition unit can estimate the user's emotions and determine the priority of pattern recognition based on the estimated user emotions. For example, the pattern recognition unit estimates the user's emotions and determines the priority of pattern recognition based on the estimated user emotions. For example, if the user is angry, the pattern recognition unit sets the priority of pattern recognition to high and performs recognition quickly. For example, if the user is relaxed, the pattern recognition unit sets the priority of pattern recognition to a normal level and performs standard recognition. For example, if the user is anxious, the pattern recognition unit sets the priority of pattern recognition to a medium level and performs appropriate recognition. This makes it possible for the pattern recognition unit to determine the priority of pattern recognition based on the user's emotions. Emotion estimation is achieved using an emotion estimation function, for example, using an emotion engine or generative AI. Generative AI is, but is not limited to, text generation AI (e.g., LLM) or multimodal generation AI. Some or all of the above processing in the pattern recognition unit may be performed using a generative AI, for example, or without a generative AI. For example, the pattern recognition unit uses a generative AI to analyze the user's emotions and determine the priority of pattern recognition.

[0116] The pattern recognition unit can prioritize the recognition of region-specific patterns by taking into account the user's geographical location information. For example, the pattern recognition unit prioritizes the recognition of region-specific patterns by taking into account the user's geographical location information. For example, if the user is in a specific region, the pattern recognition unit prioritizes the recognition of region-specific patterns for that region. For example, the pattern recognition unit optimizes the algorithm for recognizing region-specific patterns based on the user's geographical location information. For example, the pattern recognition unit quickly recognizes region-specific patterns by taking into account the user's geographical location information. This makes it possible for the pattern recognition unit to prioritize the recognition of region-specific patterns by taking into account the user's geographical location information. Some or all of the above processing in the pattern recognition unit may be performed using, for example, a generative AI, or without a generative AI. For example, the pattern recognition unit may use a generative AI to analyze the user's geographical location information and recognize region-specific patterns. The pattern recognition unit can use a generative AI to analyze the user's geographical location information and recognize region-specific patterns.

[0117] The system according to the embodiment is not limited to the example described above, and various modifications are possible, for example, as follows.

[0118] The AI ​​agent evaluation system can estimate the user's emotions and adjust how evaluation results are presented based on those estimated emotions. For example, if the user is angry, the evaluation results can be presented quickly and concisely to alleviate the user's dissatisfaction. If the user is relaxed, detailed evaluation results can be provided to allow the user to understand more deeply. If the user is anxious, the evaluation results can be presented in a reassuring way to ease the user's anxiety. This allows the AI ​​agent evaluation system to present evaluation results flexibly according to the user's emotions.

[0119] The AI ​​agent evaluation system can take into account the user's geographical location and perform evaluations based on region-specific laws, regulations, and cultural backgrounds. For example, in a particular region where privacy protection regulations are strict, the evaluation will be based on those local laws. In another region, bias detection will be performed based on specific cultural backgrounds. Furthermore, evaluation criteria that reflect the social values ​​of each region can be set. This allows the AI ​​agent evaluation system to perform evaluations that are tailored to region-specific requirements.

[0120] The AI ​​agent evaluation system can analyze a user's social media activity and provide evaluations based on the user's interests and values. For example, it can focus on detecting biases related to topics that users frequently mention on social media. It can estimate specific values ​​and ethical perspectives from the user's posts and set evaluation criteria based on them. Furthermore, it can customize how evaluation results are presented using data obtained from the user's social media activity. This allows the AI ​​agent evaluation system to provide evaluations tailored to each user's individual interests and values.

[0121] The AI ​​agent evaluation system can estimate the user's emotions and adjust the transparency of the evaluation process based on those emotions. For example, if the user is angry, the system will clearly explain the details of the evaluation process to ensure the user understands and accepts it. If the user is relaxed, the system will only present an overview of the evaluation process with a concise explanation. If the user is anxious, the system will carefully explain each step of the evaluation process to enhance the user's sense of security. In this way, the AI ​​agent evaluation system can provide transparency in the evaluation process that is tailored to the user's emotions.

[0122] The AI ​​agent evaluation system can estimate the user's emotions and adjust how evaluation results are communicated based on those emotions. For example, if the user is angry, the evaluation results are communicated quickly to alleviate their dissatisfaction. If the user is relaxed, the evaluation results are communicated in detail to allow for deeper understanding. If the user is anxious, the evaluation results are communicated in a reassuring way to ease their anxiety. This enables the AI ​​agent evaluation system to provide flexible evaluation result notifications that respond to the user's emotions.

