System

Abstract

To provide a system for supporting an improvement in response of a user.SOLUTION: A system contains: means of receiving a selection of a user from a plurality of scenarios; means of setting a specific role on the basis of the scenario selected by the user and generating a prompt text for instructing to make a response to the role; means of receiving an input message of the user; means of generating a reply to the input message of the user by using the generated prompt text and a generative AI; means of outputting the generated reply to the user; means of repeating the reception of an input message, the generation of a reply, and the output of the reply; and means of evaluating a response of the user by using the generative AI and the prompt text for instructing to check a tendency of harassment or the presence / absence of a bias for the response of the user on the basis of the history of the input message and the reply.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the 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] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Issues of harassment and bias in employee communication have a significant impact on a company's productivity and employee satisfaction. However, identifying these issues and responding appropriately is not easy. It is particularly challenging to respond appropriately in specific situations, such as when a subordinate comes out as LGBT, announces pregnancy, or is worried about their work style. Creating return-to-work programs and providing materials for decision-making for employees who have taken leave due to mental health issues is also a challenge. [Means for solving the problem]

[0005] To solve this problem, we provide a system in which specific roles are assigned to generative AI, and users respond to those roles via chat. The results of these responses are analyzed to check for harassment tendencies and bias. The analysis results are fed back to help users improve their responses. Furthermore, this system can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. [Brief explanation of the drawings]

[0006] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 2 is a sequence diagram showing a flow of processing in the data processing system according to the first embodiment of the first form example. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Embodiment 1. [Figure 13] FIG. 10 is a sequence diagram showing a processing flow of a data processing system in a second embodiment of the second form example. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Embodiment Example 2. [Figure 15] FIG. 10 is a sequence diagram showing the flow of processing in a data processing system according to a third embodiment of the third embodiment. [Figure 16] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Embodiment 3. [Figure 17] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the first embodiment of the first form example when an emotion engine is combined. [Figure 18] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1 of Form Example 1 when an emotion engine is combined. [Figure 19] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the second embodiment of the second form example when an emotion engine is combined. [Figure 20] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 of Form Example 2 when an emotion engine is combined. [Figure 21] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in the third embodiment of the third form example when an emotion engine is combined. [Figure 22] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 3 of Form Example 3 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0007] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0008] First, the terms used in the following description will be explained.

[0009] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, the processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), or a TPU (TENSOR PROCESSING UNIT (registered trademark)).

[0010] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0011] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0012] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple 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), Bluetooth (registered trademark), etc.

[0013] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0014] [First embodiment]

[0015] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0016] 1, a 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.

[0017] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also 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).

[0018] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0019] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0020] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0021] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0022] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0023] 2, in the data processing device 12, a specific process 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" according to the technology of the present 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 process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0024] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0025] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0026] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0027] "Example 1"

[0028] One embodiment of the present invention involves a means for assigning specific roles to generative AI. Specifically, roles such as "a subordinate coming out as LGBT," "a subordinate announcing a pregnancy," or "a subordinate struggling with work style" can be assigned to the AI. The user, as the supervisor, receives the consultation and then interacts with the AI ​​via chat.

[0029] "Example 2"

[0030] Next, there is a method to analyze user responses to check for harassment tendencies and bias. This analysis is done using natural language processing and machine learning techniques. Specifically, it determines whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc.

[0031] "Example 3"

[0032] Furthermore, there is a means to provide feedback on the analysis results. This feedback provides specific suggestions and advice to help users improve their own behavior. For example, if bias is found in the way employees treat subordinates who come out as LGBTQ+, the system will present the user with specific details of that bias and suggestions for improvement.

[0033] "Example 4"

[0034] The system of this invention can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. Specifically, to simulate how to respond when an employee returns from leave, the AI ​​is assigned the role of "an employee returning from leave," and the user responds via chat. The results are analyzed, and appropriate responses and return-to-work support programs are proposed.

[0035] The processing flow of each embodiment will be described below.

[0036] "Example 1"

[0037] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "a subordinate who comes out as LGBT."

[0038] Step 2: As a manager, the user responds to the subordinate's coming out via chat.

[0039] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0040] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0041] "Example 2"

[0042] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "subordinate who reports pregnancy."

[0043] Step 2: As a supervisor, the user responds to the subordinate's pregnancy report via chat.

[0044] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0045] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0046] "Example 3"

[0047] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of a "subordinate who is worried about their work style."

[0048] Step 2: As a supervisor, the user responds to the subordinate's concerns via chat.

[0049] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0050] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0051] "Example 4"

[0052] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of an "employee returning from leave."

[0053] Step 2: As a manager, the user responds to the employee's return via chat.

[0054] Step 3: AI analyzes the user's response and suggests appropriate responses and programs to support their return to work.

[0055] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0056] Example 1

[0057] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0058] In conventional systems using generative AI, when users respond to specific roles in chat, the system does not adequately generate appropriate prompts or display responses, resulting in a poor user experience. Furthermore, the system does not analyze the response results or check for harassment tendencies or bias, resulting in insufficient feedback. This limits users' opportunities to learn appropriate responses.

[0059] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0060] In this invention, the server includes: means for accepting a user's selection from multiple scenarios; means for setting a specific role based on the scenario selected by the user and generating a prompt sentence instructing how to respond to the role; means for accepting a user's input message; means for generating a response to the user's input message using the generated prompt sentence and a generative AI; means for outputting the generated response to the user; means for repeatedly accepting the input message, generating the response, and outputting the application; a prompt sentence instructing the user to check for harassment tendencies or bias in the user's responses based on a history of the input messages and responses; and means for evaluating the user's responses using the generative AI. This allows the user to learn how to respond to specific roles in real time and receive appropriate feedback. The multiple scenarios include a user interacting with a subordinate who comes out as a sexual minority, a user interacting with a subordinate who announces pregnancy, and a user interacting with a subordinate who is concerned about their work style.

[0061] "Generative AI" is an artificial intelligence system that generates natural language responses based on user input.

[0062] A "specific role" is a specific character or situation that the user sets based on a scenario and that the generative AI will imitate.

[0063] A "user" is a person who uses the system to chat with generative artificial intelligence.

[0064] A "prompt" is text input to a generative artificial intelligence (AI) that instructs it to generate a response based on a specific role.

[0065] A "response" is a text message that the generative artificial intelligence generates based on the prompt sentence.

[0066] "Chat" is an interactive form of communication between a user and a generative artificial intelligence via text messages.

[0067] A "scenario" is a setting containing specific situations and roles selected by the user, which serves as the basis for the generative AI to generate responses.

[0068] "Harassment tendencies" refers to a pattern of inappropriate speech and behavior toward others as a result of a user's responses.

[0069] "Presence or absence of bias" indicates whether or not a particular prejudice or preconception is included in the user's response results.

[0070] "Feedback" is information for evaluation and improvement provided based on the user's response results.

[0071] The present invention is a system that uses generative artificial intelligence to set specific roles and allows users to respond to those roles through chat. A specific embodiment of this system will be described below.

[0072] First, the user selects one of several scenarios through the device interface. For example, scenarios are displayed in which the user talks to a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style.

[0073] The server generates a prompt based on the scenario selected by the user. Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4 (registered trademark)) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to that role. For example, it generates a prompt that reads, "The user is your boss. You are a subordinate who is consulting with him about coming out as a sexual minority. Please generate a response to the user's input message so as to reproduce this dialogue."

[0074] Next, the server inputs the generated prompt and the user's input message into the generative AI to generate a response. For example, the generated response might be, "Boss, could you please give me a moment? Actually, there's something I've been wanting to say for a while."

[0075] The server generates a response and sends it to the terminal, which displays it to the user. The user sees the response and types in a new message, such as "Sure, what's wrong?"

[0076] The server then inputs the user's input message back into the generative AI to generate the next response, such as, "Actually, I want to come out as a sexual minority, but I don't know how to say it..."

[0077] This process is repeated, and the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and respond to a subordinate's consultation.

[0078] Furthermore, the server analyzes the user's response based on the history of input messages and responses, and evaluates whether there is a tendency for harassment or bias. This allows the server to evaluate how the user responded and provide appropriate feedback.

[0079] As a specific example, the prompt sentence when the user selects the scenario "a subordinate who comes out as a sexual minority" is shown below.

[0080] Example prompt sentence:

[0081] "The user is your manager. You are a subordinate who is asking for advice on coming out as a LGBTQ person. Please generate responses to the user's input messages to recreate this dialogue. You are nervous and wondering how to proceed. The subordinate's initial message is as follows: "Boss, can I have a moment? There's actually something I've been wanting to tell you for a while."

[0082] In this way, users can learn how to respond to specific roles in real time and receive appropriate feedback.

[0083] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0084] Step 1:

[0085] The user selects a scenario.

[0086] The user selects one of several scenarios through the terminal interface. For example, scenarios are displayed in which the user interacts with a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style. The scenario selected by the user is sent to the server (input). The server receives this scenario information and proceeds to the next process (output).

[0087] Step 2:

[0088] The server generates a prompt.

[0089] The server generates a prompt based on the scenario selected by the user (input). Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to the role (output).

[0090] Step 3:

[0091] The server inputs a prompt sentence into the generative AI model.

[0092] The server inputs the generated prompt sentence into the generative AI (input). The generative AI generates a response based on this prompt sentence. For example, a response such as "Boss, can you give me a moment? Actually, there's something I've been wanting to say for a while" is generated (output).

[0093] Step 4:

[0094] The server sends the generated response to the terminal.

[0095] The server sends the response obtained from the generative AI to the terminal (input), which receives the response and displays it to the user (output).

[0096] Step 5:

[0097] The terminal displays the response to the user.

[0098] The terminal displays the response received from the server to the user (input), who then looks at it and prepares to enter the next message (output).

[0099] Step 6:

[0100] The user enters the following message:

[0101] The user responds to the displayed response by typing the next message (input). For example, "Of course, what's wrong?" The user's input message is sent to the server via the terminal (output).

[0102] Step 7:

[0103] The server inputs the user's message into a generative AI model.

[0104] The server inputs the message entered by the user into the generative AI (input). The generative AI generates the following response based on this message. For example, it might generate a response like, "Actually, I want to come out as a sexual minority, but I don't know how to say it..." (output).

[0105] Step 8:

[0106] The server sends the generated response to the terminal.

[0107] The server sends the next response obtained from the generative AI to the terminal (input). The terminal receives this response and displays it to the user (output).

[0108] Step 9:

[0109] The terminal displays the response to the user.

[0110] The terminal displays the response received from the server to the user again (input), who then looks at it and prepares to input the next message (output).

[0111] Step 10:

[0112] This process is repeated.

[0113] This process is repeated as long as the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and provide advice to their subordinates (input and output). The server may further include means for recognizing the user's emotions using an emotion engine described below, and when generating a response, may use the recognized user's emotions, the generated prompt sentence, and the generative AI to generate a response to the user's input message.

[0114] (Application example 1)

[0115] Next, a description will be given of Application Example 1 of Embodiment Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0116] Conventional systems using generative AI allow users to set specific roles and practice responses for those roles, but they lack training for the specific issues that corporate security personnel face, such as security issues, whistleblowing, and suspected information leaks. Furthermore, the system has limited functionality for analyzing the results of these responses to check for harassment tendencies and bias, and the quality of the feedback is insufficient. Furthermore, it does not take into consideration the use of the system as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. This means that corporate security personnel and managers are insufficiently trained to respond appropriately, potentially increasing security risks across the entire company.

[0117] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0118] In this invention, the server includes a means for assigning specific roles to the generative AI and allowing users to respond to those roles via chat; a means for analyzing the response results and checking for harassment tendencies and bias; a means for providing feedback on the analysis results; a means for simulating security issues and allowing personnel to practice their responses via chat; and a means for generating prompts using the generative AI model and providing appropriate responses. This allows corporate security personnel and managers to receive training on how to respond appropriately to security issues, whistleblowers, suspected information leaks, and other issues, thereby reducing security risks across the company. It can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues, contributing to employee mental health care.

[0119] "Generative AI" is artificial intelligence that generates natural language based on specific roles and scenarios and engages in dialogue with users.

[0120] "Specific roles" refer to specific scenarios or characters that are set in the generative AI, such as "a subordinate coming out as LGBT" or "a subordinate announcing a pregnancy."

[0121] "Methods via chat" refers to interfaces and functions that allow users to have text-based conversations with generative AI.

[0122] "Means for analyzing the response results" refers to algorithms and software that analyze the content of the dialogue between the user and the generative AI and evaluate whether there is any tendency toward harassment or bias.

[0123] "Means to check for harassment tendencies and bias" refers to a function that determines whether a user's response contains harassment or bias based on the content of the conversation.

[0124] "Means of providing feedback" refers to functions and interfaces that provide analysis results to users and indicate areas for improvement or attention.

[0125] "Means to simulate security issues" refers to a feature that allows users to set security-related scenarios, such as information leaks and whistle-blowing, in the generative AI and practice how to respond to them.

[0126] A "generative AI model" is a machine learning model for generating natural language and generating appropriate responses based on specific prompts.

[0127] A "prompt sentence" is an input sentence that sets a specific role or scenario for generative AI, and serves as a trigger to start a dialogue with the user.

[0128] A system for implementing this invention includes means for setting a specific role for a generative AI and for the user to respond to that role via chat, means for analyzing the results of that response and checking for harassment tendencies or bias, means for feeding back the results of the analysis, means for simulating security issues and allowing personnel to practice their responses via chat, and means for generating prompt sentences using a generative AI model and providing appropriate responses.

[0129] System program generation

[0130] The server assigns specific roles to the generative AI and provides an interface for users to respond to those roles via chat. The text entered by the user is sent to the generative AI model on the server, which generates an appropriate response. The generated response is returned to the user, and the dialogue progresses.

[0131] Processing Description

[0132] The server performs processing using the following hardware and software:

[0133] Hardware: Internet-connected server

[0134] Software: Python, OpenAI API

[0135] The server receives input text from the user and sends it to the generative AI model. The generative AI model generates an appropriate response based on the prompt. The generated response is returned to the user via the server. The server then analyzes the content of the dialogue to check for harassment tendencies and bias. The analysis results are then fed back to the user.

[0136] Specific examples

[0137] For example, to simulate a security officer responding appropriately to a call from an employee reporting a suspected data breach, the following prompt might be used:

[0138] You are an employee reporting a suspected data breach. Please respond appropriately to the user's inquiry.

[0139] User: I recently noticed some confidential information from my company has been leaked. What should I do?

[0140] you:

[0141] Using this prompt, a generative AI model can generate appropriate responses that security personnel can practice, such as:

[0142] You: That's a very important issue. First, please provide specific evidence and details of the situation. Also, I need to report this matter to my superiors and the security department immediately. A potential information leak requires a quick response, so could you please provide more details?

[0143] In this way, security personnel can be trained to respond appropriately to actual inquiries.

[0144] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0145] Step 1:

[0146] A user accesses the chat interface using a terminal and enters a prompt sentence based on a specific role. The entered prompt sentence is sent to the server. The input data is the user's text input, and the output data is the prompt sentence sent to the server.

[0147] Step 2:

[0148] The server sends the received prompt to the generative AI model, which generates an appropriate response based on the prompt. The input data is the prompt, and the output data is the generated response. Specifically, the server calls the OpenAI API and sends the prompt.

[0149] Step 3:

[0150] The server receives the response generated by the generative AI model and returns it to the user's device. The input data is the response from the generative AI model, and the output data is the response sent to the user's device. Specifically, the server displays the generated response on the user's chat interface.

[0151] Step 4:

[0152] The user confirms the generated response and provides further input. The user's new input is again sent to the server, and the process repeats from step 2. The input data is the user's new text input, and the output data is the new prompt sent to the server.

[0153] Step 5:

[0154] The server analyzes the dialogue between the user and the generated AI model in real time to check for harassment tendencies and bias. The input data is the dialogue content, and the output data is the analysis results. Specifically, the server analyzes the dialogue content using a natural language processing algorithm.

[0155] Step 6:

[0156] The server feeds back the analysis results to the user. The input data is the analysis results, and the output data is the feedback provided to the user. In specific operations, the server displays the analysis results in text format on the user's chat interface.

[0157] Step 7:

[0158] The server simulates security issues and generates new prompts for users to practice their responses in chat. The input data is the security scenario settings, and the output data is the new prompts. Specifically, the server generates new prompts using a generative AI model.

[0159] Step 8:

[0160] The user then practices responding to the new prompt sentence again in the chat interface. The input data is the user's text input for the new prompt sentence, and the output data is the new prompt sentence sent to the server. As a specific operation, the user responds to the new scenario using the chat interface.

[0161] Example 2

[0162] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0163] With the previous system, it was difficult to properly analyze the results of user responses and check for harassment tendencies or bias. Furthermore, there was an insufficient means of providing feedback to users about the results of the analysis, making it difficult to lead to improvements in user behavior. Furthermore, there was a lack of specific methods for evaluating employee responses in specific situations.

[0164] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0165] In this invention, the server includes a means for a user to respond to a specific role via chat, a means for collecting the response results, a means for preprocessing the collected text data, a means for inputting the preprocessed text data into a machine learning model to analyze the presence or absence of harassment trends and bias, and a means for feeding back the analysis results. This makes it possible to appropriately analyze the user's response results and check for the presence or absence of harassment and bias. Furthermore, by feeding back the analysis results to the user, it is possible to promote improvement of the user's behavior. Furthermore, it is possible to provide a specific means for evaluating an employee's response in a specific situation.

[0166] "User" means an individual or entity that utilizes the System and fulfills a particular role.

[0167] A "specific role" refers to a specific situation or scenario that a user is required to handle within the system.

[0168] "Chat" is an interface that allows users to communicate text-based.

[0169] "Response results" are records of responses and actions taken by users for specific roles.

[0170] "Means for collection" refers to a method or device for acquiring and storing the results of a user's response as data.

[0171] "Text data" is data that expresses the user's response results as text information.

[0172] The "preprocessing means" refers to a method or device that processes collected text data to convert it into an analyzable format.

[0173] A "machine learning model" is an algorithm or system that learns patterns and trends from data and makes predictions and classifications.

[0174] "Means for analysis" refers to a method or device that uses preprocessed text data to evaluate the presence or absence of harassment tendencies or bias.

[0175] "Feedback means" refers to a method or device for communicating the analysis results to the user.

[0176] "Harassment" refers to inappropriate words, actions or behaviour towards others.

[0177] "Bias" refers to making judgments or taking actions based on specific prejudices or preconceptions.

[0178] This invention is a system in which users respond to specific roles via chat, analyze the results of their responses to check for harassment tendencies and bias, and provide feedback on the analysis results. A specific embodiment of this system is described below.

[0179] First, the server provides an interface for users to respond to specific roles in chat. This interface can be accessed through devices such as web browsers and mobile applications. For example, a user can be assigned a role as an "employee in a specific situation," and respond in chat based on that role.

[0180] The server then collects the user's responses, which are then stored in a database as text data. The collected text data is preprocessed using a natural language processing (NLP) library (e.g., spaCy or NLTK). Preprocessing includes tokenization, stop word removal, stemming, etc.

[0181] The preprocessed text data is input into a machine learning model (e.g., scikit-learn or TENSORFLOW (registered trademark)). The server uses this machine learning model to analyze the text data and determine whether there is any harassment tendency or bias. Specifically, it evaluates whether specific keywords, phrases, tones, etc. are offensive.

[0182] The analysis results are stored in a database by the server. In addition, the server provides feedback to the user. For example, if a message sent by a user is judged to be offensive, the server will provide feedback such as, "Your message contains offensive language. Please try to use more polite language."

[0183] As a concrete example, consider the case where a user sends a message saying, "Your response is completely useless. You're incompetent." In this case, the server collects this message and preprocesses it using an NLP library. Then, using a machine learning model, it determines that this message is offensive. Finally, the server provides appropriate feedback to the user.

[0184] An example of a prompt might be:

[0185] "Please generate a program that analyzes user responses and checks for harassment tendencies and bias. Specifically, use natural language processing and machine learning techniques to determine whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc."

[0186] In this way, the server can properly analyze the user's response and check for harassment or bias. Furthermore, by providing feedback to the user, the analysis results can encourage the user to improve their behavior. Furthermore, the server can provide a concrete means for evaluating employee responses in specific situations.

[0187] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0188] Step 1:

[0189] The server provides an interface for users to respond to specific roles on the chat. Users access this interface through their terminals and start chats based on specific roles. The input is the user's chat message, and the output is the text data of that message.

[0190] Step 2:

[0191] The server collects users' chat messages and stores them in a database. The input is the user's chat messages, and the output is the stored text data. Specifically, the server captures messages in real time and writes them to the database.

[0192] Step 3:

[0193] The server preprocesses the collected text data. Specifically, it uses an NLP library (e.g., spaCy or NLTK) to perform the following processes: tokenization, stop word removal, and stemming. The input is the stored text data, and the output is the preprocessed text data. Specifically, the server splits the text data into tokens, removes unnecessary words, and extracts word stems.

[0194] Step 4:

[0195] The server inputs the preprocessed text data into a machine learning model for analysis. Specifically, it uses libraries such as scikit-learn and TensorFlow to determine whether there is any harassment tendency or bias. The input is the preprocessed text data, and the output is the analysis results. Specifically, the server inputs the text data into the model and evaluates specific keywords, phrases, and tone.

[0196] Step 5:

[0197] The server stores the analysis results in a database. The input is the analysis results, and the output is the stored analysis results. As a specific operation, the server writes the analysis results into the database.

[0198] Step 6:

[0199] The server provides feedback to the user. The input is the saved analysis result, and the output is a feedback message to the user. Specifically, the server generates a feedback message based on the analysis result and sends it to the user. For example, it provides a message such as, "Your message contains offensive language. Please try to use more polite language."

[0200] In this way, the server can properly analyze the user's response results, check for harassment or bias, and provide feedback.

[0201] (Application example 2)

[0202] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0203] In today's corporate environment, harassment and bias issues are serious, and there is a need to detect these issues early and respond appropriately. However, traditional methods make it difficult to detect these issues in real time and respond immediately. In addition, signs of harassment and bias are often overlooked, which can have a negative impact on the company's credibility and the mental health of employees. Therefore, effective measures to resolve these issues are needed.

[0204] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0205] In this invention, the server includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies and bias, means for feeding back the analysis results, means for analyzing chat and email content in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate responses to users. This makes it possible to detect harassment and bias issues in real time and respond immediately.

[0206] "Generative AI" is an artificial intelligence system that is assigned a specific role and interacts with the user based on that role.

[0207] A "specific role" refers to a specific scenario or character that is set in the generative AI, and is the role that the user responds to.

[0208] "User responses" refer to chat interactions and reactions that users have with specific roles assigned to generative AI.

