Information processing device, output method, and program

Abstract

An information processing device according to the present invention comprises a control unit that uses a plurality of response models to acquire responses indicating information pertaining to at least one common theme, and performs downgrade processing on some or all of the responses if a compensation condition, which is a compensation-related condition, is not satisfied. The control unit outputs the response after the downgrade processing.

Description

Information Processing Apparatus, Output Method, and Program

[0001] The present invention relates to an information processing apparatus, an output method, and a program.

[0002] A plurality of response models using a generation model (e.g., a large language model) that generates subsequent data from the input data can generate different data (e.g., responses) for the same input data (e.g., a theme). Patent Document 1 discloses a technique for simulating discussions by agents with different personalities using a large language model (LLM).

[0003] Chi-Min Chan et al., "CHATEVAL: TOWARDS BETTER LLM-BASED EVALUATORS THROUGH MULTI-AGENT DEBATE," arXiv:2308.07201v1 [cs.CL] 14 Aug 2023

[0004] In an information processing apparatus that expects a synergistic effect by using a plurality of response models, more effective data acquisition is required. On the other hand, when acquiring data by using a plurality of response models, it is difficult to know the cost required for acquiring the data before acquiring the data. Therefore, it is required to output the data acquisition result within the range of conditions related to the cost (cost conditions).

[0005] In order to solve the above problems, one aspect of the present invention includes a control unit that uses a plurality of response models to obtain a response indicating information related to at least one common theme, and performs a downgrade process on part or all of the response when a cost condition, which is a condition related to cost, is not satisfied. The control unit outputs the response after the downgrade process, and is an information processing apparatus.

[0006] One aspect of the present invention is an output method that uses a plurality of response models to obtain responses indicating information about at least one common theme, and if the consideration conditions, which are conditions related to consideration, are not met, a downgrade process is performed on part or all of the response, and the response after the downgrade process is output.

[0007] One aspect of the present invention is a program for causing a computer to function as an information processing device, which includes a control unit that uses a plurality of response models to obtain a response indicating information relating to at least one common theme, and performs a downgrade process on part or all of the response if a consideration condition, which is a condition relating to consideration, is not met, and the control unit outputs the response after the downgrade process.

[0008] According to the above embodiment, the data acquisition results can be output within the range of compensation that the user can accept, depending on whether or not the compensation conditions have been met.

[0009] This figure shows an example configuration of the response system according to the first embodiment. This figure shows an example of role information according to the first embodiment. This figure shows an example of knowledge data according to the first embodiment. This is a flowchart showing the operation of the response system in the first embodiment. This figure shows an example of the agenda input screen according to the first embodiment. This figure shows an example of the discussion preparation screen according to the first embodiment. This figure shows an example of the discussion progress screen according to the first embodiment. This figure shows an example of the discussion summary screen according to the first embodiment. This figure shows an example configuration of the response system 1 according to the second embodiment. This figure shows a specific example of the value information table stored by the value information storage unit 111. This figure shows a specific example of the settlement history table stored by the settlement history storage unit 113. This is a specific example of a flowchart showing the flow of the restriction control process in the second embodiment. This figure shows an example of the response screen G5 according to the second embodiment. This figure shows an example of the first restriction screen G6 according to the second embodiment. This figure shows a specific example of the functional configuration of the service provision system 100 that displays the response screen G5 according to the second embodiment. This figure shows an example of the second restriction screen G7 according to the second embodiment. This figure shows an example of the third restriction screen G8 according to the second embodiment. This figure shows a specific example of the functional configuration of the service provision system 100 that displays the response screen G7 or G8 according to the second embodiment. This figure shows an example of the fourth restriction screen G9 according to the second embodiment. This figure shows an example of the fifth restriction screen G10 according to the second embodiment. This figure shows a specific example of the functional configuration of the service provision system 100 that displays the response screen G9 or G10 according to the second embodiment. This figure shows an example of the sixth restriction screen G11 according to the second embodiment. This figure shows a specific example of the functional configuration of the service provision system 100 that displays the response screen G11 according to the second embodiment. This is a schematic block diagram showing the configuration of a computer according to at least one embodiment. This figure shows one aspect of the response system 1 according to this embodiment.