[0123] The AI ​​agent evaluation system can estimate the user's emotions and adjust the strictness of its evaluation criteria based on those emotions. For example, if the user is angry, the evaluation criteria can be set stricter for a more rigorous evaluation. If the user is relaxed, the evaluation criteria can be set to a normal level for a standard evaluation. If the user is anxious, the evaluation criteria can be set to a moderate level for a moderate evaluation. This allows the AI ​​agent evaluation system to set flexible evaluation criteria that respond to the user's emotions.

[0124] The AI ​​agent evaluation system can estimate the user's emotions and adjust the feedback method of the evaluation results based on those estimated emotions. For example, if the user is angry, the evaluation results feedback will be quick and concise to alleviate the user's dissatisfaction. If the user is relaxed, detailed feedback will be provided to allow the user to understand more deeply. If the user is anxious, feedback will be provided in a reassuring way to ease the user's anxiety. This allows the AI ​​agent evaluation system to provide flexible evaluation results feedback that responds to the user's emotions.

[0125] The AI ​​agent evaluation system can estimate the user's emotions and adjust the display format of the evaluation results based on those emotions. For example, if the user is angry, the evaluation results will be displayed concisely to reduce the user's dissatisfaction. If the user is relaxed, detailed evaluation results will be displayed to allow the user to understand more deeply. If the user is anxious, the evaluation results will be displayed in a reassuring way to alleviate the user's anxiety. This allows the AI ​​agent evaluation system to display evaluation results flexibly according to the user's emotions.

[0126] The AI ​​agent evaluation system can estimate the user's emotions and adjust the pace of the evaluation process based on those emotions. For example, if the user is angry, the evaluation process will proceed quickly to alleviate their frustration. If the user is relaxed, the evaluation process will proceed at a normal pace. If the user is anxious, the evaluation process will proceed slowly to increase their sense of security. This allows the AI ​​agent evaluation system to conduct a flexible evaluation process that responds to the user's emotions.

[0127] The AI ​​agent evaluation system can estimate the user's emotions and adjust the importance of the evaluation result based on those emotions. For example, if the user is angry, the importance of the evaluation result is set high, and a quick response is initiated. If the user is relaxed, the importance of the evaluation result is set to a normal level, and a standard response is provided. If the user is anxious, the importance of the evaluation result is set to a medium level, and an appropriate response is provided. This allows the AI ​​agent evaluation system to flexibly adjust the importance of evaluation results according to the user's emotions.

[0128] The following briefly describes the processing flow for example form 2.

[0129] Step 1: The bias detection unit automatically detects potential biases contained in the dataset and agent responses. For example, it can detect gender bias, racial bias, algorithmic bias, etc. It can also detect unconscious biases and biases in the dataset. Step 2: The compliance check unit checks compliance based on biases detected by the bias detection unit. For example, it evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. It addresses industry-specific legal regulations for financial institutions, healthcare institutions, and public institutions, and evaluates the response based on legal regulations such as GDPR and CCPA. It also evaluates the response based on corporate policies and ethical standards. Step 3: The Improvement Proposal Department makes improvement proposals based on the results obtained by the Compliance Check Department. For example, it makes correction proposals based on the results of the Bias Detection Department and the Compliance Check Department, proposing modifications to the algorithm or dataset. Correction proposals are made based on the type of proposal and evaluation criteria. Step 4: The Monitoring Department monitors the agents in operation based on the improvement proposals submitted by the Improvement Proposal Department. For example, it periodically evaluates the agents in operation and immediately notifies them if any new risks or biases are detected. It evaluates the agents based on the monitoring frequency and evaluation criteria, and automatically analyzes the responses of the agents in operation to check for biases or risks.

[0130] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0131] Data generation model 58 is a form of so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> Examples of generative AI include text generation AI, image generation AI, and multimodal generation AI. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats from audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVMs), k-means clustering, convolutional neural networks (CNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each of the above parts is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example.Furthermore, processing performed by AI, including generative AI, may be replaced with rule-based processing, and rule-based processing may be replaced with processing performed by AI, including generative AI.