[0209] "Response results" refer to the content of the interactions and reactions the user had with the generative AI.

[0210] "Harassment" refers to inappropriate words, actions, or behavior toward others, and is problematic behavior in the workplace and in society.

[0211] "Bias" refers to judgments or actions based on specific prejudices or preconceptions, and is a factor that leads to a lack of fairness.

[0212] "Means of checking" refers to methods and techniques for analyzing the results of users' responses and determining whether or not there is harassment or bias.

[0213] "Feedback means" refers to methods and techniques for notifying users and administrators of the analysis results and encouraging them to take appropriate action.

[0214] "Real-time analysis means" refers to methods and technologies for instantly analyzing chat and email content to detect signs of harassment or bias.

[0215] "Alert methods" are methods or technologies used to alert administrators when signs of harassment or bias are detected.

[0216] "Means for storing data in a database" refers to methods and technologies for recording the nature of any detected harassment or bias, making it available for later analysis and reference.

[0217] "Means for suggesting appropriate responses" are methods and techniques for suggesting appropriate responses to harassment and bias issues to users.

[0218] A system for implementing this invention includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies or bias, means for feeding back the analysis results, means for analyzing the content of chats and emails in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate ways of responding to users.

[0219] System configuration

[0220] The server uses a generative AI model to set specific roles, and users can respond to those roles in the chat. The user's responses are analyzed using a natural language processing library (e.g., spaCy). The analyzed data is then checked for harassment and bias using a machine learning model (e.g., scikit-learn's Logistic Regression).

[0221] Data processing and calculation

[0222] The server analyzes the content of users' chats and emails in real time. The analyzed data is converted into TF-IDF vectors and input into a machine learning model. The model detects signs of harassment and bias and sends the results as an alert to an administrator. The detected content is also stored in a database for later detailed analysis.

[0223] Hardware and software used

[0224] Hardware: Servers, user devices (PCs, smartphones, etc.)

[0225] Software: Generative AI model, natural language processing library (spaCy), machine learning library (scikit-learn), database management system

[0226] Specific examples

[0227] For example, if a company were to use this system for customer support chat, they would input the following prompts into the generative AI model:

[0228] Analyze customer support chats in real time to detect signs of harassment and bias. Analyze messages such as:

[0229] "You're really incompetent."

[0230] "Thank you for your amazing support!"

[0231] "You're an idiot."

[0232] Based on this prompt, the generative AI model analyzes the chat content to detect signs of harassment or bias, and if detected, sends an alert to an administrator and suggests appropriate responses.

[0233] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0234] Step 1:

[0235] The server assigns specific roles to the generative AI model. Specifically, it assigns scenarios such as "a subordinate coming out as LGBT" or "a subordinate announcing pregnancy" to the model. Based on this assignment, it generates a prompt for the user to start a chat.

[0236] Input: Specific Role

[0237] Output: prompt statement

[0238] Step 2:

[0239] Users respond to specific roles assigned by the generative AI model through chat, and their input is sent to the server as chat messages.

[0240] Input: User's chat message

[0241] Output: Message sent to the server

[0242] Step 3:

[0243] The server analyzes the received chat messages using a natural language processing library (e.g., spaCy), and performs data processing such as text tokenization and TF-IDF vectorization on the analyzed data.

[0244] Input: User's chat message

[0245] Output: Parsed text data (tokenized, vectorized)

[0246] Step 4:

[0247] The server then feeds the parsed text data into a machine learning model (e.g., scikit-learn's LogisticRegression) to check for harassment and bias. The model makes predictions based on the pre-trained data.

[0248] Input: Parsed text data

[0249] Output: Prediction results of harassment and bias

[0250] Step 5:

[0251] Based on the prediction results, the server sends an alert to the administrator if signs of harassment or bias are detected. The alert includes the content of the detected message and the prediction results.

[0252] Input: Predicted results of harassment and bias

[0253] Output: Alert to administrator

[0254] Step 6:

[0255] The server stores any detected instances of harassment or bias in a database, which is then used for further analysis and reporting.

[0256] Input: Detected harassment or bias

[0257] Output: Data stored in the database

[0258] Step 7:

[0259] The server then proposes appropriate responses to the user, based on generative AI models and pre-defined response guidelines.

[0260] Input: Predicted results of harassment and bias

[0261] Output: Suggestions for the user on how to respond

[0262] Example 3

[0263] Next, a description will be given of a third embodiment of the third embodiment. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0264] With conventional systems, it was difficult to effectively detect bias or harassment tendencies in users' responses and provide specific improvements. Furthermore, when supporting employees returning to work after taking a leave of absence due to mental health issues, there was a lack of means to simulate appropriate responses and provide feedback. This limited opportunities for users to learn and improve appropriate responses.

[0265] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[0266] In this invention, the server includes a means for the user to input a prompt sentence, a means for the terminal to send the prompt sentence to the server, a means for the server to execute the generative AI model, a means for the server to analyze the results of the generative AI model, a means for the server to generate feedback, and a means for the terminal to display the feedback to the user. This makes it possible to effectively detect bias or harassment tendencies in the user's behavior and provide specific improvements and advice. It is also possible to simulate appropriate response methods and provide feedback to support the return of employees who have taken leave due to mental health issues.

[0267] A "prompt" is the text of an instruction or question that a user enters into a system.

[0268] A "terminal" is a device on which a user inputs a prompt and communicates with a server. Examples include PCs, smartphones, and tablets.

[0269] A "server" is a computer system that runs the generative AI model, analyzes prompts, and generates feedback.

[0270] A "generative AI model" is an artificial intelligence model that simulates a response based on a user's prompt and generates the results.

[0271] "Feedback" refers to specific improvements and advice provided to users after analyzing the results of the generative AI model.

[0272] "Bias" refers to prejudices or preconceived notions contained in user responses.

[0273] "Harassment" refers to inappropriate words, actions, or behavior in a user's interactions with others.

[0274] "Mental health" refers to the state of mental health and is a concept that includes psychological problems such as stress, anxiety, and depression.

[0275] A "return-to-work program" is a plan or procedure for assisting employees who have taken leave due to mental health issues to return to work.

[0276] This invention is a system in which a user inputs a prompt sentence, executes a generative AI model based on the prompt sentence, and provides feedback. A specific embodiment of this system will be described below.

[0277] First, the user inputs a prompt using a terminal. A terminal is a device such as a PC, smartphone, or tablet that provides an interface for users to access the system. For example, a user can input a prompt like the following:

[0278] "How should I respond to a subordinate who comes out as LGBT?"

[0279] How should I respond to an employee returning from leave?

[0280] The device sends the prompts entered by the user to a server, which has the hardware and software to run the generative AI model, specifically, a server equipped with an NVIDIA GPU and a deep learning framework such as TensorFlow or PyTorch.

[0281] The server inputs the received prompt sentence into a generative AI model, which then simulates the user's response and generates the results. This generative AI model then analyzes the prompt sentence using natural language processing technology and suggests an appropriate response.

[0282] The output of the generative AI model is analyzed by a server, which uses natural language processing techniques, such as sentiment analysis and keyword extraction, to analyze the results of the generative AI model and detect bias or harassment in user responses.

[0283] The server generates feedback based on the analysis results, including specific areas for improvement and advice. For example, it might say, "It's important to respond to subordinates who come out as LGBT without prejudice. Specifically, we recommend that you listen carefully to what your subordinates have to say and show understanding."

[0284] Finally, the device displays the feedback received from the server to the user, who can then review the feedback on the device screen and use it as a reference for improving their own response.

[0285] In this way, this invention can effectively detect bias and harassment tendencies in user behavior and provide specific improvements and advice. It can also simulate appropriate response methods and provide feedback when supporting the return of employees who have taken leave due to mental health issues. The flow of the identification process in Example 3 will be explained using Figure 15.

[0286] Step 1:

[0287] The user enters a prompt statement.

[0288] The user inputs a prompt through the device interface. For example, they might input, "Please tell me how to respond to a subordinate who comes out as LGBT." The input prompt is temporarily stored in the device's memory.

[0289] Step 2:

[0290] The terminal sends the prompt to the server.

[0291] The terminal converts the prompt text entered by the user into JSON format and sends it to the server using an HTTP POST request. The destination endpoint is the server's API endpoint. The input is the prompt text, and the output is the request sent to the server.

[0292] Step 3:

[0293] The server runs the generative AI model.

[0294] The server performs data preprocessing on the received prompt sentences in order to input them into the generative AI model. Specifically, it tokenizes and normalizes the text. It then runs the generative AI model to simulate the user's response. The input is the preprocessed prompt sentence, and the output is the simulation result of the generative AI model.

[0295] Step 4:

[0296] The server analyzes the results of the generated AI model.

[0297] The server uses natural language processing techniques to analyze the output of the generative AI model. For example, it performs sentiment analysis of the text and keyword extraction to detect bias in the user's responses. The input is the simulation result of the generative AI model, and the output is the analysis result.

[0298] Step 5:

[0299] The server generates the feedback.

[0300] The server generates feedback based on the analysis results. The feedback includes specific advice and suggestions for improving the user's response. The generated feedback is sent to the terminal in JSON format. The input is the analysis results, and the output is the feedback.

[0301] Step 6:

[0302] The device displays the feedback to the user.

[0303] The device uses an appropriate UI component to display the feedback received from the server to the user. For example, it displays the feedback in a popup window or as a notification. The user can check the feedback and use it to improve their own response. The input is the feedback, and the output is what is displayed to the user.

[0304] (Application example 3)

[0305] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0306] There are issues with responding appropriately when employees have mental health issues or in specific situations. In particular, there is a lack of simulations and feedback for machines working in factories to respond appropriately to employees, which can delay improvements in employee mental health and the working environment.

[0307] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 3 is realized by the following means. In this invention, the server includes means for setting a specific role for the generative AI and for the user to respond to that role via chat, means for analyzing the response results and checking for harassment tendencies and bias, means for feeding back the analysis results, means for providing a simulation for machines working in a factory to respond appropriately to employees, and means for analyzing the simulation results and providing improvements and advice. This makes it possible to simulate and provide feedback for appropriate responses when employees have mental health issues or in specific situations.

[0308] "Generative AI" is an artificial intelligence system that assigns specific roles and allows users to respond to those roles.

[0309] "Chat" is a text-based means of communication that allows users to interact with generative AI.

[0310] "Harassment" refers to inappropriate behavior or language directed at an individual or group in the workplace or other environment.

[0311] "Bias" refers to prejudices or preconceived notions held against a particular individual or group.

[0312] "Feedback" is a means of providing evaluations and suggestions for improvement of actions taken by users.

[0313] "Factory machines" refers to automated devices and robots used to perform tasks in a factory.

[0314] "Simulation" refers to the process of virtually recreating a specific situation or role and allowing users to respond within that situation.

[0315] "Mental health" refers to an individual's mental well-being and includes issues such as stress, anxiety and depression.

[0316] "Employee" refers to people who work for a particular company or organization.

[0317] "Areas for improvement" refers to the actions taken by the user that need to be corrected or changed to achieve better results.

[0318] "Advice" refers to specific instructions or suggestions to help users take more appropriate action.

[0319] To implement this invention, it is first necessary to build a system in which specific roles are assigned to generative AI and users respond to those roles via chat. This system has the function of analyzing the user's responses and checking for harassment tendencies and biases. It also includes a function to provide feedback on the analysis results.

[0320] Furthermore, a function that provides simulations to help machines working in factories respond appropriately to employees is also required. The results of these simulations are analyzed and suggestions and advice are provided to improve employee mental health and the working environment.

[0321] Hardware and software used

[0322] Hardware: Automated equipment and robots used in factories (e.g., ABB YuMi, Fanuc CR series)

[0323] Software: OpenAI API, Python

[0324] Data processing and calculation

[0325] server:

[0326] 1. Receive input from the user.

[0327] 2. Send prompts to a generative AI model (OpenAI's GPT-3®) to generate a robotic response.

[0328] 3. The generated response is sent back to the generative AI model to analyze the response and generate feedback.

[0329] Device:

[0330] It provides an interface that allows users to interact with generative AI through chat.

[0331] User:

[0332] Responses to specific roles are made via chat and the results are sent to the system.

[0333] Specific examples

[0334] For example, if an employee enters, "I recently took a leave of absence due to work stress. I'm worried about returning to work," the system will act as follows:

[0335] 1. The server receives input from the user and sends prompts to the generative AI model.

[0336] 2. The generative AI model generates a response based on the following prompt:

[0337] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[0338] robot:

[0339] 3. The server sends the generated response back to the generative AI model to analyze the response and generate feedback.

[0340] 4. Feedback is provided to users and appropriate actions and improvements are suggested.

[0341] In this way, simulations and feedback can be provided to enable machines working in factories to respond appropriately to employees.

[0342] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0343] Step 1:

[0344] The server receives input from the user. Specifically, it receives text data entered by the user through the chat interface. This input may include, for example, "I recently took a leave of absence due to work stress. I'm worried about returning to work." The input data is sent to the server in text format.

[0345] Step 2:

[0346] The server sends the received user input to the generative AI model. Specifically, it uses OpenAI's API to send the following prompt sentence to the generative AI model:

[0347] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[0348] robot:

[0349] Based on this prompt, the generative AI model generates an appropriate response. The input data is sent in the form of a prompt, and the output data is the generated response text.

[0350] Step 3:

[0351] The server receives a response from the generative AI model. Specifically, it receives text data returned by the generative AI model. This response may include, for example, "When you return to work, try setting small goals first." The input data is the response text from the generative AI model, and the output data is that response text.

[0352] Step 4:

[0353] The server sends the generated response back to the generative AI model to analyze the response and generate feedback. Specifically, it sends the following prompt sentence to the generative AI model:

[0354] Please analyze the responses below and offer your suggestions and suggestions for improvement.

[0355] Response: When you return, try setting small goals first.

[0356] feedback:

[0357] Based on this prompt, the generative AI model analyzes the response and generates feedback. The input data is the response text from the generative AI model, and the output data is the feedback text.

[0358] Step 5:

[0359] The server receives feedback from the generative AI model. Specifically, it receives the feedback text returned by the generative AI model. This feedback includes, for example, "This response is appropriate, but it would be good to suggest specific support methods as well." The input data is the feedback text from the generative AI model, and the output data is the feedback text.

[0360] Step 6:

[0361] The server provides the generated feedback to the user. Specifically, the server displays the feedback text in the user's chat interface. Based on this feedback, the user can improve their own response. The input data is the feedback text, and the output data is the feedback displayed to the user.

[0362] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0363] "Example 1"

[0364] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[0365] "Example 2"

[0366] Another embodiment of the present invention is a system in which an emotion engine adjusts the AI's response based on the user's emotional state. In this system, when a user expresses joy or satisfaction to the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​may respond in a way that shares the user's joy or send a message of gratitude to the user. This allows for a more appropriate response based on the user's emotional state.

[0367] "Example 3"

[0368] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[0369] The processing flow of each embodiment will be described below.

[0370] "Example 1"

[0371] Step 1: The user initiates a chat conversation with the AI.

[0372] Step 2: The AI ​​interacts with the user based on the assigned role.

[0373] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[0374] Step 4: The emotion engine analyzes the user's emotional state.

[0375] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[0376] "Example 2"

[0377] Step 1: The user initiates a chat conversation with the AI.

[0378] Step 2: The AI ​​interacts with the user based on the assigned role.

[0379] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[0380] Step 4: The emotion engine analyzes the user's emotional state.

[0381] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[0382] Example 1

[0383] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0384] Conventional workplace communication and harassment countermeasures are difficult to simulate in real situations and lack the means to properly grasp users' emotional states. This limits opportunities for supervisors and managers to learn appropriate responses, making it easy for harassment and bias to occur. Furthermore, the lack of feedback based on users' emotional states makes it difficult to improve response skills in the real workplace.

[0385] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0386] In this invention, the server includes a means for assigning specific roles to generative artificial intelligence and allowing users to respond to those roles via communication means, a means for analyzing the results of the responses and checking for harassment tendencies and bias, a means for analyzing the user's emotional state using an emotion engine, and a means for providing feedback on the analysis results. This allows users to improve their ability to respond in the actual workplace and prevent problems such as harassment and bias. Furthermore, providing appropriate feedback according to the user's emotional state enables more effective communication.

[0387] "Generative AI" is AI that can generate natural language based on user input and behave as a character with a specific role.

[0388] "Specific roles" refer to specific scenarios or character roles that are set for generative AI, such as "a subordinate who comes out as a sexual minority" or "a subordinate who announces pregnancy."

[0389] "Communication means" refers to the interface through which users interact with generative AI, including chat applications and messaging platforms.

[0390] "Harassment tendencies" are patterns that indicate a user's comments or actions may be inappropriately pressuring or harassing others.

[0391] "Bias" refers to prejudices and preconceived notions contained in users' statements and actions, and indicates unfair treatment of people with certain attributes or backgrounds.

[0392] An "emotion engine" is a software component for extracting and analyzing emotions from a user's text or speech.

[0393] "Feedback" refers to the responses and advice provided by generative artificial intelligence or emotion engines based on the user's statements and emotional state.

[0394] "Mental health" refers to an individual's psychological stability and sense of well-being, and indicates a state in which they are free from mental problems such as stress, anxiety, and depression.

[0395] An "employee" is an individual employed by a business or organization who works to perform a specific role or function.

[0396] This invention is a system that combines generative AI and an emotion engine, and aims to improve responsiveness in the workplace by allowing users to interact with AI with specific roles. This system is realized using a server, terminals, a generative AI model, and an emotion engine.

[0397] Hardware and software used

[0398] Hardware: Servers, terminals (PCs, smartphones)

[0399] Software: Generative AI models (e.g., GPT-4), emotion engines (e.g., Affectiva)

[0400] System Overview

[0401] 1. Role Settings:

[0402] The server assigns specific roles to the generative AI model.

[0403] Specific roles include "a subordinate who comes out as a sexual minority," "a subordinate who announces pregnancy," and "a subordinate who is worried about their work style."

[0404] 2. Start chatting:

[0405] A user accesses the chat application using a terminal.

[0406] As a boss, the user starts chatting with an AI with a set role.

[0407] 3. Analysis of response results:

[0408] The server analyzes the chat content in real time.

[0409] Run algorithms to check for harassment patterns and bias.

[0410] 4. Emotional state analysis:

[0411] The emotion engine analyzes the user's emotional state.

[0412] Recognizes emotions such as anger or frustration when a user expresses them.

[0413] 5. Feedback:

[0414] The server adjusts the response of the generative AI model based on the analysis results of the emotion engine.

[0415] For example, the generative AI model is instructed to "respond in a way that will ease the user's anger."

[0416] Specifically, the generative AI model sends an apology message such as, "I'm sorry, I may have misunderstood you."

[0417] Examples and prompts

[0418] Examples:

[0419] A user starts a chat with a subordinate who is coming out as a sexual minority.

[0420] User: "Is anything bothering you lately?"

[0421] AI (subordinate role): "Actually, I want to come out as a LGBTQ person."

[0422] Example prompt sentence:

[0423] "As a manager, you are being consulted by a subordinate who wants to come out as a sexual minority. How would you respond?"

[0424] This system allows users to improve their ability to respond in the workplace. In addition, the emotion engine enables appropriate feedback based on the user's emotional state. This can prevent harassment and bias issues and lead to more effective communication.

[0425] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0426] Step 1:

[0427] Role Settings

[0428] The server assigns specific roles to the generative AI model.

[0429] Input: Role information set by the administrator (e.g., "Subordinate who comes out as a sexual minority")

[0430] Data processing: The server sends the role information to the generative AI model to generate a prompt.

[0431] Output: Role and prompt set for the generative AI model

[0432] Specific behavior: The server sends the prompt message "You are a subordinate who is coming out as a sexual minority. Please talk to your supervisor." to the generative AI model.

[0433] Step 2:

[0434] Start chatting

[0435] A user accesses the chat application using a terminal.

[0436] Input: User login information and chat start request

[0437] Data processing: The server authenticates the user and starts the chat session.

[0438] Output: Notification that a chat session has started

[0439] Specific behavior: The user sends a message such as "Is there anything bothering you lately?"

[0440] Step 3:

[0441] Analysis of response results

[0442] The server analyzes the chat content in real time.

[0443] Input: Chat content between the user and the generated AI model

[0444] Data processing: The server uses natural language processing algorithms to check for harassment patterns and bias.

[0445] Output: Analysis results (presence or absence of harassment trends and bias)

[0446] Specific behavior: The server determines whether the user's statement contains a specific bias.

[0447] Step 4:

[0448] Emotional state analysis

[0449] The emotion engine analyzes the user's emotional state.

[0450] Input: User chat content

[0451] Data processing: The emotion engine extracts emotions from text and recognizes emotions such as anger or frustration.

[0452] Output: Sentiment analysis results (e.g. anger, sadness, joy)

[0453] What it does: If a user expresses anger, such as "Why would you say that?", the emotion engine will detect that emotion.

[0454] Step 5:

[0455] feedback

[0456] The server adjusts the response of the generative AI model based on the analysis results of the emotion engine.

[0457] Input: Sentiment analysis results and chat content

[0458] Data processing: The server instructs the generative AI model on how to respond based on the user's emotional state.

[0459] Output: The adjusted generative AI model response

[0460] What it does: The generative AI model sends an apology message such as, "Sorry, I may have misunderstood you."

[0461] (Application example 1)

[0462] Next, a description will be given of Application Example 1 of Embodiment Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0463] Conventional generative AI systems have difficulty properly analyzing a user's emotional state and providing feedback when the user responds to a specific role. Furthermore, they lack the ability to adjust the AI's response based on the user's emotional state, which can result in the AI ​​sometimes failing to respond appropriately to the user. This has led to problems such as inaccurate analysis when checking for harassment tendencies or bias.

[0464] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for setting a specific role for the generative AI and for the user to respond to that role via chat, means for analyzing the response results and checking for harassment tendencies and bias, means for analyzing the user's emotional state using an emotion engine and feeding back the results, and means for adjusting the AI's response according to the user's emotional state. This enables appropriate responses according to the user's emotional state and makes it possible to accurately check for harassment tendencies and bias.

[0465] "Generative AI" is an artificial intelligence system that generates natural language responses based on user input.

[0466] A "specific role" refers to a specific scenario or character that is set in the generative AI, and is the role that the user responds to.

[0467] "Chat" is a means of text-based real-time communication between users and generative AI.

[0468] "Harassment" refers to inappropriate words, actions, or behavior toward others in the workplace or in society, which causes mental or physical pain to the other person.

[0469] "Bias" is the distortion of judgment or behavior based on a particular prejudice or preconception.

[0470] The "emotion engine" is a system that analyzes the user's emotional state from their text input and provides appropriate feedback based on the results.