[0010] Several embodiments will be described below with reference to the drawings. The response system according to the following embodiment realizes role-playing of responses by multiple personalities by inputting a theme and a role into a response model. Multiple response models are used in the response system to perform role-playing of responses by multiple personalities. The multiple response models may be implemented using machine learning models that have been implemented and trained using different learning models. The multiple response models may be implemented using machine learning models that have the same learning model but are given different parameters by undergoing different training. The multiple response models may be implemented by giving different instructions (e.g., prompts) to the same generative model. With the implementation described above, the response may be obtained in any way. For example, the response may be generated using a generative model, or it may be selected from a plurality of existing response candidates. In other words, any technology may be applied to realize role-playing of responses by multiple personalities, as long as the response is generated by some different means. Such a response model may be pre-implemented in the response system 1, or it may be implemented in the external server 3.

[0011] The theme is provided by the user. The theme may be provided as, for example, an agenda item or as a topic of casual conversation. The response is the output that the response model obtains regarding the theme. Responses include first-kind responses and second-kind responses. First-kind and second-kind responses are each generated by different roles. A first-kind response is a response generated by a given role. A first-kind response may also be a response generated during the process of generating a second-kind response. A first-kind response may be used as input in obtaining a new first-kind response. A second-kind response is a response generated as the final response to the theme. The role that generates a second-kind response may be defined as a role that does not have a specific occupation or specific gender and age. For example, the role that generates a second-kind response may be defined as a neutral role. For example, if the theme is an agenda item, the first-kind response may be an opinion and the second-kind response may be a conclusion. For example, if the theme is a topic of casual conversation, the first-kind response may be an utterance and the second-kind response may be a summary. In the following explanation, we will use "agenda" as an example of a theme and "opinion" and "conclusion" as examples of responses.

[0012] The response obtained by the response system may be a string of text, an image (still or moving image), or sound (such as voice or music). The following description will focus on the case where the response obtained by the response system is a string of text. The response system may be a web server that accepts access from terminal devices such as PCs connected via a network such as the Internet, or it may be a terminal device with a program installed to enable role-playing of discussions.

[0013] A generative model is a model that receives input data and generates output data that follows that input data. Generative models may be constructed using machine learning models such as Generative Adversarial Networks (GANs) or Transformers. Furthermore, generative models are not limited to models computed by computers, as in the machine learning models mentioned above, but may also include human input and output. A Large-Scale Language Model (LLM) is an example of a generative model that receives natural language text as input and is trained to generate natural language text (strings) that follow the input natural language text. Other types of generative models include text-to-text models that receive string input and output strings, and text-to-image models that receive string input and output images. The input to a generative model is not limited to strings; it may also accept image data or audio data. The input data for an LLM is also called a "prompt".

[0014] A role is an instruction given as input data to a response model (e.g., a generative model), and it represents setting information that describes the persona (character, personality, personality, etc.), occupation, age, background, and other attributes of the person the response model is to roleplay. The generative model generates output data that follows the input data. Therefore, for example, if a response model is formed using a generative model, inputting input data including roles into the generative model will yield output data that would be generated by the person represented by the role. Multiple types of roles may be implemented, for example, by using multiple machine learning models, each with different parameters. Any machine learning model capable of generating a response to an input theme can be used. One concrete example of such a machine learning model is a language model. In this case, multiple types of roles may be implemented, for example, by using multiple language models, each with different parameters.

[0015] The response system may be given the knowledge of the person to be role-played through methods such as RAG (Retrieval-Augmented Generation) and fine-tuning (full fine-tuning, adapter tuning). When using RAG, the data that can be referenced may differ for each role. For example, when generating output data in a generative model according to the role of a lawyer, the response system will reference legal data via RAG. On the other hand, when generating output data in a generative model according to the role of a meteorologist, the response system will reference meteorological data via RAG. When using adapter tuning, adapter models for each role are prepared in advance, and the output data obtained by inputting input data into the adapter corresponding to the specified role is input into the LLM.