[0132] Furthermore, the processing performed by the data processing system 10 described above is carried out by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it may also be carried out by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart device 14 or an external device, and the smart device 14 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0133] Each of the multiple elements described above, including the bias detection unit, compliance check unit, improvement suggestion unit, monitoring unit, evaluation criteria update unit, automatic response content analysis unit, and pattern recognition unit, is implemented in at least one of the smart device 14 and the data processing device 12. For example, the bias detection unit is implemented by the control unit 46A of the smart device 14 and automatically detects potential biases contained in the dataset and agent responses. The compliance check unit is implemented by the specific processing unit 290 of the data processing device 12 and evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. The improvement suggestion unit is implemented by the specific processing unit 290 of the data processing device 12 and makes correction suggestions based on the results of the bias detection unit and the compliance check unit. The monitoring unit is implemented by the control unit 46A of the smart device 14 and periodically evaluates the agent in operation and immediately notifies when new risks or biases are detected. The evaluation criteria update unit is implemented by the specific processing unit 290 of the data processing device 12 and uses generated AI to perform evaluations and improvements in accordance with the latest legal regulations and social values. The automated response analysis unit is implemented by the control unit 46A of the smart device 14, which uses generating AI to analyze whether the agent's responses contain bias or risks. The pattern recognition unit is implemented by the identification processing unit 290 of the data processing device 12, which analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0134] [Second Embodiment] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0135] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0136] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0137] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0138] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0139] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0140] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0141] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing by the processor 28. The storage 32 stores the specific processing program 56.

[0142] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0143] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0144] In the smart glasses 214, specific processing is performed by the processor 46. The storage 50 stores a specific processing program 60. The processor 46 reads the specific processing program 60 from the storage 50 and executes the read specific processing program 60 on the RAM 48. The specific processing is realized by the processor 46 acting as a control unit 46A according to the specific processing program 60 executed on the RAM 48. The smart glasses 214 also have a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0145] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0146] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0147] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0148] The data processing system 210 according to the second embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 210 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the smart glasses 214 or an external device, and the smart glasses 214 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0149] Each of the multiple elements described above, including the bias detection unit, compliance check unit, improvement suggestion unit, monitoring unit, evaluation criteria update unit, automatic response content analysis unit, and pattern recognition unit, is implemented in at least one of the smart glasses 214 and the data processing unit 12. For example, the bias detection unit is implemented by the control unit 46A of the smart glasses 214 and automatically detects potential biases contained in the dataset and agent responses. The compliance check unit is implemented by the identification processing unit 290 of the data processing unit 12 and evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. The improvement suggestion unit is implemented by the identification processing unit 290 of the data processing unit 12 and makes correction suggestions based on the results of the bias detection unit and the compliance check unit. The monitoring unit is implemented by the control unit 46A of the smart glasses 214 and periodically evaluates the agent in operation and immediately notifies when new risks or biases are detected. The evaluation criteria update unit is implemented by the identification processing unit 290 of the data processing unit 12 and uses generated AI to perform evaluations and improvements in accordance with the latest legal regulations and social values. The automated response analysis unit is implemented by the control unit 46A of the smart glasses 214, which uses generating AI to analyze whether the agent's responses contain bias or risks. The pattern recognition unit is implemented by the identification processing unit 290 of the data processing device 12, which analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0150] [Third Embodiment] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0151] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0152] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0153] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0154] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0155] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0156] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0157] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0158] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0159] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0160] In the headset terminal 314, specific processing is performed by the processor 46. The storage 50 stores a specific program 60. The processor 46 reads the specific program 60 from the storage 50 and executes the read specific program 60 on the RAM 48. The specific processing is realized by the processor 46 acting as a control unit 46A according to the specific program 60 executed on the RAM 48. The headset terminal 314 also has a data generation model 58 and an emotion identification model 59, similar to the data generation model and emotion identification model 59, and can perform processing similar to that of the specific processing unit 290 using these models.

[0161] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0162] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0163] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0164] The data processing system 310 according to the third embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 310 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the headset terminal 314, but may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the headset terminal 314. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the headset terminal 314 or an external device, and the headset terminal 314 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0165] Each of the multiple elements described above, including the bias detection unit, compliance check unit, improvement suggestion unit, monitoring unit, evaluation criteria update unit, automatic response content analysis unit, and pattern recognition unit, is implemented in at least one of the headset terminal 314 and the data processing unit 12. For example, the bias detection unit is implemented by the control unit 46A of the headset terminal 314 and automatically detects potential biases contained in the dataset and agent responses. The compliance check unit is implemented by the specific processing unit 290 of the data processing unit 12 and evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. The improvement suggestion unit is implemented by the specific processing unit 290 of the data processing unit 12 and makes correction suggestions based on the results of the bias detection unit and the compliance check unit. The monitoring unit is implemented by the control unit 46A of the headset terminal 314 and periodically evaluates the agent in operation and immediately notifies when new risks or biases are detected. The evaluation criteria update unit is implemented by the specific processing unit 290 of the data processing unit 12 and uses generated AI to perform evaluations and improvements in accordance with the latest legal regulations and social values. The automated response analysis unit is implemented by the control unit 46A of the headset terminal 314, which uses generating AI to analyze whether the agent's responses contain bias or risks. The pattern recognition unit is implemented by the identification processing unit 290 of the data processing device 12, which analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0166] [Fourth Embodiment] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[0167] As shown in Figure 7, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0168] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN and / or LAN.