[0471] "Feedback" refers to the response or evaluation that a generative AI gives to a user's actions or input.

[0472] "Security" refers to the measures and procedures used to protect information and physical assets and to guard against unauthorized access and attacks.

[0473] "Mental health" refers to the state of mental health and is a concept that includes mental problems such as stress, anxiety, and depression.

[0474] A "return-to-work program" is a plan or procedure for providing assistance to employees returning to work after a leave of absence.

[0475] A system for implementing this invention includes a means for setting a specific role for generative AI and for the user to respond to that role via chat, a means for analyzing the results of the response and checking for any tendency toward harassment or bias, a means for analyzing the user's emotional state using an emotion engine and feeding back the results, and a means for adjusting the AI's response according to the user's emotional state.

[0476] 1. System Program

[0477] This system includes the following programs:

[0478] Generative AI: Generates natural language responses based on user input.

[0479] Emotion engine: Analyzes the user's emotional state from their input text.

[0480] Feedback module: Provides appropriate feedback to users based on the analysis results of the emotion engine.

[0481] Adjustment module: Adjusts the response of the generative AI depending on the user's emotional state.

[0482] 2. Program processing explanation

[0483] Hardware and software used

[0484] Hardware: Smartphone

[0485] Software: Python, OpenAI API, TextBlob

[0486] Data processing and calculation

[0487] 1. Receiving user input: The user enters input in chat format using their smartphone.

[0488] 2. Sentiment Analysis: Analyze the sentiment of the user's input text using TextBlob. Determine whether the sentiment is positive, negative, or neutral.

[0489] 3. AI Response Generation: Uses OpenAI's API to generate appropriate responses to user input.

[0490] 4. Feedback adjustment: Based on the results of sentiment analysis, adjust the AI's response and provide feedback to the user.

[0491] 3. Specific Examples

[0492] Specific examples

[0493] If a user types, "I've been worried about the security of my home lately. What should I do?", sentiment analysis will determine this as negative, and the AI ​​will respond with, "I understand your concerns. Can you tell me more?"

[0494] Prompt Sentence Examples

[0495] User: I'm worried about the security of my home these days. What should I do?

[0496] AI: I understand your concerns. Can you elaborate a bit more?

[0497] In this way, it is possible to respond appropriately to the user's emotional state and accurately check for harassment tendencies and biases.

[0498] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0499] Step 1:

[0500] The user inputs information in chat format using a smartphone. The input text is sent to the server. The input data is the user's inquiry, for example, "I've been worried about the security of my home lately. What should I do?"

[0501] Step 2:

[0502] The server passes the received user input text to the emotion engine. The emotion engine uses TextBlob to analyze the sentiment of the text. Specifically, it determines whether the sentiment of the text is positive, negative, or neutral. For example, if the input text contains the negative sentiment "I'm worried," the emotion engine will determine it as negative.

[0503] Step 3:

[0504] The server sends a prompt to the generative AI model based on the analysis results of the emotion engine. The prompt includes the user's input text and the results of the emotion analysis. For example, the server sends the prompt "User: I've been worried about the security of my home lately. What should I do? Emotion: Negative" to the generative AI model.

[0505] Step 4:

[0506] The generative AI model generates an appropriate response based on the prompt sentence, which corresponds to the user's emotional state, such as "I understand your concerns. Can you tell me more?"

[0507] Step 5:

[0508] The server receives the response from the generative AI model and passes it to the feedback module, which adjusts the final feedback to the user based on the analysis results of the emotion engine and the response of the generative AI model, for example, adding an additional message to alleviate negative emotions.

[0509] Step 6:

[0510] The server then sends the final feedback to the user's smartphone, where the user can view the response from the generative AI model and continue chatting. For example, the message might read, "I understand your concerns. Can you tell me more?"

[0511] In this way, it is possible to respond appropriately according to the user's emotional state, and to accurately check for harassment tendencies and biases.

[0512] Example 2

[0513] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0514] With conventional systems, it was difficult to properly analyze the results of user responses and check for harassment tendencies and bias. It was also difficult to adjust the AI's response based on the user's emotional state, which meant that it was unable to provide appropriate feedback to the user.

[0515] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for setting a specific role for the generative artificial intelligence and for the user to respond to that role via communication means; means for collecting the response results and storing them in a database; means for analyzing the stored data using natural language processing technology; means for determining whether the analyzed data has a tendency toward harassment or bias using a machine learning model; means including an emotion engine for analyzing the user's emotional state and adjusting the response; and means for feeding back the analysis results. This makes it possible to appropriately analyze the user's response results and check for a tendency toward harassment or bias. Furthermore, the AI's response can be adjusted based on the user's emotional state, making it possible to provide appropriate feedback to the user.

[0516] "Generative AI" is an AI system that is assigned a specific role and interacts with the user based on that role.

[0517] "Means of communication" refers to the devices and networks that users use to interact with generative artificial intelligence.

[0518] The "database" is an information management system for storing the results of users' responses.

[0519] "Natural language processing technology" is a technology that analyzes text data, extracts specific keywords and phrases, and evaluates context and tone.

[0520] A "machine learning model" is an algorithm that is trained on past data to identify specific patterns and trends.

[0521] The "emotion engine" is a system that analyzes the user's emotional state and adjusts the AI's response based on the results.

[0522] "Feedback means" refers to a method or system for communicating the analysis results to the user.

[0523] This invention is a system that uses generative AI to analyze user responses and check for harassment tendencies and biases. It also has the ability to adjust the AI's response based on the user's emotional state.

[0524] Hardware and software used

[0525] This system operates using a server and a terminal. The server is a cloud-based server equipped with a high-performance processor and large memory capacity. Specifically, it uses a virtual server from a general cloud service provider. The terminal is a device operated by the user, such as a PC or smartphone.

[0526] The software uses the following technologies:

[0527] Natural language processing libraries (e.g. spaCy, NLTK)

[0528] Machine learning frameworks (e.g. TensorFlow, PyTorch)

[0529] Sentiment analysis engine (e.g., IBM Watson® Tone Analyzer)

[0530] Data processing and calculation

[0531] 1. Collecting user input data:

[0532] A user uses a terminal to input text data. For example, the user types, "This service is really great!"

[0533] The terminal transmits this data to the server.

[0534] 2. Data storage by the server:

[0535] The server stores the received text data in a database, such as MySQL (registered trademark) or PostgreSQL.

[0536] 3. Server-based Natural Language Processing (NLP):

[0537] The server analyzes the stored text data using a natural language processing library, for example, extracting the keyword "great" from the text.

[0538] 4. Analysis using machine learning models on the server:

[0539] The server uses machine learning models, such as models trained on historical data, to determine whether a message is harassing or biased based on the extracted keywords, phrases, and tone.

[0540] 5. Running the emotion engine on the server:

[0541] The server uses an emotion engine to analyze the user's emotional state, for example recognizing the emotion "happy."

[0542] 6. Feedback from the server to the user:

[0543] Based on the analysis, the server sends appropriate feedback to the user, such as a message like "That's great! I'm happy too."

[0544] Specific examples

[0545] Collecting user input data

[0546] Example: User types, "This service is absolutely amazing!"

[0547] Action: The device sends this text data to the server.

[0548] Server-based data storage

[0549] How it works: The server stores the received text data in a MySQL database.

[0550] Server-based natural language processing (NLP) execution

[0551] What it does: The server uses spaCy to extract the keyword "great".

[0552] Analysis using machine learning models on the server

[0553] How it works: The server uses a TensorFlow model to determine if the text is harassment-free or bias-free.

[0554] Emotion engine execution by server

[0555] How it works: The server uses IBM Watson's Tone Analyzer to recognize the emotion "happy."

[0556] Feedback from the server to the user

[0557] What it does: The server generates a message such as "That's great! I'm happy too" and sends it to the user via the device.

[0558] Prompt Sentence Examples

[0559] Prompt: "Analyze user-entered text to determine if it contains harassment or bias. Take appropriate action based on the user's emotional state."

[0560] In this way, the system operates in cooperation with the server, terminals, and users.

[0561] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0562] Step 1: Collecting User Input Data

[0563] Input: A user uses a terminal to enter text data. For example, a user types, "This service is really great!"

[0564] Specific operation: The device sends this text data to the server.

[0565] Output: The text data sent to the server.

[0566] Step 2: Server saves data

[0567] Input: Text data sent from the terminal.

[0568] Specific operation: The server stores the received text data in a database, for example, using a database such as MySQL or PostgreSQL.

[0569] Output: Text data stored in a database.

[0570] Step 3: Server performs natural language processing (NLP)

[0571] Input: Text data stored in a database.

[0572] What happens: The server uses a natural language processing library (e.g., spaCy) to analyze the text data. For example, it extracts the keyword "great" from the text.

[0573] Output: Extracted keywords and phrases, context and tone assessment results.

[0574] Step 4: Analysis using machine learning models on the server

[0575] Input: Keywords and phrases extracted through natural language processing, and assessment results of context and tone.

[0576] What it does: The server uses a machine learning model (e.g., TensorFlow) to determine whether or not harassment or bias exists based on the extracted information, for example, using a model trained on historical data.

[0577] Output: Judgment result on whether harassment or bias existed.

[0578] Step 5: Running the Emotion Engine on the Server

[0579] Input: User-entered text data.

[0580] Specific operation: The server uses an emotion engine (e.g., IBM Watson's Tone Analyzer) to analyze the user's emotional state. For example, it recognizes the emotion "happy."

[0581] Output: Analysis results about the user's emotional state.

[0582] Step 6: Feedback from the server to the user

[0583] Input: Judgment results regarding the presence or absence of harassment or bias, and analysis results regarding the user's emotional state.

[0584] Specific operation: Based on the analysis results, the server generates appropriate feedback for the user and sends it to the user via the device. For example, it sends a message such as "That's great! I'm happy too."

[0585] Output: The feedback message sent to the user.

[0586] (Application example 2)

[0587] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0588] It is necessary to detect signs of harassment and bias in employee communications early and respond appropriately. It is also important to understand employees' emotional state in real time and provide appropriate feedback and support. However, due to a lack of efficient and effective means to resolve these issues, improvements to employee mental health and the work environment are delayed.

[0589] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0590] In this invention, the server includes a means for setting a specific role for the generative AI and for the user to respond to that role via chat, a means for analyzing the response results and checking for harassment tendencies and bias, a means for feeding back the analysis results, a means for analyzing the emotional state of the user and adjusting the AI's response based on the emotion, and a means for monitoring communication between employees in real time and providing appropriate feedback and support. This enables early detection of harassment and bias in communication between employees and appropriate responses according to the emotional state of employees.

[0591] "Generative AI" is artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[0592] "Specific role" refers to the specific role or scenario that the generative AI should play in interacting with the user.

[0593] "User response" refers to the chat responses and actions that users take toward generative AI.

[0594] "Harassment" refers to words, actions, or behavior that cause discomfort or pain to others.

[0595] "Bias" refers to words, actions, or behavior that reflect prejudice or preconceived notions toward a particular individual or group.

[0596] "Emotional state" refers to the type and intensity of emotions a user displays during a conversation.

[0597] "Feedback" refers to the responses and information that generative AI provides based on the user's response and emotional state.

[0598] "Real-time" refers to immediate processing or response without delay.

[0599] "Communication between employees" refers to dialogue and information exchange between employees in the workplace.

[0600] "Appropriate feedback and support" refers to responses and assistance provided based on the user's emotional state and the results of their response, with the aim of solving problems and improving mental health.

[0601] The system for implementing this invention uses the following hardware and software. The hardware requires a smartphone and a server. The software uses Python, NLTK (a natural language processing library), and Scikit-learn (a machine learning library).

[0602] The server assigns specific roles to the generative AI and provides a means for users to respond to those roles in chat. Natural language processing technology is used to analyze the user's responses and check for harassment tendencies and bias. Specifically, NLTK's sentiment analysis tool (SentimentIntensityAnalyzer) is used to analyze the sentiment of the messages. Specific keywords are also counted to calculate scores for harassment and bias.

[0603] Furthermore, it provides a means to analyze the user's emotional state and adjust the AI's response based on the emotion. This allows it to provide appropriate feedback and support depending on the emotion the user expresses. For example, if the user expresses positive emotion, the AI ​​will respond by sharing its joy, and if the user expresses negative emotion, it will respond by promising improvement.

[0604] This also includes the ability to monitor employee communications in real time and provide appropriate feedback and support, which will enable early detection of harassment and bias in the workplace and contribute to improving employee mental health.

[0605] As a concrete example, the following prompt sentence can be used:

[0606] "Enter a user message. For example, 'I think this project is unfair.'"

[0607] This system enables early detection of harassment and bias in employee communications and appropriate responses based on the employee's emotional state.

[0608] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0609] Step 1:

[0610] The server assigns specific roles to the generative AI and provides an interface for users to respond to those roles in chat. As input, it receives messages entered by users in chat, and as output, the generative AI generates a response based on that role.

[0611] Step 2:

[0612] The server analyzes the user's response. Specifically, it uses natural language processing technology to analyze the user's message. It receives the user's message as input and calculates an emotion score, harassment score, and bias score as output. For sentiment analysis, it uses NLTK's SentimentIntensityAnalyzer, counting specific keywords and calculating a score.

[0613] Step 3:

[0614] The server feeds back the analysis results. It receives emotion scores, harassment scores, and bias scores as input, and generates an appropriate feedback message for the user as output. For example, in the case of a positive emotion score, it generates a message such as "We share your joy. Thank you!", and in the case of a negative emotion score, it generates a message such as "We apologize for any inconvenience caused. We will strive to improve."

[0615] Step 4:

[0616] The server analyzes the user's emotional state and adjusts the AI's response based on the emotion. It receives the user's emotion score as input and generates the AI's response based on the emotion as output. For example, if the user expresses positive emotion, the AI ​​will respond by sharing its joy, and if the user expresses negative emotion, it will respond by promising improvement.

[0617] Step 5:

[0618] The server monitors communication between employees in real time and provides appropriate feedback and support. It receives chat messages between employees as input and generates alerts and notifies managers if signs of harassment or bias are detected as output. It also generates and provides support messages to employees according to their emotional state.

[0619] Example 3

[0620] Next, a description will be given of a third embodiment of the third embodiment. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0621] With conventional systems, it was difficult to properly check for harassment tendencies or bias when users responded to specific roles. Furthermore, the system did not provide feedback that took into account the user's emotional state, which led to problems with users not being able to adequately improve their responses. Furthermore, there was also the issue of difficulty in providing appropriate support for employees who had taken leave due to mental health issues when returning to work.

[0622] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[0623] In this invention, the server includes a means for assigning specific roles to the generative AI and allowing the user to interactively respond to that role, a means for analyzing the response results and checking for harassment tendencies and bias, and a means for analyzing the user's emotional state using an emotion engine. This makes it possible to check for harassment and bias in the user's responses, and by providing feedback that takes the user's emotional state into consideration, it promotes improvements in the user's responses. It also enables appropriate support for employees who have taken leave due to mental health issues when they return to work.

[0624] "Generative AI" is artificial intelligence that generates natural language responses based on prompts entered by the user.

[0625] A "specific role" refers to a specific scenario or situation in which a user interacts with the system.

[0626] "User" refers to an individual or member of an organization who utilizes the system to fulfill a particular role.

[0627] "Dialogue format" refers to a format in which users interact with generative AI in natural language.

[0628] "Harassment" refers to inappropriate words, actions, or behavior toward others in the workplace or in society.

[0629] "Bias" refers to having prejudices or preconceived notions toward a particular individual or group.

[0630] An "emotion engine" refers to software or hardware that analyzes a user's emotional state and provides appropriate feedback based on the results.

[0631] "Feedback" refers to specific improvements and advice provided based on the user's response and emotional state.

[0632] "Mental health" refers to mental health conditions and includes mental health issues such as stress, anxiety, and depression.

[0633] "Leave of absence" refers to an employee temporarily leaving work for health reasons or other reasons.

[0634] "Return-to-work support" refers to the support and programs provided to employees who have been on leave when they return to work.

[0635] This invention is a system that allows users to interactively respond to specific roles, analyzes the results of those responses, checks for harassment tendencies and biases, and analyzes the user's emotional state using an emotion engine. This system is built by combining a generative AI model and an emotion engine.

[0636] The server assigns specific roles using a generative AI model (for example, an AI model that processes natural language). Users can chat with this AI via their devices in an interactive format, simulating how to respond to, for example, an employee returning from leave or a subordinate coming out as LGBT.

[0637] The device transmits the user's chat content to the server in real time. The server analyzes the received chat content to check for harassment tendencies and bias. It also analyzes the user's emotional state using an emotion engine (e.g., emotion recognition software).

[0638] The server then provides the user with feedback based on the analysis results. This feedback includes specific areas for improvement and advice. For example, if a user shows bias toward "subordinates who come out as LGBT," the server will present the specific details of that bias and suggestions for improvement.

[0639] As a concrete example, consider a case where a user simulates how to respond to an employee returning from leave. The user inputs the following prompt into the generative AI model:

[0640] How should I talk to an employee returning from leave?

[0641] When the user enters this prompt, the generative AI model suggests an appropriate response. The user chats based on the suggestion, and the server analyzes the response. As a result of the analysis, feedback is provided, such as "The user is providing appropriate support, but more specific advice is needed."

[0642] If a user expresses anger or dissatisfaction during chat, the emotion engine will recognize that emotion and the server will adjust the AI's response. For example, the AI ​​may respond to reduce the user's anger or send an apology.

[0643] In this way, the user can successively improve his / her own response through the system. The flow of the identification process in the third embodiment will be described with reference to FIG.

[0644] Step 1:

[0645] The user enters a prompt statement.

[0646] The user inputs a prompt to the generative AI model through the device interface, for example, "Please tell me how to greet an employee returning from leave."

[0647] Input: The prompt text entered by the user

[0648] Output: The prompt text displayed on the terminal

[0649] Step 2:

[0650] The terminal sends the prompt to the server.

[0651] The terminal sends the prompt text entered by the user to the server using a protocol such as an HTTP request.

[0652] Input: The prompt text entered by the user

[0653] Output: The prompt sent to the server

[0654] Step 3:

[0655] The server generates a response using a generative AI model.

[0656] The server inputs the received prompt sentence into a generative AI model (e.g., an AI model that performs natural language processing) to generate an appropriate response. The generative AI model generates a natural language response based on the prompt sentence.

[0657] Input: The prompt sent to the server

[0658] Output: The response generated by the generative AI model

[0659] Step 4:

[0660] The server generates a response and sends it to the terminal.

[0661] The server sends the response generated by the generative AI model to the terminal, again using a protocol such as HTTP response.

[0662] Input: The response generated by the generative AI model

[0663] Output: Response sent to the terminal

[0664] Step 5:

[0665] The user chats based on the generated response.

[0666] The user checks the generative AI model's response displayed on the device and continues the chat based on it. For example, if the generative AI model suggests, "First, send a message welcoming the user back," the user will follow that suggestion and chat.

[0667] Input: The response of the generative AI model displayed on the device

[0668] Output: Chat content of the user

[0669] Step 6:

[0670] The device sends the chat contents to the server.

[0671] The terminal transmits the contents of the chat that the user has conducted to the server in real time, allowing the server to grasp the user's responses successively.

[0672] Input: Chat content that the user has had

[0673] Output: Chat messages sent to the server

[0674] Step 7:

[0675] The server analyzes the chat content.

[0676] The server analyzes the chat content received and checks for harassment tendencies and bias. For example, if a user makes an inappropriate remark about a subordinate who has come out as LGBT, the server will detect that content.

[0677] Input: Chat content sent to the server

[0678] Output: Analysis results regarding harassment trends and bias

[0679] Step 8:

[0680] The server uses an emotion engine to analyze the user's emotional state.

[0681] The server uses an emotion engine (e.g., emotion recognition software) to analyze the user's emotional state, for example, recognizing when a user expresses anger or frustration during a chat.

[0682] Input: Chat content sent to the server

[0683] Output: Analysis results about the user's emotional state

[0684] Step 9:

[0685] The server generates the analysis results as feedback.

[0686] The server generates feedback for the user based on the analysis of the chat content and emotional state, including specific suggestions for improvement and advice.

[0687] Input: Analysis results on harassment trends and bias, and analysis results on the user's emotional state

[0688] Output: Feedback to the user

[0689] Step 10:

[0690] The server sends the feedback to the device.

[0691] The server then sends the generated feedback to the device, and the user receives this feedback, which provides specific guidelines for improving their own behavior.

[0692] Input: Feedback to the user

[0693] Output: Feedback sent to the terminal

[0694] Step 11:

[0695] Users receive feedback and improve their response.

[0696] Users can review the feedback displayed on their devices and use it to improve their own behavior, for example by taking specific actions to reduce bias toward subordinates who come out as LGBT.

[0697] Input: Feedback sent to the device

[0698] Output: User-improved response

[0699] (Application example 3)

[0700] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0701] Conventional systems could check for harassment tendencies and bias when users responded to specific roles, but they were unable to analyze the user's emotional state in real time and provide appropriate feedback based on those emotions. As a result, responses based on the user's emotional state were insufficient, and appropriate responses were particularly needed in emotionally charged situations. Furthermore, in collaborative work between humans and robots in factories, communication improvements that take into account the emotional state of workers were also necessary.

[0702] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[0703] In this invention, the server includes a means for assigning specific roles to generative AI and allowing users to respond to those roles via chat, a means for analyzing the results of the responses and checking for harassment tendencies and bias, a means for providing feedback on the analysis results, a means for using an emotion engine that analyzes the user's emotional state in real time, and a means for providing appropriate feedback according to the user's emotional state. This enables more appropriate responses according to the user's emotional state, making it possible to improve communication that takes into account the emotional state of workers, particularly in collaborative work between humans and robots in factories.

[0704] "Generative AI" is artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[0705] A "specific role" is a role that a user will play in a specific scenario or situation set within the system.

[0706] A "chat interaction" is an interface that allows users to interact with a particular role through text-based dialogue.

[0707] The "means for analyzing response results" is a function that analyzes the content of the responses made by users and evaluates whether there is a tendency for harassment or bias.

[0708] "Means to check for harassment tendencies and bias" is a function that determines whether harassment or prejudice is present based on the user's response results.

[0709] "Means for providing feedback on analysis results" is a function for providing specific improvements and advice to users based on the analysis of response results.

[0710] The "emotion engine" is a technology that analyzes the user's emotional state in real time and responds appropriately based on those emotions.

[0711] "Real-time analysis" means analyzing the user's emotional state instantly the moment they give input.

[0712] "Means for providing appropriate feedback" is a function that provides appropriate responses and advice depending on the user's emotional state.