[0016] The response system according to this embodiment simulates individuals (agents) with different personalities by assigning different roles to the response model. Hereinafter, the i-th role will be called role Ri, and the person simulated by role Ri will be called agent Ai. When the response system is implemented using a generative model, it can generate output data representing agent A1's statement after receiving agent A2's statement by inputting output data generated as a statement by agent A2 and input data including role R1 into the generative model. The response system can simulate a discussion between multiple agents by, for example, repeating this process. In this case, the response system may have a single generative model simulate all of the multiple agents by assigning all of the multiple roles to a single generative model. Alternatively, the response system may distribute the simulation of multiple agents among multiple generative models by distributing the multiple roles among several generative models. In this way, since the response system simulates a discussion between multiple agents with different settings, users can obtain diverse perspectives from the opinions and conclusions obtained during the discussion process.

[0017] (First Embodiment) Figure 1 is a diagram showing an example configuration of a service provision system 100 according to the first embodiment. In the following description, a system configured using a generative model (particularly LLM) will be mainly described as a specific example of a response model. The service provision system 100 includes a response system 1, a terminal device 2, and an external server 3. The response system 1 has a web server function that accepts access from a terminal device 2 such as a PC connected via a network such as the Internet. In response to instructions from the terminal device 2, the response system 1 uses LLM to realize role-playing of a discussion by multiple agents A on an agenda entered by a user. LLM is an example of a generative model. The response system 1 may use an LLM that is stored in advance, or it may use an LLM on the external server 3. The response system 1 according to the first embodiment simulates a discussion by agents belonging to each group by dividing multiple roles into several groups according to the entered agenda. The response system 1 may be configured as a single information processing device, or it may be configured using multiple information processing devices.

[0018] The response system 1 comprises an input unit 101, an agenda determination unit 102, a role storage unit 103, a viewpoint acquisition unit 104, a selection unit 105, a grouping unit 106, a discussion generation unit 107, an extraction unit 108, and an output unit 109.

[0019] The input unit 101 receives input operations from the terminal device 2. The agenda determination unit 102 prompts the user to input an agenda item and determines the string entered by the user via the input unit 101 as the agenda item.

[0020] The role storage unit 103 stores role information relating to the roles assigned to the LLM. Figure 2 shows an example of role information according to the first embodiment. In the first embodiment, the role storage unit 103 stores the agent name, the LLM to be used, the job title, the persona, the attributes of the knowledge data, and the discussion rules in association. The agent name is a string indicating the name of the agent to be simulated by the role. In the example shown in Figure 2, the agent name and job title are the same, but in other embodiments, the agent name may be a real name or nickname. The LLM to be used indicates the type of LLM (service name, etc.) to which the role is assigned. The job title and persona are strings indicating the job title and persona of the agent to be simulated by the LLM, respectively. The knowledge data is identification information indicating the data referenced by the RAG. In other embodiments, the role information may include other information representing the agent's background, such as family structure and upbringing. The knowledge data may be stored in the response system 1 or on the external server 3. The discussion rules indicate instructions on how to proceed with the discussion by the role. For example, possible rules for discussion include "persuading the other party" and "considering the other party's point of view." Note that roles do not necessarily need to have their own discussion rules.

[0021] The perspective acquisition unit 104 acquires a document representing the agent's perspective on the agenda (perspective document) by inputting a prompt to the LLM, which includes the agenda and the role, to generate the agent's perspective on the agenda. Examples of perspectives include points of contention on the agenda, main points of the argument, viewpoint on the agenda, opinion on the agenda, points of disinterest on the agenda, and concerns on the agenda. The perspective acquisition unit 104 inputs a prompt to the LLM designated as the "LLM to use" in the role storage unit 103.

[0022] The following is an example of a prompt to generate an agent's perspective on an agenda item. In the following example prompt, {} indicates a variable. "# Instructions You are {Agent Name}. Your occupation is {Job Title} and your persona is {Persona}. Using {Knowledge Data} as a reference, please provide one perspective you would consider when discussing the following agenda item. # Agenda {Agenda} # Knowledge Data {Knowledge Data Search Results}" The agenda item included in the prompt is an example of agenda-related data concerning the topic.