[0169] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[0170] The microphone 238 receives voice signals from the user and accepts instructions from the user. The microphone 238 captures the voice signals from the user, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0171] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS image sensor or CCD image sensor, which captures images of the area around the user (for example, an imaging range defined by a field of view equivalent to the field of vision of a typical healthy person).

[0172] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0173] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. The robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0174] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0175] The processor 28 reads a specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 acting as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0176] Storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290. The identification processing unit 290 can estimate the user's emotions using the emotion identification model 59 and perform identification processing using the user's emotions. The emotion estimation function (emotion identification function) using the emotion identification model 59 performs various estimations and predictions regarding the user's emotions, including but not limited to these examples. Furthermore, emotion estimation and prediction also include, for example, emotion analysis.

[0177] In robot 414, specific processing is performed by processor 46. A specific program 60 is stored in storage 50. Processor 46 reads the specific program 60 from storage 50 and executes it on RAM 48. The specific processing is achieved by processor 46 acting as a control unit 46A according to the specific program 60 executed on RAM 48. Robot 414 also has data generation model 58 and emotion identification model 59, similar to those of the robot, and can perform processing similar to that of the specific processing unit 290 using these models.

[0178] Furthermore, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device may have the data generation model 58. In this case, the data processing device 12 obtains processing results (such as prediction results) using the data generation model 58 by communicating with the server device that has the data generation model 58. Also, the data processing device 12 may be a server device or a terminal device owned by the user (for example, a mobile phone, robot, home appliance, etc.).

[0179] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0180] The data generation model 58 is a so-called generative AI. An example of a data generation model 58 is a generative AI such as ChatGPT. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and inference data such as audio data representing speech, text data representing text, and image data representing images (e.g., still image data or video data). The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference result in one or more data formats such as audio data, text data, and image data. The data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the specific processing described above using the data generation model 58. The data generation model 58 may be a fine-tuned model that outputs inference results from prompts that do not contain instructions, in which case the data generation model 58 can output inference results from prompts that do not contain instructions. In the data processing device 12, etc., there are multiple types of data generation models 58, and the data generation model 58 includes AI other than generative AI. AI other than generative AI includes, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or naive Bayes, and can perform various processes, but is not limited to these examples. Also, the AI ​​may be an AI agent. Furthermore, when the processing of each part described above is performed by the AI, the processing may be performed by the AI ​​in part or in whole, but is not limited to this example. Also, processing performed by an AI including a generative AI may be replaced by rule-based processing, and rule-based processing may be replaced by processing performed by an AI including a generative AI.

[0181] The data processing system 410 according to the fourth embodiment performs the same processing as the data processing system 10 according to the first embodiment. The processing by the data processing system 410 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it may also be performed by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. In addition, the specific processing unit 290 of the data processing device 12 acquires or collects information necessary for processing from the robot 414 or an external device, and the robot 414 acquires or collects information necessary for processing from the data processing device 12 or an external device.

[0182] Each of the multiple elements described above, including the bias detection unit, compliance check unit, improvement suggestion unit, monitoring unit, evaluation criteria update unit, response content automatic analysis unit, and pattern recognition unit, is implemented in at least one of the robot 414 and the data processing unit 12. For example, the bias detection unit is implemented by the control unit 46A of the robot 414 and automatically detects potential biases contained in the dataset and agent responses. The compliance check unit is implemented by the specific processing unit 290 of the data processing unit 12 and evaluates whether the response is appropriate based on legal regulations and privacy protection requirements. The improvement suggestion unit is implemented by the specific processing unit 290 of the data processing unit 12 and makes correction suggestions based on the results of the bias detection unit and the compliance check unit. The monitoring unit is implemented by the control unit 46A of the robot 414 and periodically evaluates the agent in operation and immediately notifies when new risks or biases are detected. The evaluation criteria update unit is implemented by the specific processing unit 290 of the data processing unit 12 and uses generated AI to perform evaluations and improvements in accordance with the latest legal regulations and social values. The automated response analysis unit is implemented by the control unit 46A of the robot 414, which uses generating AI to analyze whether the agent's responses contain bias or risks. The pattern recognition unit is implemented by the identification processing unit 290 of the data processing device 12, which analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The correspondence between each unit and the device or control unit is not limited to the example described above and can be modified in various ways.