[0713] "Collaboration between humans and robots in a factory" refers to a situation in which human workers and robots work together in a factory.

[0714] "Improved communication" refers to measures to make dialogue and the exchange of instructions between users and robots smoother, more efficient, and more effective.

[0715] The system for implementing this invention is a combination of generative AI, an emotion engine, and a feedback system. A specific embodiment of this system will be described below.

[0716] System configuration

[0717] The system consists of the following main components:

[0718] 1. Generative AI: Artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[0719] 2. Emotion Engine: Technology that analyzes the user's emotional state in real time.

[0720] 3. Feedback system: A function that analyzes the user's response results and provides specific improvements and advice.

[0721] Hardware and software used

[0722] Hardware: Smart glasses (e.g., Google® Glass®)

[0723] Software: OpenAI API, EmotionEngine, FeedbackSystem

[0724] System Operation

[0725] 1. User input: The user responds to a specific role through the chat via smart glasses.

[0726] 2. Emotion Analysis: The emotion engine analyzes user input in real time to identify their emotional state.

[0727] 3. Generative AI Response: Generative AI generates appropriate responses based on user input and emotional state.

[0728] 4. Providing feedback: The feedback system analyzes the user's response and provides specific improvements and advice.

[0729] Specific examples

[0730] For example, if a worker in a factory expresses dissatisfaction with a robot, the following prompt sentence may be generated:

[0731] User Input: This robot is completely useless!

[0732] Emotion: Anger

[0733] Please suggest an appropriate response.

[0734] This prompt can be fed into a generative AI model to generate an appropriate response and feedback. The generative AI takes into account the user's emotional state and generates a response such as, "Sorry. Can you tell me what the problem is?" The emotion engine suggests responses to alleviate the user's anger, and the feedback system provides specific improvements.

[0735] Data processing and calculation

[0736] Data processing: The user's input data is obtained in text format and analyzed using the emotion engine.

[0737] Data calculation: The emotion engine analyzes the emotional data and passes it to the generative AI to generate an appropriate response. The feedback system calculates specific improvements based on the generated response and user input.

[0738] In this way, it becomes possible to respond appropriately to the user's emotional state, and communication improvements that take into account the emotional state of workers can be realized, particularly in collaborative work between humans and robots in factories.

[0739] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[0740] Step 1:

[0741] Users respond to specific roles via chat through smart glasses.

[0742] Input: User text input (e.g., "This robot is completely useless!")

[0743] Output: User input text data

[0744] Specific behavior: A user uses the smart glasses interface to provide text input for a specific role, which is then sent to the system.

[0745] Step 2:

[0746] The server uses an emotion engine to analyze the user's input text data in real time and identify the user's emotional state.

[0747] Input: User-entered text data

[0748] Output: User's emotional state data (e.g., "Anger")

[0749] Specific operation: The server starts the emotion engine and analyzes the user's input text. The emotion engine extracts emotions from the text and outputs the results as emotional state data.

[0750] Step 3:

[0751] The server uses generative AI to generate an appropriate response based on the user's input text data and emotional state data.

[0752] Input: User input text data, emotional state data

[0753] Output: The generated response text (e.g. "Sorry, can you tell me what's wrong?")

[0754] How it works: The server calls the generative AI model and inputs the user's input text and emotional state as prompts. The generative AI then generates an appropriate response based on this and outputs it as response text.

[0755] Step 4:

[0756] The server uses a feedback system to analyze the user's response and provide specific improvements and advice.

[0757] Input: User input text data, generated response text

[0758] Output: Feedback text with suggestions for improvement

[0759] Specific operation: The server starts the feedback system, analyzes the user's input text and the generated response text, and based on this, generates specific improvements and advice for the user and outputs them as feedback text.

[0760] Step 5:

[0761] The server sends the generated response text and feedback text to the user.

[0762] Input: Generated response text, feedback text

[0763] Output: Responses and feedback displayed on the user's smart glasses

[0764] Specific operation: The server sends the generated response text and feedback text to the user's smart glasses so that the user can review them.

[0765] 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 a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the 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.

[0766] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0767] Another example of generative AI is Gemini (registered trademark) (Internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[0768] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0769] [Second embodiment]

[0770] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0771] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0772] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also 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).

[0773] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0774] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0775] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0776] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0777] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0778] The specific processing program 56 is an example of a "program" according to the technology of the present 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 in accordance with the specific processing program 56 executed on the RAM 30.

[0779] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0780] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0781] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[0782] "Example 1"

[0783] One embodiment of the present invention involves a means for assigning specific roles to generative AI. Specifically, roles such as "a subordinate coming out as LGBT," "a subordinate announcing a pregnancy," or "a subordinate struggling with work style" can be assigned to the AI. The user, as the supervisor, receives the consultation and then interacts with the AI ​​via chat.

[0784] "Example 2"

[0785] Next, there is a method to analyze user responses to check for harassment tendencies and bias. This analysis is done using natural language processing and machine learning techniques. Specifically, it determines whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc.

[0786] "Example 3"

[0787] Furthermore, there is a means to provide feedback on the analysis results. This feedback provides specific suggestions and advice to help users improve their own behavior. For example, if bias is found in the way employees treat subordinates who come out as LGBTQ+, the system will present the user with specific details of that bias and suggestions for improvement.

[0788] "Example 4"

[0789] The system of this invention can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. Specifically, to simulate how to respond when an employee returns from leave, the AI ​​is assigned the role of "an employee returning from leave," and the user responds via chat. The results are analyzed, and appropriate responses and return-to-work support programs are proposed.

[0790] The processing flow of each embodiment will be described below.

[0791] "Example 1"

[0792] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "a subordinate who comes out as LGBT."

[0793] Step 2: As a manager, the user responds to the subordinate's coming out via chat.

[0794] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0795] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0796] "Example 2"

[0797] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "subordinate who reports pregnancy."

[0798] Step 2: As a supervisor, the user responds to the subordinate's pregnancy report via chat.

[0799] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0800] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0801] "Example 3"

[0802] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of a "subordinate who is worried about their work style."

[0803] Step 2: As a supervisor, the user responds to the subordinate's concerns via chat.

[0804] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[0805] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0806] "Example 4"

[0807] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of an "employee returning from leave."

[0808] Step 2: As a manager, the user responds to the employee's return via chat.

[0809] Step 3: AI analyzes the user's response and suggests appropriate responses and programs to support their return to work.

[0810] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[0811] Example 1

[0812] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0813] In conventional systems using generative AI, when users respond to specific roles in chat, the system does not adequately generate appropriate prompts or display responses, resulting in a poor user experience. Furthermore, the system does not analyze the response results or check for harassment tendencies or bias, resulting in insufficient feedback. This limits users' opportunities to learn appropriate responses.

[0814] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0815] In this invention, the server includes: means for accepting a user's selection from multiple scenarios; means for setting a specific role based on the scenario selected by the user and generating a prompt sentence instructing how to respond to the role; means for accepting a user's input message; means for generating a response to the user's input message using the generated prompt sentence and a generative AI; means for outputting the generated response to the user; means for repeatedly accepting the input message, generating the response, and outputting the application; a prompt sentence instructing the user to check for harassment tendencies or bias in the user's responses based on a history of the input messages and responses; and means for evaluating the user's responses using the generative AI. This allows the user to learn how to respond to specific roles in real time and receive appropriate feedback. The multiple scenarios include a user interacting with a subordinate who comes out as a sexual minority, a user interacting with a subordinate who announces pregnancy, and a user interacting with a subordinate who is concerned about their work style.

[0816] "Generative AI" is an artificial intelligence system that generates natural language responses based on user input.

[0817] A "specific role" is a specific character or situation that the user sets based on a scenario and that the generative AI will imitate.

[0818] A "user" is a person who uses the system to chat with generative artificial intelligence.

[0819] A "prompt" is text input to a generative artificial intelligence (AI) that instructs it to generate a response based on a specific role.

[0820] A "response" is a text message that the generative artificial intelligence generates based on the prompt sentence.

[0821] "Chat" is an interactive form of communication between a user and a generative artificial intelligence via text messages.

[0822] A "scenario" is a setting containing specific situations and roles selected by the user, which serves as the basis for the generative AI to generate responses.

[0823] "Harassment tendencies" refers to a pattern of inappropriate speech and behavior toward others as a result of a user's responses.

[0824] "Presence or absence of bias" indicates whether or not a particular prejudice or preconception is included in the user's response results.

[0825] "Feedback" is information for evaluation and improvement provided based on the user's response results.

[0826] The present invention is a system that uses generative artificial intelligence to set specific roles and allows users to respond to those roles through chat. A specific embodiment of this system will be described below.

[0827] First, the user selects one of several scenarios through the device interface. For example, scenarios are displayed in which the user talks to a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style.

[0828] The server generates a prompt based on the scenario selected by the user. Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4 (registered trademark)) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to that role. For example, it generates a prompt that reads, "The user is your boss. You are a subordinate who is consulting with him about coming out as a sexual minority. Please generate a response to the user's input message so as to reproduce this dialogue."

[0829] Next, the server inputs the generated prompt and the user's input message into the generative AI to generate a response. For example, the generated response might be, "Boss, could you please give me a moment? Actually, there's something I've been wanting to say for a while."

[0830] The server generates a response and sends it to the terminal, which displays it to the user. The user sees the response and types in a new message, such as "Sure, what's wrong?"

[0831] The server then inputs the user's input message back into the generative AI to generate the next response, such as, "Actually, I want to come out as a sexual minority, but I don't know how to say it..."

[0832] This process is repeated, and the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and respond to a subordinate's consultation.

[0833] Furthermore, the server analyzes the user's response based on the history of input messages and responses, and evaluates whether there is a tendency for harassment or bias. This allows the server to evaluate how the user responded and provide appropriate feedback.

[0834] As a specific example, the prompt sentence when the user selects the scenario "a subordinate who comes out as a sexual minority" is shown below.

[0835] Example prompt sentence:

[0836] "The user is your manager. You are a subordinate who is asking for advice on coming out as a LGBTQ person. Please generate responses to the user's input messages to recreate this dialogue. You are nervous and wondering how to proceed. The subordinate's initial message is as follows: "Boss, can I have a moment? There's actually something I've been wanting to tell you for a while."

[0837] In this way, users can learn how to respond to specific roles in real time and receive appropriate feedback.

[0838] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0839] Step 1:

[0840] The user selects a scenario.

[0841] The user selects one of several scenarios through the terminal interface. For example, scenarios are displayed in which the user interacts with a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style. The scenario selected by the user is sent to the server (input). The server receives this scenario information and proceeds to the next process (output).

[0842] Step 2:

[0843] The server generates a prompt.

[0844] The server generates a prompt based on the scenario selected by the user (input). Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to the role (output).

[0845] Step 3:

[0846] The server inputs a prompt sentence into the generative AI model.

[0847] The server inputs the generated prompt sentence into the generative AI (input). The generative AI generates a response based on this prompt sentence. For example, a response such as "Boss, can you give me a moment? Actually, there's something I've been wanting to say for a while" is generated (output).

[0848] Step 4:

[0849] The server sends the generated response to the terminal.

[0850] The server sends the response obtained from the generative AI to the terminal (input), which receives the response and displays it to the user (output).

[0851] Step 5:

[0852] The terminal displays the response to the user.

[0853] The terminal displays the response received from the server to the user (input), who then looks at it and prepares to enter the next message (output).

[0854] Step 6:

[0855] The user enters the following message:

[0856] The user responds to the displayed response by typing the next message (input). For example, "Of course, what's wrong?" The user's input message is sent to the server via the terminal (output).

[0857] Step 7:

[0858] The server inputs the user's message into a generative AI model.

[0859] The server inputs the message entered by the user into the generative AI (input). The generative AI generates the following response based on this message. For example, it might generate a response like, "Actually, I want to come out as a sexual minority, but I don't know how to say it..." (output).

[0860] Step 8:

[0861] The server sends the generated response to the terminal.

[0862] The server sends the next response obtained from the generative AI to the terminal (input). The terminal receives this response and displays it to the user (output).

[0863] Step 9:

[0864] The terminal displays the response to the user.

[0865] The terminal displays the response received from the server to the user again (input), who then looks at it and prepares to input the next message (output).

[0866] Step 10:

[0867] This process is repeated.

[0868] This process is repeated as long as the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and provide advice to their subordinates (input and output). The server may further include means for recognizing the user's emotions using an emotion engine described below, and when generating a response, may use the recognized user's emotions, the generated prompt sentence, and the generative AI to generate a response to the user's input message.

[0869] (Application example 1)

[0870] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0871] Conventional systems using generative AI allow users to set specific roles and practice responses for those roles, but they lack training for the specific issues that corporate security personnel face, such as security issues, whistleblowing, and suspected information leaks. Furthermore, the system has limited functionality for analyzing the results of these responses to check for harassment tendencies and bias, and the quality of the feedback is insufficient. Furthermore, it does not take into consideration the use of the system as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. This means that corporate security personnel and managers are insufficiently trained to respond appropriately, potentially increasing security risks across the entire company.

[0872] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0873] In this invention, the server includes a means for assigning specific roles to the generative AI and allowing users to respond to those roles via chat; a means for analyzing the response results and checking for harassment tendencies and bias; a means for providing feedback on the analysis results; a means for simulating security issues and allowing personnel to practice their responses via chat; and a means for generating prompts using the generative AI model and providing appropriate responses. This allows corporate security personnel and managers to receive training on how to respond appropriately to security issues, whistleblowers, suspected information leaks, and other issues, thereby reducing security risks across the company. It can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues, contributing to employee mental health care.

[0874] "Generative AI" is artificial intelligence that generates natural language based on specific roles and scenarios and engages in dialogue with users.

[0875] "Specific roles" refer to specific scenarios or characters that are set in the generative AI, such as "a subordinate coming out as LGBT" or "a subordinate announcing a pregnancy."

[0876] "Methods via chat" refers to interfaces and functions that allow users to have text-based conversations with generative AI.

[0877] "Means for analyzing the response results" refers to algorithms and software that analyze the content of the dialogue between the user and the generative AI and evaluate whether there is any tendency toward harassment or bias.

[0878] "Means to check for harassment tendencies and bias" refers to a function that determines whether a user's response contains harassment or bias based on the content of the conversation.

[0879] "Means of providing feedback" refers to functions and interfaces that provide analysis results to users and indicate areas for improvement or attention.

[0880] "Means to simulate security issues" refers to a feature that allows users to set security-related scenarios, such as information leaks and whistle-blowing, in the generative AI and practice how to respond to them.

[0881] A "generative AI model" is a machine learning model for generating natural language and generating appropriate responses based on specific prompts.

[0882] A "prompt sentence" is an input sentence that sets a specific role or scenario for generative AI, and serves as a trigger to start a dialogue with the user.

[0883] A system for implementing this invention includes means for setting a specific role for a generative AI and for the user to respond to that role via chat, means for analyzing the results of that response and checking for harassment tendencies or bias, means for feeding back the results of the analysis, means for simulating security issues and allowing personnel to practice their responses via chat, and means for generating prompt sentences using a generative AI model and providing appropriate responses.

[0884] System program generation

[0885] The server assigns specific roles to the generative AI and provides an interface for users to respond to those roles via chat. The text entered by the user is sent to the generative AI model on the server, which generates an appropriate response. The generated response is returned to the user, and the dialogue progresses.

[0886] Processing Description

[0887] The server performs processing using the following hardware and software:

[0888] Hardware: Internet-connected server

[0889] Software: Python, OpenAI API

[0890] The server receives input text from the user and sends it to the generative AI model. The generative AI model generates an appropriate response based on the prompt. The generated response is returned to the user via the server. The server then analyzes the content of the dialogue to check for harassment tendencies and bias. The analysis results are then fed back to the user.

[0891] Specific examples

[0892] For example, to simulate a security officer responding appropriately to a call from an employee reporting a suspected data breach, the following prompt might be used:

[0893] You are an employee reporting a suspected data breach. Please respond appropriately to the user's inquiry.

[0894] User: I recently noticed some confidential information from my company has been leaked. What should I do?

[0895] you:

[0896] Using this prompt, a generative AI model can generate appropriate responses that security personnel can practice, such as:

[0897] You: That's a very important issue. First, please provide specific evidence and details of the situation. Also, I need to report this matter to my superiors and the security department immediately. A potential information leak requires a quick response, so could you please provide more details?

[0898] In this way, security personnel can be trained to respond appropriately to actual inquiries.

[0899] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0900] Step 1:

[0901] A user accesses the chat interface using a terminal and enters a prompt sentence based on a specific role. The entered prompt sentence is sent to the server. The input data is the user's text input, and the output data is the prompt sentence sent to the server.

[0902] Step 2:

[0903] The server sends the received prompt to the generative AI model, which generates an appropriate response based on the prompt. The input data is the prompt, and the output data is the generated response. Specifically, the server calls the OpenAI API and sends the prompt.

[0904] Step 3:

[0905] The server receives the response generated by the generative AI model and returns it to the user's device. The input data is the response from the generative AI model, and the output data is the response sent to the user's device. Specifically, the server displays the generated response on the user's chat interface.

[0906] Step 4:

[0907] The user confirms the generated response and provides further input. The user's new input is again sent to the server, and the process repeats from step 2. The input data is the user's new text input, and the output data is the new prompt sent to the server.

[0908] Step 5:

[0909] The server analyzes the dialogue between the user and the generated AI model in real time to check for harassment tendencies and bias. The input data is the dialogue content, and the output data is the analysis results. Specifically, the server analyzes the dialogue content using a natural language processing algorithm.

[0910] Step 6:

[0911] The server feeds back the analysis results to the user. The input data is the analysis results, and the output data is the feedback provided to the user. In specific operations, the server displays the analysis results in text format on the user's chat interface.

[0912] Step 7:

[0913] The server simulates security issues and generates new prompts for users to practice their responses in chat. The input data is the security scenario settings, and the output data is the new prompts. Specifically, the server generates new prompts using a generative AI model.

[0914] Step 8:

[0915] The user then practices responding to the new prompt sentence again in the chat interface. The input data is the user's text input for the new prompt sentence, and the output data is the new prompt sentence sent to the server. As a specific operation, the user responds to the new scenario using the chat interface.

[0916] Example 2

[0917] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0918] With the previous system, it was difficult to properly analyze the results of user responses and check for harassment tendencies or bias. Furthermore, there was an insufficient means of providing feedback to users about the results of the analysis, making it difficult to lead to improvements in user behavior. Furthermore, there was a lack of specific methods for evaluating employee responses in specific situations.

[0919] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0920] In this invention, the server includes a means for a user to respond to a specific role via chat, a means for collecting the response results, a means for preprocessing the collected text data, a means for inputting the preprocessed text data into a machine learning model to analyze the presence or absence of harassment trends and bias, and a means for feeding back the analysis results. This makes it possible to appropriately analyze the user's response results and check for the presence or absence of harassment and bias. Furthermore, by feeding back the analysis results to the user, it is possible to promote improvement of the user's behavior. Furthermore, it is possible to provide a specific means for evaluating an employee's response in a specific situation.

[0921] "User" means an individual or entity that utilizes the System and fulfills a particular role.

[0922] A "specific role" refers to a specific situation or scenario that a user is required to handle within the system.

[0923] "Chat" is an interface that allows users to communicate text-based.

[0924] "Response results" are records of responses and actions taken by users for specific roles.

[0925] "Means for collection" refers to a method or device for acquiring and storing the results of a user's response as data.

[0926] "Text data" is data that expresses the user's response results as text information.

[0927] The "preprocessing means" refers to a method or device that processes collected text data to convert it into an analyzable format.

[0928] A "machine learning model" is an algorithm or system that learns patterns and trends from data and makes predictions and classifications.

[0929] "Means for analysis" refers to a method or device that uses preprocessed text data to evaluate the presence or absence of harassment tendencies or bias.

[0930] "Feedback means" refers to a method or device for communicating the analysis results to the user.

[0931] "Harassment" refers to inappropriate words, actions or behaviour towards others.

[0932] "Bias" refers to making judgments or taking actions based on specific prejudices or preconceptions.

[0933] This invention is a system in which users respond to specific roles via chat, analyze the results of their responses to check for harassment tendencies and bias, and provide feedback on the analysis results. A specific embodiment of this system is described below.

[0934] First, the server provides an interface for users to respond to specific roles in chat. This interface can be accessed through devices such as web browsers and mobile applications. For example, a user can be assigned a role as an "employee in a specific situation," and respond in chat based on that role.

[0935] The server then collects the user's responses, which are then stored in a database as text data. The collected text data is preprocessed using a natural language processing (NLP) library (e.g., spaCy or NLTK). Preprocessing includes tokenization, stop word removal, stemming, etc.

[0936] The preprocessed text data is input into a machine learning model (e.g., scikit-learn or TensorFlow). The server uses this machine learning model to analyze the text data and determine whether there is any harassment tendency or bias. Specifically, it evaluates whether specific keywords, phrases, or tones are offensive.

[0937] The analysis results are stored in a database by the server. In addition, the server provides feedback to the user. For example, if a message sent by a user is judged to be offensive, the server will provide feedback such as, "Your message contains offensive language. Please try to use more polite language."

[0938] As a concrete example, consider the case where a user sends a message saying, "Your response is completely useless. You're incompetent." In this case, the server collects this message and preprocesses it using an NLP library. Then, using a machine learning model, it determines that this message is offensive. Finally, the server provides appropriate feedback to the user.

[0939] An example of a prompt might be:

[0940] "Please generate a program that analyzes user responses and checks for harassment tendencies and bias. Specifically, use natural language processing and machine learning techniques to determine whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc."

[0941] In this way, the server can properly analyze the user's response and check for harassment or bias. Furthermore, by providing feedback to the user, the analysis results can encourage the user to improve their behavior. Furthermore, the server can provide a concrete means for evaluating employee responses in specific situations.

[0942] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0943] Step 1:

[0944] The server provides an interface for users to respond to specific roles on the chat. Users access this interface through their terminals and start chats based on specific roles. The input is the user's chat message, and the output is the text data of that message.

[0945] Step 2:

[0946] The server collects users' chat messages and stores them in a database. The input is the user's chat messages, and the output is the stored text data. Specifically, the server captures messages in real time and writes them to the database.

[0947] Step 3:

[0948] The server preprocesses the collected text data. Specifically, it uses an NLP library (e.g., spaCy or NLTK) to perform the following processes: tokenization, stop word removal, and stemming. The input is the stored text data, and the output is the preprocessed text data. Specifically, the server splits the text data into tokens, removes unnecessary words, and extracts word stems.

[0949] Step 4:

[0950] The server inputs the preprocessed text data into a machine learning model for analysis. Specifically, it uses libraries such as scikit-learn and TensorFlow to determine whether there is any harassment tendency or bias. The input is the preprocessed text data, and the output is the analysis results. Specifically, the server inputs the text data into the model and evaluates specific keywords, phrases, and tone.

[0951] Step 5:

[0952] The server stores the analysis results in a database. The input is the analysis results, and the output is the stored analysis results. As a specific operation, the server writes the analysis results into the database.

[0953] Step 6:

[0954] The server provides feedback to the user. The input is the saved analysis result, and the output is a feedback message to the user. Specifically, the server generates a feedback message based on the analysis result and sends it to the user. For example, it provides a message such as, "Your message contains offensive language. Please try to use more polite language."

[0955] In this way, the server can properly analyze the user's response results, check for harassment or bias, and provide feedback.

[0956] (Application example 2)

[0957] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0958] In today's corporate environment, harassment and bias issues are serious, and there is a need to detect these issues early and respond appropriately. However, traditional methods make it difficult to detect these issues in real time and respond immediately. In addition, signs of harassment and bias are often overlooked, which can have a negative impact on the company's credibility and the mental health of employees. Therefore, effective measures to resolve these issues are needed.

[0959] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0960] In this invention, the server includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies and bias, means for feeding back the analysis results, means for analyzing chat and email content in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate responses to users. This makes it possible to detect harassment and bias issues in real time and respond immediately.

[0961] "Generative AI" is an artificial intelligence system that is assigned a specific role and interacts with the user based on that role.

[0962] A "specific role" refers to a specific scenario or character that is set in the generative AI, and is the role that the user responds to.

[0963] "User responses" refer to chat interactions and reactions that users have with specific roles assigned to generative AI.

[0964] "Response results" refer to the content of the interactions and reactions the user had with the generative AI.

[0965] "Harassment" refers to inappropriate words, actions, or behavior toward others, and is problematic behavior in the workplace and in society.

[0966] "Bias" refers to judgments or actions based on specific prejudices or preconceptions, and is a factor that leads to a lack of fairness.

[0967] "Means of checking" refers to methods and techniques for analyzing the results of users' responses and determining whether or not there is harassment or bias.

[0968] "Feedback means" refers to methods and techniques for notifying users and administrators of the analysis results and encouraging them to take appropriate action.

[0969] "Real-time analysis means" refers to methods and technologies for instantly analyzing chat and email content to detect signs of harassment or bias.

[0970] "Alert methods" are methods or technologies used to alert administrators when signs of harassment or bias are detected.

[0971] "Means for storing data in a database" refers to methods and technologies for recording the nature of any detected harassment or bias, making it available for later analysis and reference.

[0972] "Means for suggesting appropriate responses" are methods and techniques for suggesting appropriate responses to harassment and bias issues to users.

[0973] A system for implementing this invention includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies or bias, means for feeding back the analysis results, means for analyzing the content of chats and emails in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate ways of responding to users.

[0974] System configuration

[0975] The server uses a generative AI model to set specific roles, and users can respond to those roles in the chat. The user's responses are analyzed using a natural language processing library (e.g., spaCy). The analyzed data is then checked for harassment and bias using a machine learning model (e.g., scikit-learn's Logistic Regression).

[0976] Data processing and calculation

[0977] The server analyzes the content of users' chats and emails in real time. The analyzed data is converted into TF-IDF vectors and input into a machine learning model. The model detects signs of harassment and bias and sends the results as an alert to an administrator. The detected content is also stored in a database for later detailed analysis.

[0978] Hardware and software used

[0979] Hardware: Servers, user devices (PCs, smartphones, etc.)

[0980] Software: Generative AI model, natural language processing library (spaCy), machine learning library (scikit-learn), database management system

[0981] Specific examples

[0982] For example, if a company were to use this system for customer support chat, they would input the following prompts into the generative AI model:

[0983] Analyze customer support chats in real time to detect signs of harassment and bias. Analyze messages such as:

[0984] "You're really incompetent."

[0985] "Thank you for your amazing support!"

[0986] "You're an idiot."

[0987] Based on this prompt, the generative AI model analyzes the chat content to detect signs of harassment or bias, and if detected, sends an alert to an administrator and suggests appropriate responses.

[0988] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0989] Step 1:

[0990] The server assigns specific roles to the generative AI model. Specifically, it assigns scenarios such as "a subordinate coming out as LGBT" or "a subordinate announcing pregnancy" to the model. Based on this assignment, it generates a prompt for the user to start a chat.

[0991] Input: Specific Role

[0992] Output: prompt statement

[0993] Step 2:

[0994] Users respond to specific roles assigned by the generative AI model through chat, and their input is sent to the server as chat messages.

[0995] Input: User's chat message

[0996] Output: Message sent to the server

[0997] Step 3:

[0998] The server analyzes the received chat messages using a natural language processing library (e.g., spaCy), and performs data processing such as text tokenization and TF-IDF vectorization on the analyzed data.

[0999] Input: User's chat message

[1000] Output: Parsed text data (tokenized, vectorized)

[1001] Step 4:

[1002] The server then feeds the parsed text data into a machine learning model (e.g., scikit-learn's LogisticRegression) to check for harassment and bias. The model makes predictions based on the pre-trained data.

[1003] Input: Parsed text data

[1004] Output: Prediction results of harassment and bias

[1005] Step 5:

[1006] Based on the prediction results, the server sends an alert to the administrator if signs of harassment or bias are detected. The alert includes the content of the detected message and the prediction results.

[1007] Input: Predicted results of harassment and bias

[1008] Output: Alert to administrator

[1009] Step 6:

[1010] The server stores any detected instances of harassment or bias in a database, which is then used for further analysis and reporting.

[1011] Input: Detected harassment or bias

[1012] Output: Data stored in the database

[1013] Step 7:

[1014] The server then proposes appropriate responses to the user, based on generative AI models and pre-defined response guidelines.

[1015] Input: Predicted results of harassment and bias

[1016] Output: Suggestions for the user on how to respond

[1017] Example 3

[1018] Next, a description will be given of Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1019] With conventional systems, it was difficult to effectively detect bias or harassment tendencies in users' responses and provide specific improvements. Furthermore, when supporting employees returning to work after taking a leave of absence due to mental health issues, there was a lack of means to simulate appropriate responses and provide feedback. This limited opportunities for users to learn and improve appropriate responses.

[1020] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1021] In this invention, the server includes a means for the user to input a prompt sentence, a means for the terminal to send the prompt sentence to the server, a means for the server to execute the generative AI model, a means for the server to analyze the results of the generative AI model, a means for the server to generate feedback, and a means for the terminal to display the feedback to the user. This makes it possible to effectively detect bias or harassment tendencies in the user's behavior and provide specific improvements and advice. It is also possible to simulate appropriate response methods and provide feedback to support the return of employees who have taken leave due to mental health issues.

[1022] A "prompt" is the text of an instruction or question that a user enters into a system.

[1023] A "terminal" is a device on which a user inputs a prompt and communicates with a server. Examples include PCs, smartphones, and tablets.

[1024] A "server" is a computer system that runs the generative AI model, analyzes prompts, and generates feedback.

[1025] A "generative AI model" is an artificial intelligence model that simulates a response based on a user's prompt and generates the results.

[1026] "Feedback" refers to specific improvements and advice provided to users after analyzing the results of the generative AI model.

[1027] "Bias" refers to prejudices or preconceived notions contained in user responses.

[1028] "Harassment" refers to inappropriate words, actions, or behavior in a user's interactions with others.

[1029] "Mental health" refers to the state of mental health and is a concept that includes psychological problems such as stress, anxiety, and depression.

[1030] A "return-to-work program" is a plan or procedure for assisting employees who have taken leave due to mental health issues to return to work.

[1031] This invention is a system in which a user inputs a prompt sentence, executes a generative AI model based on the prompt sentence, and provides feedback. A specific embodiment of this system will be described below.

[1032] First, the user inputs a prompt using a terminal. A terminal is a device such as a PC, smartphone, or tablet that provides an interface for users to access the system. For example, a user can input a prompt like the following:

[1033] "How should I respond to a subordinate who comes out as LGBT?"

[1034] How should I respond to an employee returning from leave?

[1035] The device sends the prompts entered by the user to a server, which has the hardware and software to run the generative AI model, specifically, a server equipped with an NVIDIA GPU and a deep learning framework such as TensorFlow or PyTorch.

[1036] The server inputs the received prompt sentence into a generative AI model, which then simulates the user's response and generates the results. This generative AI model then analyzes the prompt sentence using natural language processing technology and suggests an appropriate response.

[1037] The output of the generative AI model is analyzed by a server, which uses natural language processing techniques, such as sentiment analysis and keyword extraction, to analyze the results of the generative AI model and detect bias or harassment in user responses.

[1038] The server generates feedback based on the analysis results, including specific areas for improvement and advice. For example, it might say, "It's important to respond to subordinates who come out as LGBT without prejudice. Specifically, we recommend that you listen carefully to what your subordinates have to say and show understanding."

[1039] Finally, the device displays the feedback received from the server to the user, who can then review the feedback on the device screen and use it as a reference for improving their own response.

[1040] In this way, this invention can effectively detect bias and harassment tendencies in user behavior and provide specific improvements and advice. It can also simulate appropriate response methods and provide feedback when supporting the return of employees who have taken leave due to mental health issues. The flow of the identification process in Example 3 will be explained using Figure 15.

[1041] Step 1:

[1042] The user enters a prompt statement.

[1043] The user inputs a prompt through the device interface. For example, they might input, "Please tell me how to respond to a subordinate who comes out as LGBT." The input prompt is temporarily stored in the device's memory.

[1044] Step 2:

[1045] The terminal sends the prompt to the server.

[1046] The terminal converts the prompt text entered by the user into JSON format and sends it to the server using an HTTP POST request. The destination endpoint is the server's API endpoint. The input is the prompt text, and the output is the request sent to the server.

[1047] Step 3:

[1048] The server runs the generative AI model.

[1049] The server performs data preprocessing on the received prompt sentences in order to input them into the generative AI model. Specifically, it tokenizes and normalizes the text. It then runs the generative AI model to simulate the user's response. The input is the preprocessed prompt sentence, and the output is the simulation result of the generative AI model.

[1050] Step 4:

[1051] The server analyzes the results of the generated AI model.

[1052] The server uses natural language processing techniques to analyze the output of the generative AI model. For example, it performs sentiment analysis of the text and keyword extraction to detect bias in the user's responses. The input is the simulation result of the generative AI model, and the output is the analysis result.

[1053] Step 5:

[1054] The server generates the feedback.

[1055] The server generates feedback based on the analysis results. The feedback includes specific advice and suggestions for improving the user's response. The generated feedback is sent to the terminal in JSON format. The input is the analysis results, and the output is the feedback.

[1056] Step 6:

[1057] The device displays the feedback to the user.

[1058] The device uses an appropriate UI component to display the feedback received from the server to the user. For example, it displays the feedback in a popup window or as a notification. The user can check the feedback and use it to improve their own response. The input is the feedback, and the output is what is displayed to the user.

[1059] (Application example 3)

[1060] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1061] There are issues with responding appropriately when employees have mental health issues or in specific situations. In particular, there is a lack of simulations and feedback for machines working in factories to respond appropriately to employees, which can delay improvements in employee mental health and the working environment.

[1062] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 3 is realized by the following means. In this invention, the server includes means for setting a specific role for the generative AI and for the user to respond to that role via chat, means for analyzing the response results and checking for harassment tendencies and bias, means for feeding back the analysis results, means for providing a simulation for machines working in a factory to respond appropriately to employees, and means for analyzing the simulation results and providing improvements and advice. This makes it possible to simulate and provide feedback for appropriate responses when employees have mental health issues or in specific situations.

[1063] "Generative AI" is an artificial intelligence system that assigns specific roles and allows users to respond to those roles.

[1064] "Chat" is a text-based means of communication that allows users to interact with generative AI.

[1065] "Harassment" refers to inappropriate behavior or language directed at an individual or group in the workplace or other environment.

[1066] "Bias" refers to prejudices or preconceived notions held against a particular individual or group.

[1067] "Feedback" is a means of providing evaluations and suggestions for improvement of actions taken by users.

[1068] "Factory machines" refers to automated devices and robots used to perform tasks in a factory.

[1069] "Simulation" refers to the process of virtually recreating a specific situation or role and allowing users to respond within that situation.

[1070] "Mental health" refers to an individual's mental well-being and includes issues such as stress, anxiety and depression.

[1071] "Employee" refers to people who work for a particular company or organization.

[1072] "Areas for improvement" refers to the actions taken by the user that need to be corrected or changed to achieve better results.

[1073] "Advice" refers to specific instructions or suggestions to help users take more appropriate action.

[1074] To implement this invention, it is first necessary to build a system in which specific roles are assigned to generative AI and users respond to those roles via chat. This system has the function of analyzing the user's responses and checking for harassment tendencies and biases. It also includes a function to provide feedback on the analysis results.

[1075] Furthermore, a function that provides simulations to help machines working in factories respond appropriately to employees is also required. The results of these simulations are analyzed and suggestions and advice are provided to improve employee mental health and the working environment.

[1076] Hardware and software used

[1077] Hardware: Automated equipment and robots used in factories (e.g., ABB YuMi, Fanuc CR series)

[1078] Software: OpenAI API, Python

[1079] Data processing and calculation

[1080] server:

[1081] 1. Receive input from the user.

[1082] 2. Send prompts to a generative AI model (OpenAI's GPT-3) to generate a robotic response.

[1083] 3. The generated response is sent back to the generative AI model to analyze the response and generate feedback.

[1084] Device:

[1085] It provides an interface that allows users to interact with generative AI through chat.

[1086] User:

[1087] Responses to specific roles are made via chat and the results are sent to the system.

[1088] Specific examples

[1089] For example, if an employee enters, "I recently took a leave of absence due to work stress. I'm worried about returning to work," the system will act as follows:

[1090] 1. The server receives input from the user and sends prompts to the generative AI model.

[1091] 2. The generative AI model generates a response based on the following prompt:

[1092] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[1093] robot:

[1094] 3. The server sends the generated response back to the generative AI model to analyze the response and generate feedback.

[1095] 4. Feedback is provided to users and appropriate actions and improvements are suggested.

[1096] In this way, simulations and feedback can be provided to enable machines working in factories to respond appropriately to employees.

[1097] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1098] Step 1:

[1099] The server receives input from the user. Specifically, it receives text data entered by the user through the chat interface. This input may include, for example, "I recently took a leave of absence due to work stress. I'm worried about returning to work." The input data is sent to the server in text format.

[1100] Step 2:

[1101] The server sends the received user input to the generative AI model. Specifically, it uses OpenAI's API to send the following prompt sentence to the generative AI model:

[1102] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[1103] robot:

[1104] Based on this prompt, the generative AI model generates an appropriate response. The input data is sent in the form of a prompt, and the output data is the generated response text.

[1105] Step 3:

[1106] The server receives a response from the generative AI model. Specifically, it receives text data returned by the generative AI model. This response may include, for example, "When you return to work, try setting small goals first." The input data is the response text from the generative AI model, and the output data is that response text.

[1107] Step 4:

[1108] The server sends the generated response back to the generative AI model to analyze the response and generate feedback. Specifically, it sends the following prompt sentence to the generative AI model:

[1109] Please analyze the responses below and offer your suggestions and suggestions for improvement.

[1110] Response: When you return, try setting small goals first.

[1111] feedback:

[1112] Based on this prompt, the generative AI model analyzes the response and generates feedback. The input data is the response text from the generative AI model, and the output data is the feedback text.

[1113] Step 5:

[1114] The server receives feedback from the generative AI model. Specifically, it receives the feedback text returned by the generative AI model. This feedback includes, for example, "This response is appropriate, but it would be good to suggest specific support methods as well." The input data is the feedback text from the generative AI model, and the output data is the feedback text.

[1115] Step 6:

[1116] The server provides the generated feedback to the user. Specifically, the server displays the feedback text in the user's chat interface. Based on this feedback, the user can improve their own response. The input data is the feedback text, and the output data is the feedback displayed to the user.

[1117] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1118] "Example 1"

[1119] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[1120] "Example 2"

[1121] In another embodiment of the present invention, the emotion engine performs AI based on the user's emotional state.

[1122] There is a system that adjusts the AI's response to the user's emotions. In this system, when a user expresses joy or satisfaction to the AI, the emotion engine recognizes the emotion and adjusts the AI's response. For example, the AI ​​may respond in a way that shares the user's joy or send a message of gratitude to the user. This allows for more appropriate responses according to the user's emotional state.

[1123] "Example 3"

[1124] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[1125] The processing flow of each embodiment will be described below.

[1126] "Example 1"

[1127] Step 1: The user initiates a chat conversation with the AI.

[1128] Step 2: The AI ​​interacts with the user based on the assigned role.

[1129] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[1130] Step 4: The emotion engine analyzes the user's emotional state.

[1131] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[1132] "Example 2"

[1133] Step 1: The user initiates a chat conversation with the AI.

[1134] Step 2: The AI ​​interacts with the user based on the assigned role.

[1135] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[1136] Step 4: The emotion engine analyzes the user's emotional state.

[1137] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[1138] Example 1

[1139] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1140] Conventional workplace communication and harassment countermeasures are difficult to simulate in real situations and lack the means to properly grasp users' emotional states. This limits opportunities for supervisors and managers to learn appropriate responses, making it easy for harassment and bias to occur. Furthermore, the lack of feedback based on users' emotional states makes it difficult to improve response skills in the real workplace.

[1141] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1142] In this invention, the server includes a means for assigning specific roles to generative artificial intelligence and allowing users to respond to those roles via communication means, a means for analyzing the results of the responses and checking for harassment tendencies and bias, a means for analyzing the user's emotional state using an emotion engine, and a means for providing feedback on the analysis results. This allows users to improve their ability to respond in the actual workplace and prevent problems such as harassment and bias. Furthermore, providing appropriate feedback according to the user's emotional state enables more effective communication.

[1143] "Generative AI" is AI that can generate natural language based on user input and behave as a character with a specific role.

[1144] "Specific roles" refer to specific scenarios or character roles that are set for generative AI, such as "a subordinate who comes out as a sexual minority" or "a subordinate who announces pregnancy."

[1145] "Communication means" refers to the interface through which users interact with generative AI, including chat applications and messaging platforms.

[1146] "Harassment tendencies" are patterns that indicate a user's comments or actions may be inappropriately pressuring or harassing others.

[1147] "Bias" refers to prejudices and preconceived notions contained in users' statements and actions, and indicates unfair treatment of people with certain attributes or backgrounds.

[1148] An "emotion engine" is a software component for extracting and analyzing emotions from a user's text or speech.

[1149] "Feedback" refers to the responses and advice provided by generative artificial intelligence or emotion engines based on the user's statements and emotional state.

[1150] "Mental health" refers to an individual's psychological stability and sense of well-being, and indicates a state in which they are free from mental problems such as stress, anxiety, and depression.

[1151] An "employee" is an individual employed by a business or organization who works to perform a specific role or function.

[1152] This invention is a system that combines generative AI and an emotion engine, and aims to improve responsiveness in the workplace by allowing users to interact with AI with specific roles. This system is realized using a server, terminals, a generative AI model, and an emotion engine.

[1153] Hardware and software used

[1154] Hardware: Servers, terminals (PCs, smartphones)

[1155] Software: Generative AI models (e.g., GPT-4), emotion engines (e.g., Affectiva)

[1156] System Overview

[1157] 1. Role Settings:

[1158] The server assigns specific roles to the generative AI model.

[1159] Specific roles include "a subordinate who comes out as a sexual minority," "a subordinate who announces pregnancy," and "a subordinate who is worried about their work style."

[1160] 2. Start chatting:

[1161] A user accesses the chat application using a terminal.

[1162] As a boss, the user starts chatting with an AI with a set role.

[1163] 3. Analysis of response results:

[1164] The server analyzes the chat content in real time.

[1165] Run algorithms to check for harassment patterns and bias.

[1166] 4. Emotional state analysis:

[1167] The emotion engine analyzes the user's emotional state.

[1168] Recognizes emotions such as anger or frustration when a user expresses them.

[1169] 5. Feedback:

[1170] The server adjusts the response of the generative AI model based on the analysis results of the emotion engine.

[1171] For example, the generative AI model is instructed to "respond in a way that will ease the user's anger."

[1172] Specifically, the generative AI model sends an apology message such as, "I'm sorry, I may have misunderstood you."

[1173] Examples and prompts

[1174] Examples:

[1175] A user starts a chat with a subordinate who is coming out as a sexual minority.

[1176] User: "Is anything bothering you lately?"

[1177] AI (subordinate role): "Actually, I want to come out as a LGBTQ person."

[1178] Example prompt sentence:

[1179] "As a manager, you are being consulted by a subordinate who wants to come out as a sexual minority. How would you respond?"

[1180] This system allows users to improve their ability to respond in the workplace. In addition, the emotion engine enables appropriate feedback based on the user's emotional state. This can prevent harassment and bias issues and lead to more effective communication.

[1181] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1182] Step 1:

[1183] Role Settings

[1184] The server assigns specific roles to the generative AI model.

[1185] Input: Role information set by the administrator (e.g., "Subordinate who comes out as a sexual minority")

[1186] Data processing: The server sends the role information to the generative AI model to generate a prompt.

[1187] Output: Role and prompt set for the generative AI model

[1188] Specific behavior: The server sends the prompt message "You are a subordinate who is coming out as a sexual minority. Please talk to your supervisor." to the generative AI model.

[1189] Step 2:

[1190] Start chatting

[1191] A user accesses the chat application using a terminal.

[1192] Input: User login information and chat start request

[1193] Data processing: The server authenticates the user and starts the chat session.

[1194] Output: Notification that a chat session has started

[1195] Specific behavior: The user sends a message such as "Is there anything bothering you lately?"

[1196] Step 3:

[1197] Analysis of response results

[1198] The server analyzes the chat content in real time.

[1199] Input: Chat content between the user and the generated AI model

[1200] Data processing: The server uses natural language processing algorithms to check for harassment patterns and bias.

[1201] Output: Analysis results (presence or absence of harassment trends and bias)

[1202] Specific behavior: The server determines whether the user's statement contains a specific bias.

[1203] Step 4:

[1204] Emotional state analysis

[1205] The emotion engine analyzes the user's emotional state.

[1206] Input: User chat content

[1207] Data processing: The emotion engine extracts emotions from text and recognizes emotions such as anger or frustration.

[1208] Output: Sentiment analysis results (e.g. anger, sadness, joy)

[1209] What it does: If a user expresses anger, such as "Why would you say that?", the emotion engine will detect that emotion.

[1210] Step 5:

[1211] feedback

[1212] The server adjusts the response of the generative AI model based on the analysis results of the emotion engine.