[0183] Furthermore, the emotion identification model 59, acting as an emotion engine, may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[0184] Figure 9 shows the emotion map 400, in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[0185] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[0186] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[0187] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, and motorcycles, emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated based, for example, on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[0188] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[0189] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[0190] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing method for the specific process may be used, which includes computer 22 and multiple other computers.

[0191] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0192] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[0193] Furthermore, it is not necessary to store the entirety of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[0194] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[0195] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[0196] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[0197] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[0198] Furthermore, although the above-described examples were divided into four embodiments, some or all of these embodiments may be combined. Also, the smart device 14, smart glasses 214, headset terminal 314, and robot 414 are just examples, and they may be combined, or other devices may be used. Also, although the above-described examples were divided into two embodiments, Embodiment 1 and Embodiment 2, these may be combined.

[0199] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and other things that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[0200] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[0201] (Note 1) A bias detection unit that detects bias, A compliance check unit that checks compliance based on the bias detected by the bias detection unit, An improvement proposal unit that makes improvement proposals based on the results obtained by the compliance check unit, The system includes a monitoring unit that monitors agents in operation based on improvement proposals submitted by the aforementioned improvement proposal unit. A system characterized by the following features. (Note 2) It also includes a section for updating evaluation criteria, which uses generated AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The system described in Appendix 1, characterized by the features described herein. (Note 3) It also includes an automated response analysis unit that uses generated AI to analyze whether the agent's responses contain bias or risk. The system described in Appendix 1, characterized by the features described herein. (Note 4) It further includes a pattern recognition unit that analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The system described in Appendix 1, characterized by the features described herein. (Note 5) The bias detection unit is Automatically detects potential biases in datasets and agent responses. The system described in Appendix 1, characterized by the features described herein. (Note 6) The aforementioned compliance check unit, We evaluate whether the response is appropriate based on legal regulations and privacy protection requirements. The system described in Appendix 1, characterized by the features described herein. (Note 7) The aforementioned improvement proposal department, Based on the results from the bias detection unit and the compliance check unit, correction suggestions are made. The system described in Appendix 1, characterized by the features described herein. (Note 8) The monitoring unit, We regularly evaluate agents in operation and immediately notify users of any new risks or biases detected. The system described in Appendix 1, characterized by the features described herein. (Note 9) The bias detection unit is It estimates the user's emotions and adjusts the accuracy of bias detection based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 10) The bias detection unit is Optimize the detection algorithm to take into account biases specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 11) The bias detection unit is Learn new bias patterns by referring to past bias detection history. The system described in Appendix 1, characterized by the features described herein. (Note 12) The bias detection unit is The system estimates the user's emotions and determines the priority of bias detection based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 13) The bias detection unit is Prioritize detecting region-specific biases by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 14) The bias detection unit is Analyze users' social media activity and detect relevant biases. The system described in Appendix 1, characterized by the features described herein. (Note 15) The aforementioned compliance check unit, We estimate user sentiment and adjust compliance check criteria based on the estimated user sentiment. The system described in Appendix 1, characterized by the features described herein. (Note 16) The aforementioned compliance check unit, Optimize the check algorithm to take into account the legal regulations of specific industries and regions. The system described in Appendix 1, characterized by the features described herein. (Note 17) The aforementioned compliance check unit, Learn new violation patterns by referring to past violation history. The system described in Appendix 1, characterized by the features described herein. (Note 18) The aforementioned compliance check unit, The system estimates user sentiment and prioritizes compliance checks based on the estimated user sentiment. The system described in Appendix 1, characterized by the features described herein. (Note 19) The aforementioned compliance check unit, Prioritize checking region-specific laws and regulations by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 20) The aforementioned compliance check unit, Analyze users' social media activity and check for any relevant legal or regulatory violations. The system described in Appendix 1, characterized by the features described herein. (Note 21) The aforementioned improvement proposal department, The system estimates the user's emotions and adjusts the content of improvement suggestions based on those estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 22) The aforementioned improvement proposal department, Optimize the proposed algorithm by considering improvement suggestions specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 23) The aforementioned improvement proposal department, Learn new improvement ideas by referring to past improvement suggestion history. The system described in Appendix 1, characterized by the features described herein. (Note 24) The aforementioned improvement proposal department, It estimates user emotions and prioritizes improvement suggestions based on those estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 25) The aforementioned improvement proposal department, Prioritize suggesting region-specific improvement plans by taking into account the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 26) The aforementioned improvement proposal department, We analyze users' social media activity and propose relevant improvement suggestions. The system described in Appendix 1, characterized by the features described herein. (Note 27) The monitoring unit, It estimates the user's emotions and adjusts the monitoring frequency based on the estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 28) The monitoring unit, Optimize monitoring algorithms to take into account risks specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 29) The monitoring unit, Learn new risk patterns by referring to past monitoring history. The system described in Appendix 1, characterized by the features described herein. (Note 30) The monitoring unit, It estimates user sentiment and determines monitoring priorities based on the estimated user sentiment. The system described in Appendix 1, characterized by the features described herein. (Note 31) The monitoring unit, Prioritize monitoring of region-specific risks by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 32) The monitoring unit, Analyze users' social media activity and monitor associated risks. The system described in Appendix 1, characterized by the features described herein. (Note 33) The evaluation criteria update unit is: The system estimates user sentiment and adjusts the frequency of updating evaluation criteria based on the estimated user sentiment. The system described in Appendix 1, characterized by the features described herein. (Note 34) The evaluation criteria update unit is: Optimize the update algorithm by taking into account standards specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 35) The evaluation criteria update unit is: The system estimates the user's emotions and determines the priority of evaluation criteria based on those estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 36) The evaluation criteria update unit is: Prioritize updating region-specific criteria, taking into account the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 37) The automated analysis unit for the response content said above, It estimates the user's emotions and adjusts the accuracy of the response analysis based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 38) The automated analysis unit for the response content said above, Optimize the analysis algorithm by taking into account response content specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 39) The automated analysis unit for the response content said above, It estimates the user's emotions and determines the priority of response analysis based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 40) The automated analysis unit for the response content said above, Prioritize analyzing region-specific responses by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. (Note 41) The pattern recognition unit, It estimates the user's emotions and adjusts the accuracy of pattern recognition based on the estimated user emotions. The system described in Appendix 1, characterized by the features described herein. (Note 42) The pattern recognition unit, Optimize the recognition algorithm by taking into account patterns specific to particular industries and cultures. The system described in Appendix 1, characterized by the features described herein. (Note 43) The pattern recognition unit, The system estimates the user's emotions and determines the priority of pattern recognition based on the estimated emotions. The system described in Appendix 1, characterized by the features described herein. (Note 44) The pattern recognition unit, Prioritize the recognition of region-specific patterns by considering the user's geographical location. The system described in Appendix 1, characterized by the features described herein. [Explanation of Symbols]