[1213] Input: Sentiment analysis results and chat content

[1214] Data processing: The server instructs the generative AI model on how to respond based on the user's emotional state.

[1215] Output: The adjusted generative AI model response

[1216] What it does: The generative AI model sends an apology message such as, "Sorry, I may have misunderstood you."

[1217] (Application example 1)

[1218] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1219] Conventional generative AI systems have difficulty properly analyzing a user's emotional state and providing feedback when the user responds to a specific role. Furthermore, they lack the ability to adjust the AI's response based on the user's emotional state, which can result in the AI ​​sometimes failing to respond appropriately to the user. This has led to problems such as inaccurate analysis when checking for harassment tendencies or bias.

[1220] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means. In this invention, the server includes means for setting a specific role for the generative AI and for the user to respond to that role via chat, means for analyzing the response results and checking for harassment tendencies and bias, means for analyzing the user's emotional state using an emotion engine and feeding back the results, and means for adjusting the AI's response according to the user's emotional state. This enables appropriate responses according to the user's emotional state and makes it possible to accurately check for harassment tendencies and bias.

[1221] "Generative AI" is an artificial intelligence system that generates natural language responses based on user input.

[1222] A "specific role" refers to a specific scenario or character that is set in the generative AI, and is the role that the user responds to.

[1223] "Chat" is a means of text-based real-time communication between users and generative AI.

[1224] "Harassment" refers to inappropriate words, actions, or behavior toward others in the workplace or in society, which causes mental or physical pain to the other person.

[1225] "Bias" is the distortion of judgment or behavior based on a particular prejudice or preconception.

[1226] The "emotion engine" is a system that analyzes the user's emotional state from their text input and provides appropriate feedback based on the results.

[1227] "Feedback" refers to the response or evaluation that a generative AI gives to a user's actions or input.

[1228] "Security" refers to the measures and procedures used to protect information and physical assets and to guard against unauthorized access and attacks.

[1229] "Mental health" refers to the state of mental health and is a concept that includes mental problems such as stress, anxiety, and depression.

[1230] A "return-to-work program" is a plan or procedure for providing assistance to employees returning to work after a leave of absence.

[1231] A system for implementing this invention includes a means for setting a specific role for generative AI and for the user to respond to that role via chat, a means for analyzing the results of the response and checking for any tendency toward harassment or bias, a means for analyzing the user's emotional state using an emotion engine and feeding back the results, and a means for adjusting the AI's response according to the user's emotional state.

[1232] 1. System Program

[1233] This system includes the following programs:

[1234] Generative AI: Generates natural language responses based on user input.

[1235] Emotion engine: Analyzes the user's emotional state from their input text.

[1236] Feedback module: Provides appropriate feedback to users based on the analysis results of the emotion engine.

[1237] Adjustment module: Adjusts the response of the generative AI depending on the user's emotional state.

[1238] 2. Program processing explanation

[1239] Hardware and software used

[1240] Hardware: Smartphone

[1241] Software: Python, OpenAI API, TextBlob

[1242] Data processing and calculation

[1243] 1. Receiving user input: The user enters input in chat format using their smartphone.

[1244] 2. Sentiment Analysis: Analyze the sentiment of the user's input text using TextBlob. Determine whether the sentiment is positive, negative, or neutral.

[1245] 3. AI Response Generation: Uses OpenAI's API to generate appropriate responses to user input.

[1246] 4. Feedback adjustment: Based on the results of sentiment analysis, adjust the AI's response and provide feedback to the user.

[1247] 3. Specific Examples

[1248] Specific examples

[1249] If a user types, "I've been worried about the security of my home lately. What should I do?", sentiment analysis will determine this as negative, and the AI ​​will respond with, "I understand your concerns. Can you tell me more?"

[1250] Prompt Sentence Examples

[1251] User: I'm worried about the security of my home these days. What should I do?

[1252] AI: I understand your concerns. Can you elaborate a bit more?

[1253] In this way, it is possible to respond appropriately to the user's emotional state and accurately check for harassment tendencies and biases.

[1254] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1255] Step 1:

[1256] The user inputs information in chat format using a smartphone. The input text is sent to the server. The input data is the user's inquiry, for example, "I've been worried about the security of my home lately. What should I do?"

[1257] Step 2:

[1258] The server passes the received user input text to the emotion engine. The emotion engine uses TextBlob to analyze the sentiment of the text. Specifically, it determines whether the sentiment of the text is positive, negative, or neutral. For example, if the input text contains the negative sentiment "I'm worried," the emotion engine will determine it as negative.

[1259] Step 3:

[1260] The server sends a prompt to the generative AI model based on the analysis results of the emotion engine. The prompt includes the user's input text and the results of the emotion analysis. For example, the server sends the prompt "User: I've been worried about the security of my home lately. What should I do? Emotion: Negative" to the generative AI model.

[1261] Step 4:

[1262] The generative AI model generates an appropriate response based on the prompt sentence, which corresponds to the user's emotional state, such as "I understand your concerns. Can you tell me more?"

[1263] Step 5:

[1264] The server receives the response from the generative AI model and passes it to the feedback module, which adjusts the final feedback to the user based on the analysis results of the emotion engine and the response of the generative AI model, for example, adding an additional message to alleviate negative emotions.

[1265] Step 6:

[1266] The server then sends the final feedback to the user's smartphone, where the user can view the response from the generative AI model and continue chatting. For example, the message might read, "I understand your concerns. Can you tell me more?"

[1267] In this way, it is possible to respond appropriately according to the user's emotional state, and to accurately check for harassment tendencies and biases.

[1268] Example 2

[1269] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1270] With conventional systems, it was difficult to properly analyze the results of user responses and check for harassment tendencies and bias. It was also difficult to adjust the AI's response based on the user's emotional state, which meant that it was unable to provide appropriate feedback to the user.

[1271] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes: means for setting a specific role for the generative artificial intelligence and for the user to respond to that role via communication means; means for collecting the response results and storing them in a database; means for analyzing the stored data using natural language processing technology; means for determining whether the analyzed data has a tendency toward harassment or bias using a machine learning model; means including an emotion engine for analyzing the user's emotional state and adjusting the response; and means for feeding back the analysis results. This makes it possible to appropriately analyze the user's response results and check for a tendency toward harassment or bias. Furthermore, the AI's response can be adjusted based on the user's emotional state, making it possible to provide appropriate feedback to the user.

[1272] "Generative AI" is an AI system that is assigned a specific role and interacts with the user based on that role.

[1273] "Means of communication" refers to the devices and networks that users use to interact with generative artificial intelligence.

[1274] The "database" is an information management system for storing the results of users' responses.

[1275] "Natural language processing technology" is a technology that analyzes text data, extracts specific keywords and phrases, and evaluates context and tone.

[1276] A "machine learning model" is an algorithm that is trained on past data to identify specific patterns and trends.

[1277] The "emotion engine" is a system that analyzes the user's emotional state and adjusts the AI's response based on the results.

[1278] "Feedback means" refers to a method or system for communicating the analysis results to the user.

[1279] This invention is a system that uses generative AI to analyze user responses and check for harassment tendencies and biases. It also has the ability to adjust the AI's response based on the user's emotional state.

[1280] Hardware and software used

[1281] This system operates using a server and a terminal. The server is a cloud-based server equipped with a high-performance processor and large memory capacity. Specifically, it uses a virtual server from a general cloud service provider. The terminal is a device operated by the user, such as a PC or smartphone.

[1282] The software uses the following technologies:

[1283] Natural language processing libraries (e.g. spaCy, NLTK)

[1284] Machine learning frameworks (e.g. TensorFlow, PyTorch)

[1285] Sentiment analysis engine (e.g., IBM Watson's Tone Analyzer)

[1286] Data processing and calculation

[1287] 1. Collecting user input data:

[1288] A user uses a terminal to input text data. For example, the user types, "This service is really great!"

[1289] The terminal transmits this data to the server.

[1290] 2. Data storage by the server:

[1291] The server stores the received text data in a database, such as MySQL or PostgreSQL.

[1292] 3. Server-based Natural Language Processing (NLP):

[1293] The server analyzes the stored text data using a natural language processing library, for example, extracting the keyword "great" from the text.

[1294] 4. Analysis using machine learning models on the server:

[1295] The server uses machine learning models, such as models trained on historical data, to determine whether a message is harassing or biased based on the extracted keywords, phrases, and tone.

[1296] 5. Running the emotion engine on the server:

[1297] The server uses an emotion engine to analyze the user's emotional state, for example recognizing the emotion "happy."

[1298] 6. Feedback from the server to the user:

[1299] Based on the analysis, the server sends appropriate feedback to the user, such as a message like "That's great! I'm happy too."

[1300] Specific examples

[1301] Collecting user input data

[1302] Example: User types, "This service is absolutely amazing!"

[1303] Action: The device sends this text data to the server.

[1304] Server-based data storage

[1305] How it works: The server stores the received text data in a MySQL database.

[1306] Server-based natural language processing (NLP) execution

[1307] What it does: The server uses spaCy to extract the keyword "great".

[1308] Analysis using machine learning models on the server

[1309] How it works: The server uses a TensorFlow model to determine if the text is harassment-free or bias-free.

[1310] Emotion engine execution by server

[1311] How it works: The server uses IBM Watson's Tone Analyzer to recognize the emotion "happy."

[1312] Feedback from the server to the user

[1313] What it does: The server generates a message such as "That's great! I'm happy too" and sends it to the user via the device.

[1314] Prompt Sentence Examples

[1315] Prompt: "Analyze user-entered text to determine if it contains harassment or bias. Take appropriate action based on the user's emotional state."

[1316] In this way, the system operates in cooperation with the server, terminals, and users.

[1317] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1318] Step 1: Collecting User Input Data

[1319] Input: A user uses a terminal to enter text data. For example, a user types, "This service is really great!"

[1320] Specific operation: The device sends this text data to the server.

[1321] Output: The text data sent to the server.

[1322] Step 2: Server saves data

[1323] Input: Text data sent from the terminal.

[1324] Specific operation: The server stores the received text data in a database, for example, using a database such as MySQL or PostgreSQL.

[1325] Output: Text data stored in a database.

[1326] Step 3: Server performs natural language processing (NLP)

[1327] Input: Text data stored in a database.

[1328] What happens: The server uses a natural language processing library (e.g., spaCy) to analyze the text data. For example, it extracts the keyword "great" from the text.

[1329] Output: Extracted keywords and phrases, context and tone assessment results.

[1330] Step 4: Analysis using machine learning models on the server

[1331] Input: Keywords and phrases extracted through natural language processing, and assessment results of context and tone.

[1332] What it does: The server uses a machine learning model (e.g., TensorFlow) to determine whether or not harassment or bias exists based on the extracted information, for example, using a model trained on historical data.

[1333] Output: Judgment result on whether harassment or bias existed.

[1334] Step 5: Running the Emotion Engine on the Server

[1335] Input: User-entered text data.

[1336] Specific operation: The server uses an emotion engine (e.g., IBM Watson's Tone Analyzer) to analyze the user's emotional state. For example, it recognizes the emotion "happy."

[1337] Output: Analysis results about the user's emotional state.

[1338] Step 6: Feedback from the server to the user

[1339] Input: Judgment results regarding the presence or absence of harassment or bias, and analysis results regarding the user's emotional state.

[1340] Specific operation: Based on the analysis results, the server generates appropriate feedback for the user and sends it to the user via the device. For example, it sends a message such as "That's great! I'm happy too."

[1341] Output: The feedback message sent to the user.

[1342] (Application example 2)

[1343] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1344] It is necessary to detect signs of harassment and bias in employee communications early and respond appropriately. It is also important to understand employees' emotional state in real time and provide appropriate feedback and support. However, due to a lack of efficient and effective means to resolve these issues, improvements to employee mental health and the work environment are delayed.

[1345] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1346] In this invention, the server includes a means for setting a specific role for the generative AI and for the user to respond to that role via chat, a means for analyzing the response results and checking for harassment tendencies and bias, a means for feeding back the analysis results, a means for analyzing the emotional state of the user and adjusting the AI's response based on the emotion, and a means for monitoring communication between employees in real time and providing appropriate feedback and support. This enables early detection of harassment and bias in communication between employees and appropriate responses according to the emotional state of employees.

[1347] "Generative AI" is artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[1348] "Specific role" refers to the specific role or scenario that the generative AI should play in interacting with the user.

[1349] "User response" refers to the chat responses and actions that users take toward generative AI.

[1350] "Harassment" refers to words, actions, or behavior that cause discomfort or pain to others.

[1351] "Bias" refers to words, actions, or behavior that reflect prejudice or preconceived notions toward a particular individual or group.

[1352] "Emotional state" refers to the type and intensity of emotions a user displays during a conversation.

[1353] "Feedback" refers to the responses and information that generative AI provides based on the user's response and emotional state.

[1354] "Real-time" refers to immediate processing or response without delay.

[1355] "Communication between employees" refers to dialogue and information exchange between employees in the workplace.

[1356] "Appropriate feedback and support" refers to responses and assistance provided based on the user's emotional state and the results of their response, with the aim of solving problems and improving mental health.

[1357] The system for implementing this invention uses the following hardware and software. The hardware requires a smartphone and a server. The software uses Python, NLTK (a natural language processing library), and Scikit-learn (a machine learning library).

[1358] The server assigns specific roles to the generative AI and provides a means for users to respond to those roles in chat. Natural language processing technology is used to analyze the user's responses and check for harassment tendencies and bias. Specifically, NLTK's sentiment analysis tool (SentimentIntensityAnalyzer) is used to analyze the sentiment of the messages. Specific keywords are also counted to calculate scores for harassment and bias.

[1359] Furthermore, it provides a means to analyze the user's emotional state and adjust the AI's response based on the emotion. This allows it to provide appropriate feedback and support depending on the emotion the user expresses. For example, if the user expresses positive emotion, the AI ​​will respond by sharing its joy, and if the user expresses negative emotion, it will respond by promising improvement.

[1360] This also includes the ability to monitor employee communications in real time and provide appropriate feedback and support, which will enable early detection of harassment and bias in the workplace and contribute to improving employee mental health.

[1361] As a concrete example, the following prompt sentence can be used:

[1362] "Enter a user message. For example, 'I think this project is unfair.'"

[1363] This system enables early detection of harassment and bias in employee communications and appropriate responses based on the employee's emotional state.

[1364] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1365] Step 1:

[1366] The server assigns specific roles to the generative AI and provides an interface for users to respond to those roles in chat. As input, it receives messages entered by users in chat, and as output, the generative AI generates a response based on that role.

[1367] Step 2:

[1368] The server analyzes the user's response. Specifically, it uses natural language processing technology to analyze the user's message. It receives the user's message as input and calculates an emotion score, harassment score, and bias score as output. For sentiment analysis, it uses NLTK's SentimentIntensityAnalyzer, counting specific keywords and calculating a score.

[1369] Step 3:

[1370] The server feeds back the analysis results. It receives emotion scores, harassment scores, and bias scores as input, and generates an appropriate feedback message for the user as output. For example, in the case of a positive emotion score, it generates a message such as "We share your joy. Thank you!", and in the case of a negative emotion score, it generates a message such as "We apologize for any inconvenience caused. We will strive to improve."

[1371] Step 4:

[1372] The server analyzes the user's emotional state and adjusts the AI's response based on the emotion. It receives the user's emotion score as input and generates the AI's response based on the emotion as output. For example, if the user expresses positive emotion, the AI ​​will respond by sharing its joy, and if the user expresses negative emotion, it will respond by promising improvement.

[1373] Step 5:

[1374] The server monitors communication between employees in real time and provides appropriate feedback and support. It receives chat messages between employees as input and generates alerts and notifies managers if signs of harassment or bias are detected as output. It also generates and provides support messages to employees according to their emotional state.

[1375] Example 3

[1376] Next, a description will be given of Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1377] With conventional systems, it was difficult to properly check for harassment tendencies or bias when users responded to specific roles. Furthermore, the system did not provide feedback that took into account the user's emotional state, which led to problems with users not being able to adequately improve their responses. Furthermore, there was also the issue of difficulty in providing appropriate support for employees who had taken leave due to mental health issues when returning to work.

[1378] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1379] In this invention, the server includes a means for assigning specific roles to the generative AI and allowing the user to interactively respond to that role, a means for analyzing the response results and checking for harassment tendencies and bias, and a means for analyzing the user's emotional state using an emotion engine. This makes it possible to check for harassment and bias in the user's responses, and by providing feedback that takes the user's emotional state into consideration, it promotes improvements in the user's responses. It also enables appropriate support for employees who have taken leave due to mental health issues when they return to work.

[1380] "Generative AI" is artificial intelligence that generates natural language responses based on prompts entered by the user.

[1381] A "specific role" refers to a specific scenario or situation in which a user interacts with the system.

[1382] "User" refers to an individual or member of an organization who utilizes the system to fulfill a particular role.

[1383] "Dialogue format" refers to a format in which users interact with generative AI in natural language.

[1384] "Harassment" refers to inappropriate words, actions, or behavior toward others in the workplace or in society.

[1385] "Bias" refers to having prejudices or preconceived notions toward a particular individual or group.

[1386] An "emotion engine" refers to software or hardware that analyzes a user's emotional state and provides appropriate feedback based on the results.

[1387] "Feedback" refers to specific improvements and advice provided based on the user's response and emotional state.

[1388] "Mental health" refers to mental health conditions and includes mental health issues such as stress, anxiety, and depression.

[1389] "Leave of absence" refers to an employee temporarily leaving work for health reasons or other reasons.

[1390] "Return-to-work support" refers to the support and programs provided to employees who have been on leave when they return to work.

[1391] This invention is a system that allows users to interactively respond to specific roles, analyzes the results of those responses, checks for harassment tendencies and biases, and analyzes the user's emotional state using an emotion engine. This system is built by combining a generative AI model and an emotion engine.

[1392] The server assigns specific roles using a generative AI model (for example, an AI model that processes natural language). Users can chat with this AI via their devices in an interactive format, simulating how to respond to, for example, an employee returning from leave or a subordinate coming out as LGBT.

[1393] The device transmits the user's chat content to the server in real time. The server analyzes the received chat content to check for harassment tendencies and bias. It also analyzes the user's emotional state using an emotion engine (e.g., emotion recognition software).

[1394] The server then provides the user with feedback based on the analysis results. This feedback includes specific areas for improvement and advice. For example, if a user shows bias toward "subordinates who come out as LGBT," the server will present the specific details of that bias and suggestions for improvement.

[1395] As a concrete example, consider a case where a user simulates how to respond to an employee returning from leave. The user inputs the following prompt into the generative AI model:

[1396] How should I talk to an employee returning from leave?

[1397] When the user enters this prompt, the generative AI model suggests an appropriate response. The user chats based on the suggestion, and the server analyzes the response. As a result of the analysis, feedback is provided, such as "The user is providing appropriate support, but more specific advice is needed."

[1398] If a user expresses anger or dissatisfaction during chat, the emotion engine will recognize that emotion and the server will adjust the AI's response. For example, the AI ​​may respond to reduce the user's anger or send an apology.

[1399] In this way, the user can successively improve his / her own response through the system. The flow of the identification process in the third embodiment will be described with reference to FIG.

[1400] Step 1:

[1401] The user enters a prompt statement.

[1402] The user inputs a prompt to the generative AI model through the device interface, for example, "Please tell me how to greet an employee returning from leave."

[1403] Input: The prompt text entered by the user

[1404] Output: The prompt text displayed on the terminal

[1405] Step 2:

[1406] The terminal sends the prompt to the server.

[1407] The terminal sends the prompt text entered by the user to the server using a protocol such as an HTTP request.

[1408] Input: The prompt text entered by the user

[1409] Output: The prompt sent to the server

[1410] Step 3:

[1411] The server generates a response using a generative AI model.

[1412] The server inputs the received prompt sentence into a generative AI model (e.g., an AI model that performs natural language processing) to generate an appropriate response. The generative AI model generates a natural language response based on the prompt sentence.

[1413] Input: The prompt sent to the server

[1414] Output: The response generated by the generative AI model

[1415] Step 4:

[1416] The server generates a response and sends it to the terminal.

[1417] The server sends the response generated by the generative AI model to the terminal, again using a protocol such as HTTP response.

[1418] Input: The response generated by the generative AI model

[1419] Output: Response sent to the terminal

[1420] Step 5:

[1421] The user chats based on the generated response.

[1422] The user checks the generative AI model's response displayed on the device and continues the chat based on it. For example, if the generative AI model suggests, "First, send a message welcoming the user back," the user will follow that suggestion and chat.

[1423] Input: The response of the generative AI model displayed on the device

[1424] Output: Chat content of the user

[1425] Step 6:

[1426] The device sends the chat contents to the server.

[1427] The terminal transmits the contents of the chat that the user has conducted to the server in real time, allowing the server to grasp the user's responses successively.

[1428] Input: Chat content that the user has had

[1429] Output: Chat messages sent to the server

[1430] Step 7:

[1431] The server analyzes the chat content.

[1432] The server analyzes the chat content received and checks for harassment tendencies and bias. For example, if a user makes an inappropriate remark about a subordinate who has come out as LGBT, the server will detect that content.

[1433] Input: Chat content sent to the server

[1434] Output: Analysis results regarding harassment trends and bias

[1435] Step 8:

[1436] The server uses an emotion engine to analyze the user's emotional state.

[1437] The server uses an emotion engine (e.g., emotion recognition software) to analyze the user's emotional state, for example, recognizing when a user expresses anger or frustration during a chat.

[1438] Input: Chat content sent to the server

[1439] Output: Analysis results about the user's emotional state

[1440] Step 9:

[1441] The server generates the analysis results as feedback.

[1442] The server generates feedback for the user based on the analysis of the chat content and emotional state, including specific suggestions for improvement and advice.

[1443] Input: Analysis results on harassment trends and bias, and analysis results on the user's emotional state

[1444] Output: Feedback to the user

[1445] Step 10:

[1446] The server sends the feedback to the device.

[1447] The server then sends the generated feedback to the device, and the user receives this feedback, which provides specific guidelines for improving their own behavior.

[1448] Input: Feedback to the user

[1449] Output: Feedback sent to the terminal

[1450] Step 11:

[1451] Users receive feedback and improve their response.

[1452] Users can review the feedback displayed on their devices and use it to improve their own behavior, for example by taking specific actions to reduce bias toward subordinates who come out as LGBT.

[1453] Input: Feedback sent to the device

[1454] Output: User-improved response

[1455] (Application example 3)

[1456] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[1457] Conventional systems could check for harassment tendencies and bias when users responded to specific roles, but they were unable to analyze the user's emotional state in real time and provide appropriate feedback based on those emotions. As a result, responses based on the user's emotional state were insufficient, and appropriate responses were particularly needed in emotionally charged situations. Furthermore, in collaborative work between humans and robots in factories, communication improvements that take into account the emotional state of workers were also necessary.