[0202] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots

Claims

1. A bias detection unit that detects bias, A compliance check unit that checks compliance based on the bias detected by the bias detection unit, An improvement proposal unit that makes improvement proposals based on the results obtained by the compliance check unit, The system includes a monitoring unit that monitors agents in operation based on improvement proposals submitted by the aforementioned improvement proposal unit. A system characterized by the following features.

2. It also includes a section for updating evaluation criteria, which uses generative AI to perform evaluations and improvements in accordance with the latest laws, regulations, and social values. The system according to feature 1.

3. It further includes an automated response analysis unit that uses generated AI to analyze whether the agent's responses contain bias or risk. The system according to feature 1.

4. It further includes a pattern recognition unit that analyzes large amounts of data to identify frequently occurring inappropriate response patterns and ethical risks. The system according to feature 1.

5. The bias detection unit is Automatically detects potential biases in datasets and agent responses. The system according to feature 1.

6. The aforementioned compliance check unit, We evaluate whether the response is appropriate based on legal regulations and privacy protection requirements. The system according to feature 1.

7. The aforementioned improvement proposal department, Based on the results from the bias detection unit and the compliance check unit, correction suggestions are made. The system according to feature 1.

8. The monitoring unit, We regularly evaluate agents in operation and immediately notify users of any new risks or biases detected. The system according to feature 1.

9. The bias detection unit is It estimates the user's emotions and adjusts the accuracy of bias detection based on the estimated user emotions. The system according to feature 1.

Images