[1458] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 3 is realized by the following means.

[1459] In this invention, the server includes a means for assigning specific roles to generative AI and allowing users to respond to those roles via chat, a means for analyzing the results of the responses and checking for harassment tendencies and bias, a means for providing feedback on the analysis results, a means for using an emotion engine that analyzes the user's emotional state in real time, and a means for providing appropriate feedback according to the user's emotional state. This enables more appropriate responses according to the user's emotional state, making it possible to improve communication that takes into account the emotional state of workers, particularly in collaborative work between humans and robots in factories.

[1460] "Generative AI" is artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[1461] A "specific role" is a role that a user will play in a specific scenario or situation set within the system.

[1462] A "chat interaction" is an interface that allows users to interact with a particular role through text-based dialogue.

[1463] The "means for analyzing response results" is a function that analyzes the content of the responses made by users and evaluates whether there is a tendency for harassment or bias.

[1464] "Means to check for harassment tendencies and bias" is a function that determines whether harassment or prejudice is present based on the user's response results.

[1465] "Means for providing feedback on analysis results" is a function for providing specific improvements and advice to users based on the analysis of response results.

[1466] The "emotion engine" is a technology that analyzes the user's emotional state in real time and responds appropriately based on those emotions.

[1467] "Real-time analysis" means analyzing the user's emotional state instantly the moment they give input.

[1468] "Means for providing appropriate feedback" is a function that provides appropriate responses and advice depending on the user's emotional state.

[1469] "Collaboration between humans and robots in a factory" refers to a situation in which human workers and robots work together in a factory.

[1470] "Improved communication" refers to measures to make dialogue and the exchange of instructions between users and robots smoother, more efficient, and more effective.

[1471] The system for implementing this invention is a combination of generative AI, an emotion engine, and a feedback system. A specific embodiment of this system will be described below.

[1472] System configuration

[1473] The system consists of the following main components:

[1474] 1. Generative AI: Artificial intelligence that generates natural language based on user input and engages in dialogue with specific roles.

[1475] 2. Emotion Engine: Technology that analyzes the user's emotional state in real time.

[1476] 3. Feedback system: A function that analyzes the user's response results and provides specific improvements and advice.

[1477] Hardware and software used

[1478] Hardware: Smart glasses (e.g., Google Glass)

[1479] Software: OpenAI API, EmotionEngine, FeedbackSystem

[1480] System Operation

[1481] 1. User input: The user responds to a specific role through the chat via smart glasses.

[1482] 2. Emotion Analysis: The emotion engine analyzes user input in real time to identify their emotional state.

[1483] 3. Generative AI Response: Generative AI generates appropriate responses based on user input and emotional state.

[1484] 4. Providing feedback: The feedback system analyzes the user's response and provides specific improvements and advice.

[1485] Specific examples

[1486] For example, if a worker in a factory expresses dissatisfaction with a robot, the following prompt sentence may be generated:

[1487] User Input: This robot is completely useless!

[1488] Emotion: Anger

[1489] Please suggest an appropriate response.

[1490] This prompt can be fed into a generative AI model to generate an appropriate response and feedback. The generative AI takes into account the user's emotional state and generates a response such as, "Sorry. Can you tell me what the problem is?" The emotion engine suggests responses to alleviate the user's anger, and the feedback system provides specific improvements.

[1491] Data processing and calculation

[1492] Data processing: The user's input data is obtained in text format and analyzed using the emotion engine.

[1493] Data calculation: The emotion engine analyzes the emotional data and passes it to the generative AI to generate an appropriate response. The feedback system calculates specific improvements based on the generated response and user input.

[1494] In this way, it becomes possible to respond appropriately to the user's emotional state, and communication improvements that take into account the emotional state of workers can be realized, particularly in collaborative work between humans and robots in factories.

[1495] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1496] Step 1:

[1497] Users respond to specific roles via chat through smart glasses.

[1498] Input: User text input (e.g., "This robot is completely useless!")

[1499] Output: User input text data

[1500] Specific behavior: A user uses the smart glasses interface to provide text input for a specific role, which is then sent to the system.

[1501] Step 2:

[1502] The server uses an emotion engine to analyze the user's input text data in real time and identify the user's emotional state.

[1503] Input: User-entered text data

[1504] Output: User's emotional state data (e.g., "Anger")

[1505] Specific operation: The server starts the emotion engine and analyzes the user's input text. The emotion engine extracts emotions from the text and outputs the results as emotional state data.

[1506] Step 3:

[1507] The server uses generative AI to generate an appropriate response based on the user's input text data and emotional state data.

[1508] Input: User input text data, emotional state data

[1509] Output: The generated response text (e.g. "Sorry, can you tell me what's wrong?")

[1510] How it works: The server calls the generative AI model and inputs the user's input text and emotional state as prompts. The generative AI then generates an appropriate response based on this and outputs it as response text.

[1511] Step 4:

[1512] The server uses a feedback system to analyze the user's response and provide specific improvements and advice.

[1513] Input: User input text data, generated response text

[1514] Output: Feedback text with suggestions for improvement

[1515] Specific operation: The server starts the feedback system, analyzes the user's input text and the generated response text, and based on this, generates specific improvements and advice for the user and outputs them as feedback text.

[1516] Step 5:

[1517] The server sends the generated response text and feedback text to the user.

[1518] Input: Generated response text, feedback text

[1519] Output: Responses and feedback displayed on the user's smart glasses

[1520] Specific operation: The server sends the generated response text and feedback text to the user's smart glasses so that the user can review them.

[1521] 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 a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1522] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1523] Another example of generative AI is Gemini (internet search engine). <url: https: gemini.google.com ?hl="ja">) are listed.

[1524] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[1525] [Third embodiment]

[1526] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1527] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1528] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. 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. The database 24 and the communication I / F 26 are also 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).

[1529] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1530] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1531] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1532] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1533] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1534] The specific processing program 56 is an example of a "program" according to the technology of the present 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 in accordance with the specific processing program 56 executed on the RAM 30.

[1535] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1536] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1537] Next, the specific processing by the specific processing unit 290 of the data processing device 12 will be described.

[1538] "Example 1"

[1539] One embodiment of the present invention involves a means for assigning specific roles to generative AI. Specifically, roles such as "a subordinate coming out as LGBT," "a subordinate announcing a pregnancy," or "a subordinate struggling with work style" can be assigned to the AI. The user, as the supervisor, receives the consultation and then interacts with the AI ​​via chat.

[1540] "Example 2"

[1541] Next, there is a method to analyze user responses to check for harassment tendencies and bias. This analysis is done using natural language processing and machine learning techniques. Specifically, it determines whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc.

[1542] "Example 3"

[1543] Furthermore, there is a means to provide feedback on the analysis results. This feedback provides specific suggestions and advice to help users improve their own behavior. For example, if bias is found in the way employees treat subordinates who come out as LGBTQ+, the system will present the user with specific details of that bias and suggestions for improvement.

[1544] "Example 4"

[1545] The system of this invention can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. Specifically, to simulate how to respond when an employee returns from leave, the AI ​​is assigned the role of "an employee returning from leave," and the user responds via chat. The results are analyzed, and appropriate responses and return-to-work support programs are proposed.

[1546] The processing flow of each embodiment will be described below.

[1547] "Example 1"

[1548] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "a subordinate who comes out as LGBT."

[1549] Step 2: As a manager, the user responds to the subordinate's coming out via chat.

[1550] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[1551] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[1552] "Example 2"

[1553] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of "subordinate who reports pregnancy."

[1554] Step 2: As a supervisor, the user responds to the subordinate's pregnancy announcement via chat.

[1555] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[1556] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[1557] "Example 3"

[1558] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of a "subordinate who is worried about their work style."

[1559] Step 2: As a supervisor, the user responds to the subordinate's concerns via chat.

[1560] Step 3: AI analyzes the user's response and checks for harassment trends and bias.

[1561] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[1562] "Example 4"

[1563] Step 1: Assign a specific role to the generative AI. Specifically, assign the AI ​​the role of an "employee returning from leave."

[1564] Step 2: As a manager, the user responds to the employee's return via chat.

[1565] Step 3: AI analyzes the user's response and suggests appropriate responses and programs to support their return to work.

[1566] Step 4: The AI ​​provides the results of its analysis as feedback to the user, offering specific improvements and advice.

[1567] Example 1

[1568] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1569] In conventional systems using generative AI, when users respond to specific roles in chat, the system does not adequately generate appropriate prompts or display responses, resulting in a poor user experience. Furthermore, the system does not analyze the response results or check for harassment tendencies or bias, resulting in insufficient feedback. This limits users' opportunities to learn appropriate responses.

[1570] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1571] In this invention, the server includes: means for accepting a user's selection from multiple scenarios; means for setting a specific role based on the scenario selected by the user and generating a prompt sentence instructing how to respond to the role; means for accepting a user's input message; means for generating a response to the user's input message using the generated prompt sentence and a generative AI; means for outputting the generated response to the user; means for repeatedly accepting the input message, generating the response, and outputting the application; a prompt sentence instructing the user to check for harassment tendencies or bias in the user's responses based on a history of the input messages and responses; and means for evaluating the user's responses using the generative AI. This allows the user to learn how to respond to specific roles in real time and receive appropriate feedback. The multiple scenarios include a user interacting with a subordinate who comes out as a sexual minority, a user interacting with a subordinate who announces pregnancy, and a user interacting with a subordinate who is concerned about their work style.

[1572] "Generative AI" is an artificial intelligence system that generates natural language responses based on user input.

[1573] A "specific role" is a specific character or situation that the user sets based on a scenario and that the generative AI will imitate.

[1574] A "user" is a person who uses the system to chat with generative artificial intelligence.

[1575] A "prompt" is text input to a generative artificial intelligence (AI) that instructs it to generate a response based on a specific role.

[1576] A "response" is a text message that the generative artificial intelligence generates based on the prompt sentence.

[1577] "Chat" is an interactive form of communication between a user and a generative artificial intelligence via text messages.

[1578] A "scenario" is a setting containing a specific situation or role selected by the user, which serves as the basis for the generative AI to generate responses.

[1579] "Harassment tendencies" refers to a pattern of inappropriate speech and behavior toward others as a result of a user's responses.

[1580] "Presence or absence of bias" indicates whether or not a particular prejudice or preconception is included in the user's response results.

[1581] "Feedback" is information for evaluation and improvement provided based on the user's response results.

[1582] This invention is a system that uses generative artificial intelligence to set specific roles and allows users to respond to those roles through chat. A specific embodiment of this system will be described below.

[1583] First, the user selects one of several scenarios through the device interface. For example, scenarios are displayed in which the user talks to a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style.

[1584] The server generates a prompt based on the scenario selected by the user. Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4 (registered trademark)) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to that role. For example, it generates a prompt that reads, "The user is your boss. You are a subordinate who is consulting with him about coming out as a sexual minority. Please generate a response to the user's input message so as to reproduce this dialogue."

[1585] Next, the server inputs the generated prompt and the user's input message into the generative AI to generate a response. For example, the generated response might be, "Boss, could you please give me a moment? Actually, there's something I've been wanting to say for a while."

[1586] The server generates a response and sends it to the terminal, which displays it to the user. The user sees the response and types in a new message, such as "Sure, what's wrong?"

[1587] The server then inputs the user's input message back into the generative AI to generate the next response, such as, "Actually, I want to come out as a sexual minority, but I don't know how to say it..."

[1588] This process is repeated, and the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and respond to a subordinate's consultation.

[1589] Furthermore, the server analyzes the user's response based on the history of input messages and responses, and evaluates whether there is a tendency for harassment or bias. This allows the server to evaluate how the user responded and provide appropriate feedback.

[1590] As a specific example, the prompt sentence when the user selects the scenario "a subordinate who comes out as a sexual minority" is shown below.

[1591] Example prompt sentence:

[1592] "The user is your manager. You are a subordinate who is asking for advice on coming out as a LGBTQ person. Please generate responses to the user's input messages to recreate this dialogue. You are nervous and wondering how to proceed. The subordinate's initial message is as follows: "Boss, can I have a moment? There's actually something I've been wanting to tell you for a while."

[1593] In this way, users can learn how to respond to specific roles in real time and receive appropriate feedback.

[1594] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1595] Step 1:

[1596] The user selects a scenario.

[1597] The user selects one of several scenarios through the terminal interface. For example, scenarios are displayed in which the user interacts with a subordinate who comes out as a sexual minority, a subordinate who announces her pregnancy, or a subordinate who is worried about their work style. The scenario selected by the user is sent to the server (input). The server receives this scenario information and proceeds to the next process (output).

[1598] Step 2:

[1599] The server generates a prompt.

[1600] The server generates a prompt based on the scenario selected by the user (input). Specifically, it uses a generative AI model (e.g., OpenAI's GPT-4) to set a specific role based on the scenario selected by the user and creates a prompt that instructs how to respond to the role (output).

[1601] Step 3:

[1602] The server inputs a prompt sentence into the generative AI model.

[1603] The server inputs the generated prompt sentence into the generative AI (input). The generative AI generates a response based on this prompt sentence. For example, a response such as "Boss, can you give me a moment? Actually, there's something I've been wanting to say for a while" is generated (output).

[1604] Step 4:

[1605] The server sends the generated response to the terminal.

[1606] The server sends the response obtained from the generative AI to the terminal (input), which receives the response and displays it to the user (output).

[1607] Step 5:

[1608] The terminal displays the response to the user.

[1609] The terminal displays the response received from the server to the user (input), who then looks at it and prepares to enter the next message (output).

[1610] Step 6:

[1611] The user enters the following message:

[1612] The user responds to the displayed response by typing the next message (input). For example, "Of course, what's wrong?" The user's input message is sent to the server via the terminal (output).

[1613] Step 7:

[1614] The server inputs the user's message into a generative AI model.

[1615] The server inputs the message entered by the user into the generative AI (input). The generative AI generates the following response based on this message. For example, it might generate a response like, "Actually, I want to come out as a sexual minority, but I don't know how to say it..." (output).

[1616] Step 8:

[1617] The server sends the generated response to the terminal.

[1618] The server sends the next response obtained from the generative AI to the terminal (input). The terminal receives this response and displays it to the user (output).

[1619] Step 9:

[1620] The terminal displays the response to the user.

[1621] The terminal displays the response received from the server to the user again (input), who then looks at it and prepares to input the next message (output).

[1622] Step 10:

[1623] This process is repeated.

[1624] This process is repeated as long as the interaction between the user and the generative AI continues. The user can experience a scenario in which they act as a boss and provide advice to their subordinates (input and output). The server may further include means for recognizing the user's emotions using an emotion engine described below, and when generating a response, may use the recognized user's emotions, the generated prompt sentence, and the generative AI to generate a response to the user's input message.

[1625] (Application example 1)

[1626] Next, a description will be given of Application Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1627] Conventional systems using generative AI allow users to set specific roles and practice responses for those roles, but they lack training for the specific issues that corporate security personnel face, such as security issues, whistleblowing, and suspected information leaks. Furthermore, the system has limited functionality for analyzing the results of these responses to check for harassment tendencies and bias, and the quality of the feedback is insufficient. Furthermore, it does not take into consideration the use of the system as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues. This means that corporate security personnel and managers are insufficiently trained to respond appropriately, potentially increasing security risks across the entire company.

[1628] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1629] In this invention, the server includes a means for assigning specific roles to the generative AI and allowing users to respond to those roles via chat; a means for analyzing the response results and checking for harassment tendencies and bias; a means for providing feedback on the analysis results; a means for simulating security issues and allowing personnel to practice their responses via chat; and a means for generating prompts using the generative AI model and providing appropriate responses. This allows corporate security personnel and managers to receive training on how to respond appropriately to security issues, whistleblowers, suspected information leaks, and other issues, thereby reducing security risks across the company. It can also be used as a return-to-work program or decision-making tool for employees who have taken leave due to mental health issues, contributing to employee mental health care.

[1630] "Generative AI" is artificial intelligence that generates natural language based on specific roles and scenarios and engages in dialogue with users.

[1631] "Specific roles" refer to specific scenarios or characters that are set in the generative AI, such as "a subordinate coming out as LGBT" or "a subordinate announcing a pregnancy."

[1632] "Methods via chat" refers to interfaces and functions that allow users to have text-based conversations with generative AI.

[1633] "Means for analyzing the response results" refers to algorithms and software that analyze the content of the dialogue between the user and the generative AI and evaluate whether there is any tendency toward harassment or bias.

[1634] "Means to check for harassment tendencies and bias" refers to a function that determines whether a user's response contains harassment or bias based on the content of the conversation.

[1635] "Means of providing feedback" refers to functions and interfaces that provide analysis results to users and indicate areas for improvement or attention.

[1636] "Means to simulate security issues" refers to a function that allows users to set security-related scenarios such as information leaks and whistle-blowing in the generative AI and practice how to respond to them.

[1637] A "generative AI model" is a machine learning model for generating natural language and generating appropriate responses based on specific prompts.

[1638] A "prompt sentence" is an input sentence that sets a specific role or scenario for generative AI, and serves as a trigger to start a dialogue with the user.

[1639] A system for implementing this invention includes means for setting a specific role for a generative AI and for the user to respond to that role via chat, means for analyzing the results of that response and checking for harassment tendencies or bias, means for feeding back the results of the analysis, means for simulating security issues and allowing personnel to practice their responses via chat, and means for generating prompt sentences using a generative AI model and providing appropriate responses.

[1640] System program generation

[1641] The server assigns specific roles to the generative AI and provides an interface for users to respond to those roles via chat. The text entered by the user is sent to the generative AI model on the server, which generates an appropriate response. The generated response is returned to the user, and the dialogue progresses.

[1642] Processing Description

[1643] The server performs processing using the following hardware and software:

[1644] Hardware: Internet-connected server

[1645] Software: Python, OpenAI API

[1646] The server receives input text from the user and sends it to the generative AI model. The generative AI model generates an appropriate response based on the prompt. The generated response is returned to the user via the server. The server then analyzes the content of the dialogue to check for harassment tendencies and bias. The analysis results are then fed back to the user.

[1647] Specific examples

[1648] For example, to simulate a security officer responding appropriately to a call from an employee reporting a suspected data breach, the following prompt might be used:

[1649] You are an employee reporting a suspected data breach. Please respond appropriately to the user's inquiry.

[1650] User: I recently noticed some confidential information from my company has been leaked. What should I do?

[1651] you:

[1652] Using this prompt, a generative AI model can generate appropriate responses that security personnel can practice, such as:

[1653] You: That's a very important issue. First, please provide specific evidence and details of the situation. Also, I need to report this matter to my superiors and the security department immediately. A potential information leak requires a quick response, so could you please provide more details?

[1654] In this way, security personnel can be trained to respond appropriately to actual inquiries.

[1655] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1656] Step 1:

[1657] A user accesses the chat interface using a terminal and enters a prompt sentence based on a specific role. The entered prompt sentence is sent to the server. The input data is the user's text input, and the output data is the prompt sentence sent to the server.

[1658] Step 2:

[1659] The server sends the received prompt to the generative AI model, which generates an appropriate response based on the prompt. The input data is the prompt, and the output data is the generated response. Specifically, the server calls the OpenAI API and sends the prompt.

[1660] Step 3:

[1661] The server receives the response generated by the generative AI model and returns it to the user's device. The input data is the response from the generative AI model, and the output data is the response sent to the user's device. Specifically, the server displays the generated response on the user's chat interface.

[1662] Step 4:

[1663] The user confirms the generated response and then provides further input. The user's new input is again sent to the server, and the process repeats from step 2. The input data is the user's new text input, and the output data is the new prompt sent to the server.

[1664] Step 5:

[1665] The server analyzes the dialogue between the user and the generated AI model in real time to check for harassment tendencies and bias. The input data is the dialogue content, and the output data is the analysis results. Specifically, the server analyzes the dialogue content using a natural language processing algorithm.

[1666] Step 6:

[1667] The server feeds back the analysis results to the user. The input data is the analysis results, and the output data is the feedback provided to the user. In specific operations, the server displays the analysis results in text format on the user's chat interface.

[1668] Step 7:

[1669] The server simulates security issues and generates new prompts for users to practice their responses in chat. The input data is the security scenario settings, and the output data is the new prompts. Specifically, the server generates new prompts using a generative AI model.

[1670] Step 8:

[1671] The user then practices responding to the new prompt sentence again in the chat interface. The input data is the user's text input for the new prompt sentence, and the output data is the new prompt sentence sent to the server. As a specific operation, the user responds to the new scenario using the chat interface.

[1672] Example 2

[1673] Next, a description will be given of Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1674] With the previous system, it was difficult to properly analyze the results of user responses and check for harassment tendencies or bias. Furthermore, there was an insufficient means of providing feedback to users about the results of the analysis, making it difficult to lead to improvements in user behavior. Furthermore, there was a lack of specific methods for evaluating employee responses in specific situations.

[1675] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1676] In this invention, the server includes a means for a user to respond to a specific role via chat, a means for collecting the response results, a means for preprocessing the collected text data, a means for inputting the preprocessed text data into a machine learning model to analyze the presence or absence of harassment trends and bias, and a means for feeding back the analysis results. This makes it possible to appropriately analyze the user's response results and check for the presence or absence of harassment and bias. Furthermore, by feeding back the analysis results to the user, it is possible to promote improvement of the user's behavior. Furthermore, it is possible to provide a specific means for evaluating an employee's response in a specific situation.

[1677] "User" means an individual or entity that utilizes the System and fulfills a particular role.

[1678] A "specific role" refers to a specific situation or scenario that a user is required to handle within the system.

[1679] "Chat" is an interface that allows users to communicate text-based.

[1680] "Response results" are records of responses and actions taken by users for specific roles.

[1681] "Means for collection" refers to a method or device for acquiring and storing the results of a user's response as data.

[1682] "Text data" is data that expresses the user's response results as text information.

[1683] The "preprocessing means" refers to a method or device that processes collected text data to convert it into an analyzable format.

[1684] A "machine learning model" is an algorithm or system that learns patterns and trends from data and makes predictions and classifications.

[1685] "Means for analysis" refers to a method or device that uses preprocessed text data to evaluate the presence or absence of harassment tendencies or bias.

[1686] "Feedback means" refers to a method or device for communicating the analysis results to the user.

[1687] "Harassment" refers to inappropriate words, actions or behaviour towards others.

[1688] "Bias" refers to making judgments or taking actions based on specific prejudices or preconceptions.

[1689] This invention is a system in which users respond to specific roles via chat, analyze the results of their responses to check for harassment tendencies and bias, and provide feedback on the analysis results. A specific embodiment of this system is described below.

[1690] First, the server provides an interface for users to respond to specific roles in chat. This interface can be accessed through devices such as web browsers and mobile applications. For example, a user can be assigned a role as an "employee in a specific situation," and respond in chat based on that role.

[1691] The server then collects the user's responses, which are then stored in a database as text data. The collected text data is preprocessed using a natural language processing (NLP) library (e.g., spaCy or NLTK). Preprocessing includes tokenization, stop word removal, stemming, etc.

[1692] The preprocessed text data is input into a machine learning model (e.g., scikit-learn or TensorFlow). The server uses this machine learning model to analyze the text data and determine whether there is any harassment tendency or bias. Specifically, it evaluates whether specific keywords, phrases, or tones are offensive.

[1693] The analysis results are stored in a database by the server. In addition, the server provides feedback to the user. For example, if a message sent by a user is judged to be offensive, the server will provide feedback such as, "Your message contains offensive language. Please try to use more polite language."

[1694] As a concrete example, consider the case where a user sends a message saying, "Your response is completely useless. You're incompetent." In this case, the server collects this message and preprocesses it using an NLP library. Then, using a machine learning model, it determines that this message is offensive. Finally, the server provides appropriate feedback to the user.

[1695] An example of a prompt might be:

[1696] "Please generate a program that analyzes user responses and checks for harassment tendencies and bias. Specifically, use natural language processing and machine learning techniques to determine whether user responses indicate harassment or bias based on specific keywords, phrases, tone, etc."

[1697] In this way, the server can properly analyze the user's response and check for harassment or bias. Furthermore, by providing feedback to the user, the analysis results can encourage the user to improve their behavior. Furthermore, the server can provide a concrete means for evaluating employee responses in specific situations.

[1698] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1699] Step 1:

[1700] The server provides an interface for users to respond to specific roles on the chat. Users access this interface through their terminals and start chats based on specific roles. The input is the user's chat message, and the output is the text data of that message.

[1701] Step 2:

[1702] The server collects users' chat messages and stores them in a database. The input is the user's chat messages, and the output is the stored text data. Specifically, the server captures messages in real time and writes them to the database.

[1703] Step 3:

[1704] The server preprocesses the collected text data. Specifically, it uses an NLP library (e.g., spaCy or NLTK) to perform the following processes: tokenization, stop word removal, and stemming. The input is the stored text data, and the output is the preprocessed text data. Specifically, the server splits the text data into tokens, removes unnecessary words, and extracts word stems.

[1705] Step 4:

[1706] The server inputs the preprocessed text data into a machine learning model for analysis. Specifically, it uses libraries such as scikit-learn and TensorFlow to determine whether there is any harassment tendency or bias. The input is the preprocessed text data, and the output is the analysis results. Specifically, the server inputs the text data into the model and evaluates specific keywords, phrases, and tone.

[1707] Step 5:

[1708] The server stores the analysis results in a database. The input is the analysis results, and the output is the stored analysis results. As a specific operation, the server writes the analysis results into the database.

[1709] Step 6:

[1710] The server provides feedback to the user. The input is the saved analysis result, and the output is a feedback message to the user. Specifically, the server generates a feedback message based on the analysis result and sends it to the user. For example, it provides a message such as, "Your message contains offensive language. Please try to use more polite language."

[1711] In this way, the server can properly analyze the user's response results, check for harassment or bias, and provide feedback.

[1712] (Application example 2)

[1713] Next, a description will be given of Application Example 2 of Form Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1714] In today's corporate environment, harassment and bias issues are serious, and there is a need to detect these issues early and respond appropriately. However, traditional methods make it difficult to detect these issues in real time and respond immediately. In addition, signs of harassment and bias are often overlooked, which can have a negative impact on the company's credibility and the mental health of employees. Therefore, effective measures to resolve these issues are needed.

[1715] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1716] In this invention, the server includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies and bias, means for feeding back the analysis results, means for analyzing chat and email content in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate responses to users. This makes it possible to detect harassment and bias issues in real time and respond immediately.

[1717] "Generative AI" is an artificial intelligence system that is assigned a specific role and interacts with the user based on that role.

[1718] A "specific role" refers to a specific scenario or character that is set in the generative AI, and is the role that the user responds to.

[1719] "User responses" refer to chat interactions and reactions that users have with specific roles assigned to generative AI.

[1720] "Response results" refer to the content of the interactions and reactions the user had with the generative AI.

[1721] "Harassment" refers to inappropriate words, actions, or behavior toward others, and is problematic behavior in the workplace and in society.

[1722] "Bias" refers to judgments or actions based on specific prejudices or preconceptions, and is a factor that leads to a lack of fairness.

[1723] "Means of checking" refers to methods and techniques for analyzing the results of users' responses and determining whether or not there is harassment or bias.

[1724] "Feedback means" refers to methods and techniques for notifying users and administrators of the analysis results and encouraging them to take appropriate action.

[1725] "Real-time analysis means" refers to methods and technologies for instantly analyzing chat and email content to detect signs of harassment or bias.

[1726] "Alert methods" are methods or technologies used to alert administrators when signs of harassment or bias are detected.

[1727] "Means for storing data in a database" refers to methods and technologies for recording the nature of any detected harassment or bias, making it available for later analysis and reference.

[1728] "Means for suggesting appropriate responses" are methods and techniques for suggesting appropriate responses to harassment and bias issues to users.

[1729] A system for implementing this invention includes means for setting specific roles for generative AI and for users to respond to those roles via chat, means for analyzing the results of those responses and checking for harassment tendencies or bias, means for feeding back the analysis results, means for analyzing the content of chats and emails in real time, means for sending an alert to an administrator if signs of harassment or bias are detected, means for saving the detected content in a database so that it can be analyzed later, and means for suggesting appropriate ways of responding to users.

[1730] System configuration

[1731] The server uses a generative AI model to set specific roles, and users can respond to those roles in the chat. The user's responses are analyzed using a natural language processing library (e.g., spaCy). The analyzed data is then checked for harassment and bias using a machine learning model (e.g., scikit-learn's Logistic Regression).

[1732] Data processing and calculation

[1733] The server analyzes the content of users' chats and emails in real time. The analyzed data is converted into TF-IDF vectors and input into a machine learning model. The model detects signs of harassment and bias and sends the results as an alert to an administrator. The detected content is also stored in a database for later detailed analysis.

[1734] Hardware and software used

[1735] Hardware: Servers, user devices (PCs, smartphones, etc.)

[1736] Software: Generative AI model, natural language processing library (spaCy), machine learning library (scikit-learn), database management system

[1737] Specific examples

[1738] For example, if a company were to use this system for customer support chat, they would input the following prompts into the generative AI model:

[1739] Analyze customer support chats in real time to detect signs of harassment and bias. Analyze messages such as:

[1740] "You're really incompetent."

[1741] "Thank you for your amazing support!"

[1742] "You're an idiot."

[1743] Based on this prompt, the generative AI model analyzes the chat content to detect signs of harassment or bias, and if detected, sends an alert to an administrator and suggests appropriate responses.

[1744] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1745] Step 1:

[1746] The server assigns specific roles to the generative AI model. Specifically, it assigns scenarios such as "a subordinate coming out as LGBT" or "a subordinate announcing pregnancy" to the model. Based on this assignment, it generates a prompt for the user to start a chat.

[1747] Input: Specific Role

[1748] Output: prompt statement

[1749] Step 2:

[1750] Users respond to specific roles assigned by the generative AI model through chat, and their input is sent to the server as chat messages.

[1751] Input: User's chat message

[1752] Output: Message sent to the server

[1753] Step 3:

[1754] The server analyzes the received chat messages using a natural language processing library (e.g., spaCy), and performs data processing such as text tokenization and TF-IDF vectorization on the analyzed data.

[1755] Input: User's chat message

[1756] Output: Parsed text data (tokenized, vectorized)

[1757] Step 4:

[1758] The server then feeds the parsed text data into a machine learning model (e.g., scikit-learn's LogisticRegression) to check for harassment and bias. The model makes predictions based on the pre-trained data.

[1759] Input: Parsed text data

[1760] Output: Prediction results of harassment and bias

[1761] Step 5:

[1762] Based on the prediction results, the server sends an alert to the administrator if signs of harassment or bias are detected. The alert includes the content of the detected message and the prediction results.

[1763] Input: Predicted results of harassment and bias

[1764] Output: Alert to administrator

[1765] Step 6:

[1766] The server stores any detected instances of harassment or bias in a database, which is then used for further analysis and reporting.

[1767] Input: Detected harassment or bias

[1768] Output: Data stored in the database

[1769] Step 7:

[1770] The server then proposes appropriate responses to the user, based on generative AI models and pre-defined response guidelines.

[1771] Input: Predicted results of harassment and bias

[1772] Output: Suggestions for the user on how to respond

[1773] Example 3

[1774] Next, a third embodiment of the third embodiment will be described. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1775] With conventional systems, it was difficult to effectively detect bias or harassment tendencies in users' responses and provide specific improvements. Furthermore, when supporting employees returning to work after taking a leave of absence due to mental health issues, there was a lack of means to simulate appropriate responses and provide feedback. This limited opportunities for users to learn and improve appropriate responses.

[1776] The specific processing by the specific processing unit 290 of the data processing device 12 in the third embodiment is realized by the following means.

[1777] In this invention, the server includes a means for the user to input a prompt sentence, a means for the terminal to send the prompt sentence to the server, a means for the server to execute the generative AI model, a means for the server to analyze the results of the generative AI model, a means for the server to generate feedback, and a means for the terminal to display the feedback to the user. This makes it possible to effectively detect bias or harassment tendencies in the user's behavior and provide specific improvements and advice. It is also possible to simulate appropriate response methods and provide feedback to support the return of employees who have taken leave due to mental health issues.

[1778] A "prompt" is the text of an instruction or question that a user enters into a system.

[1779] A "terminal" is a device on which a user inputs a prompt and communicates with a server. Examples include PCs, smartphones, and tablets.

[1780] A "server" is a computer system that runs the generative AI model, analyzes prompts, and generates feedback.

[1781] A "generative AI model" is an artificial intelligence model that simulates a response based on a user's prompt and generates the results.

[1782] "Feedback" refers to specific improvements and advice provided to users after analyzing the results of the generative AI model.

[1783] "Bias" refers to prejudices or preconceived notions contained in user responses.

[1784] "Harassment" refers to inappropriate words, actions, or behavior in a user's interactions with others.

[1785] "Mental health" refers to the state of mental health and is a concept that includes psychological problems such as stress, anxiety, and depression.

[1786] A "return-to-work program" is a plan or procedure for assisting employees who have taken leave due to mental health issues to return to work.

[1787] This invention is a system in which a user inputs a prompt sentence, executes a generative AI model based on the prompt sentence, and provides feedback. A specific embodiment of this system will be described below.

[1788] First, the user inputs a prompt using a terminal. A terminal is a device such as a PC, smartphone, or tablet that provides an interface for users to access the system. For example, a user can input a prompt like the following:

[1789] "How should I respond to a subordinate who comes out as LGBT?"

[1790] How should I respond to an employee returning from leave?

[1791] The device sends the prompts entered by the user to a server, which has the hardware and software to run the generative AI model, specifically, a server equipped with an NVIDIA GPU and a deep learning framework such as TensorFlow or PyTorch.

[1792] The server inputs the received prompt sentence into a generative AI model, which then simulates the user's response and generates the results. This generative AI model then analyzes the prompt sentence using natural language processing technology and suggests an appropriate response.

[1793] The output of the generative AI model is analyzed by a server, which uses natural language processing techniques, such as sentiment analysis and keyword extraction, to analyze the results of the generative AI model and detect bias or harassment in user responses.

[1794] The server generates feedback based on the analysis results, including specific areas for improvement and advice. For example, it might say, "It's important to respond to subordinates who come out as LGBT without prejudice. Specifically, we recommend that you listen carefully to what your subordinates have to say and show understanding."

[1795] Finally, the device displays the feedback received from the server to the user, who can then review the feedback on the device screen and use it as a reference for improving their own response.

[1796] In this way, this invention can effectively detect bias and harassment tendencies in user behavior and provide specific improvements and advice. It can also simulate appropriate response methods and provide feedback when supporting the return of employees who have taken leave due to mental health issues. The flow of the identification process in Example 3 will be explained using Figure 15.

[1797] Step 1:

[1798] The user enters a prompt statement.

[1799] The user inputs a prompt through the device interface. For example, they might input, "Please tell me how to respond to a subordinate who comes out as LGBT." The input prompt is temporarily stored in the device's memory.

[1800] Step 2:

[1801] The terminal sends the prompt to the server.

[1802] The terminal converts the prompt text entered by the user into JSON format and sends it to the server using an HTTP POST request. The destination endpoint is the server's API endpoint. The input is the prompt text, and the output is the request sent to the server.

[1803] Step 3:

[1804] The server runs the generative AI model.

[1805] The server performs data preprocessing on the received prompt sentences in order to input them into the generative AI model. Specifically, it tokenizes and normalizes the text. It then runs the generative AI model to simulate the user's response. The input is the preprocessed prompt sentence, and the output is the simulation result of the generative AI model.

[1806] Step 4:

[1807] The server analyzes the results of the generated AI model.

[1808] The server uses natural language processing techniques to analyze the output of the generative AI model. For example, it performs sentiment analysis of the text and keyword extraction to detect bias in the user's responses. The input is the simulation result of the generative AI model, and the output is the analysis result.

[1809] Step 5:

[1810] The server generates the feedback.

[1811] The server generates feedback based on the analysis results. The feedback includes specific advice and suggestions for improving the user's response. The generated feedback is sent to the terminal in JSON format. The input is the analysis results, and the output is the feedback.

[1812] Step 6:

[1813] The device displays the feedback to the user.

[1814] The device uses an appropriate UI component to display the feedback received from the server to the user. For example, it displays the feedback in a popup window or as a notification. The user can check the feedback and use it to improve their own response. The input is the feedback, and the output is what is displayed to the user.

[1815] (Application example 3)

[1816] Next, a description will be given of Application Example 3 of Form Example 3. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1817] There are issues with responding appropriately when employees have mental health issues or in specific situations. In particular, there is a lack of simulations and feedback for machines working in factories to respond appropriately to employees, which can delay improvements in employee mental health and the working environment.

[1818] The specific processing by the specific processing unit 290 of the data processing device 12 in Application Example 3 is realized by the following means. In this invention, the server includes means for setting a specific role for the generative AI and for the user to respond to that role via chat, means for analyzing the response results and checking for harassment tendencies and bias, means for feeding back the analysis results, means for providing a simulation for machines working in a factory to respond appropriately to employees, and means for analyzing the simulation results and providing improvements and advice. This makes it possible to simulate and provide feedback for appropriate responses when employees have mental health issues or in specific situations.

[1819] "Generative AI" is an artificial intelligence system that assigns specific roles and allows users to respond to those roles.

[1820] "Chat" is a text-based means of communication that allows users to interact with generative AI.

[1821] "Harassment" refers to inappropriate behavior or language directed at an individual or group in the workplace or other environment.

[1822] "Bias" refers to prejudices or preconceived notions held against a particular individual or group.

[1823] "Feedback" is a means of providing evaluations and suggestions for improvement of actions taken by users.

[1824] "Factory machines" refers to automated devices and robots used to perform tasks in a factory.

[1825] "Simulation" refers to the process of virtually recreating a specific situation or role and allowing users to respond within that situation.

[1826] "Mental health" refers to an individual's mental well-being and includes issues such as stress, anxiety and depression.

[1827] "Employee" refers to people who work for a particular company or organization.

[1828] "Areas for improvement" refers to the actions taken by the user that need to be corrected or changed to achieve better results.

[1829] "Advice" refers to specific instructions or suggestions to help users take more appropriate action.

[1830] To implement this invention, it is first necessary to build a system in which specific roles are assigned to generative AI and users respond to those roles via chat. This system has the function of analyzing the user's responses and checking for harassment tendencies and biases. It also includes a function to provide feedback on the analysis results.

[1831] Furthermore, a function that provides simulations to help machines working in factories respond appropriately to employees is also required. The results of these simulations are analyzed and suggestions and advice are provided to improve employee mental health and the working environment.

[1832] Hardware and software used

[1833] Hardware: Automated equipment and robots used in factories (e.g., ABB YuMi, Fanuc CR series)

[1834] Software: OpenAI API, Python

[1835] Data processing and calculation

[1836] server:

[1837] 1. Receive input from the user.

[1838] 2. Send prompts to a generative AI model (OpenAI's GPT-3) to generate a robotic response.

[1839] 3. The generated response is sent back to the generative AI model to analyze the response and generate feedback.

[1840] Device:

[1841] It provides an interface that allows users to interact with generative AI through chat.

[1842] User:

[1843] Responses to specific roles are made via chat and the results are sent to the system.

[1844] Specific examples

[1845] For example, if an employee enters, "I recently took a leave of absence due to work stress. I'm worried about returning to work," the system will act as follows:

[1846] 1. The server receives input from the user and sends prompts to the generative AI model.

[1847] 2. The generative AI model generates a response based on the following prompt:

[1848] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[1849] robot:

[1850] 3. The server sends the generated response back to the generative AI model to analyze the response and generate feedback.

[1851] 4. Feedback is provided to users and appropriate actions and improvements are suggested.

[1852] In this way, simulations and feedback can be provided to enable machines working in factories to respond appropriately to employees.

[1853] The flow of the specific processing in Application Example 3 will be described with reference to FIG.

[1854] Step 1:

[1855] The server receives input from the user. Specifically, it receives text data entered by the user through the chat interface. This input may include, for example, "I recently took a leave of absence due to work stress. I'm worried about returning to work." The input data is sent to the server in text format.

[1856] Step 2:

[1857] The server sends the received user input to the generative AI model. Specifically, it uses OpenAI's API to send the following prompt sentence to the generative AI model:

[1858] Employee: I recently took a leave of absence due to work stress and I'm nervous about returning to work.

[1859] robot:

[1860] Based on this prompt, the generative AI model generates an appropriate response. The input data is sent in the form of a prompt, and the output data is the generated response text.

[1861] Step 3:

[1862] The server receives a response from the generative AI model. Specifically, it receives text data returned by the generative AI model. This response may include, for example, "When you return to work, try setting small goals first." The input data is the response text from the generative AI model, and the output data is that response text.

[1863] Step 4:

[1864] The server sends the generated response back to the generative AI model to analyze the response and generate feedback. Specifically, it sends the following prompt sentence to the generative AI model:

[1865] Please analyze the responses below and offer your suggestions and suggestions for improvement.

[1866] Response: When you return, try setting small goals first.

[1867] feedback:

[1868] Based on this prompt, the generative AI model analyzes the response and generates feedback. The input data is the response text from the generative AI model, and the output data is the feedback text.

[1869] Step 5:

[1870] The server receives feedback from the generative AI model. Specifically, it receives the feedback text returned by the generative AI model. This feedback includes, for example, "This response is appropriate, but it would be good to suggest specific support methods as well." The input data is the feedback text from the generative AI model, and the output data is the feedback text.

[1871] Step 6:

[1872] The server provides the generated feedback to the user. Specifically, the server displays the feedback text in the user's chat interface. Based on this feedback, the user can improve their own response. The input data is the feedback text, and the output data is the feedback displayed to the user.

[1873] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1874] "Example 1"

[1875] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[1876] "Example 2"

[1877] Another embodiment of the present invention is a system in which an emotion engine adjusts the AI's response based on the user's emotional state. In this system, when a user expresses joy or satisfaction to the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​may respond in a way that shares the user's joy or send a message of gratitude to the user. This allows for a more appropriate response based on the user's emotional state.

[1878] "Example 3"

[1879] One embodiment of the present invention is a system that incorporates an emotion engine. In this system, a specific role is assigned to the generative AI, and the user responds to that role via chat. The results of the response are analyzed to check for harassment tendencies and bias. Furthermore, the emotion engine is used to analyze the user's emotional state, and the results are fed back. Specifically, if the user expresses anger or dissatisfaction toward the AI, the emotion engine recognizes that emotion and adjusts the AI's response. For example, the AI ​​can respond in a way that alleviates the user's anger or send an apology message to the user. This allows for more appropriate responses based on the user's emotional state.

[1880] The processing flow of each embodiment will be described below.

[1881] "Example 1"

[1882] Step 1: The user initiates a chat conversation with the AI.

[1883] Step 2: The AI ​​interacts with the user based on the assigned role.

[1884] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[1885] Step 4: The emotion engine analyzes the user's emotional state.

[1886] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[1887] "Example 2"

[1888] Step 1: The user initiates a chat conversation with the AI.

[1889] Step 2: The AI ​​interacts with the user based on the assigned role.

[1890] Step 3: The system analyzes the user's conversations to check for harassment tendencies and bias.

[1891] Step 4: The emotion engine analyzes the user's emotional state.

[1892] Step 5: Adjust the AI's response based on the emotion engine's analysis results.

[1893] Example 1

[1894] Next, a description will be given of Example 1 of Form Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1895] Conventional workplace communication and harassment countermeasures are difficult to simulate in real situations and lack the means to properly grasp users' emotional states. This limits opportunities for supervisors and managers to learn appropriate responses, making it easy for harassment and bias to occur. Furthermore, the lack of feedback based on users' emotional states makes it difficult to improve response skills in the real workplace.

[1896] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1897] In this invention, the server includes a means for assigning specific roles to generative artificial intelligence and allowing users to respond to those roles via communication means, a means for analyzing the results of the responses and checking for harassment tendencies and bias, a means for analyzing the user's emotional state using an emotion engine, and a means for providing feedback on the analysis results. This allows users to improve their ability to respond in the actual workplace and prevent problems such as harassment and bias. Furthermore, providing appropriate feedback according to the u...

Claims

1. means for accepting a user's selection from a plurality of scenarios; means for setting a specific role based on a scenario selected by a user and generating a prompt sentence instructing a response to the role; means for accepting a user input message; a means for generating a response to the user's input message using the generated prompt sentence and a generation system AI; means for outputting the generated response to the user; means for repeating the reception of said input messages, the generation of said responses, and the output of said applications; A prompt sentence that instructs checking the user's response for the presence or absence of harassment tendencies or bias based on the history of the input messages and the responses, and a means for evaluating the user's response using the generative AI; A system including:

2. means for recognizing an emotion of the user using an emotion engine; The system of claim 1 , wherein the means for generating a response uses the recognized emotion of the user, the generated prompt sentence, and the generative AI to generate a response to the user's input message.

3. The system of claim 1 , wherein the multiple scenarios include a user interacting with a "subordinate who comes out as a sexual minority," a user interacting with a "subordinate who announces a pregnancy," and a user interacting with a "subordinate who is worried about their work style."

Images