[0023] The "search results for knowledge data" are the knowledge data obtained by RAG. For example, the viewpoint acquisition unit 104 generates a vector relating to the combination of role and topic, and searches for a document with a high similarity to the vector from the knowledge data relating to the attributes associated with the role. Figure 3 is a diagram showing an example of knowledge data according to the first embodiment. The knowledge data is stored in the storage device of the response system 1 or in an external storage device. The knowledge data shown in Figure 3 consists of attributes, document vectors, and documents. The attributes of the knowledge data represent attributes relating to the content of the knowledge data, such as company data, legal data, and weather data. The attributes correspond to the attributes of the knowledge data included in the role information shown in Figure 2. The document vector is a vector representation of a document. The document is represented by a string of characters. The viewpoint acquisition unit 104 can then implement RAG by including the searched document in the prompt. Note that if the highest similarity is lower than a predetermined threshold, the "search results for knowledge data" may be left blank.

[0024] The selection unit 105 calculates a relevance score representing the relationship between the perspective sentences for each agent acquired by the perspective acquisition unit 104 and the roles, and selects a predetermined number of roles with high relevance scores as roles to be used for simulating the discussion. The number of roles to be extracted may be predetermined or determined by a threshold of the relevance score. This is expected to prevent the discussion from diverging due to the inclusion of statements from agents that are not very relevant to the content of the discussion. The selection unit 105 may calculate the relevance score using LLM.

[0025] The following is an example of a prompt for calculating the relevance score: "#Instructions Please express the relevance between the person described below and the following perspective using a number between 0 and 10. A higher number indicates a stronger relevance. #About the person The person's occupation is {occupation} and their name is {persona}. #About the perspective {perspective text} #Output format Relevance: <number>" The selection unit 105 can obtain the relevance score by extracting the number written after "Relevance:" from the text output from LLM according to the above prompt.

[0026] In addition, in other embodiments, the selection unit 105 may use a vector space model instead of LLM and calculate the relevance score using other methods, such as calculating the similarity (cosine similarity, etc.) between a vector representing a role and a vector representing a viewpoint text.

[0027] The grouping unit 106 classifies the roles selected by the selection unit 105 into several groups. The grouping unit 106 classifies roles with relatively high similarity between the generated perspective texts into the same group. The grouping unit 106 may calculate a relevance score using LLM.

[0028] The following is an example of a prompt to perform grouping: "#Instructions Refer to the combination of persons and their viewpoints shown below and divide the persons into groups of those with similar viewpoints. Each group should have at least two persons. Each group should be given a group name that represents the common viewpoints. #Persons and viewpoints {Agent A1}'s viewpoint: {Viewpoint 1} {Agent A2}'s viewpoint: {Viewpoint 2} {Agent A3}'s viewpoint: {Viewpoint 3} ... #Output format 1. <Group name>: <Person>, <Person>, ... 2. <Group name>: <Person>, <Person>, ..." The grouping unit 106 can obtain the group name by extracting the string written before the ":" from the text output from LLM according to the above prompt, and can obtain multiple agent names by splitting the string written after the ":" with ",".

[0029] In addition, in other embodiments, the grouping unit 106 may use a vector space model to obtain vectors representing viewpoint sentences generated from each role, and use a method such as clustering to distribute each vector into multiple clusters, thereby grouping roles with similar viewpoints into the same group. In this case, the grouping unit 106 may generate group names by inputting a prompt to the LLM to generate group names that represent commonalities among the viewpoint sentences related to each group.

[0030] The discussion generation unit 107 generates text (spoken text) that simulates a discussion between agents by inputting prompts including the topic and role into the LLM for each group divided by the grouping unit 106. The discussion generation unit 107 inputs prompts into the LLM designated as the "LLM to use" in the role storage unit 103.

[0031] The following is an example of a prompt to simulate a discussion. "#Instructions You are {Agent Name}. Your occupation is {Job Title} and you are a person named {Persona}. Please refer to {Knowledge Data} and continue speaking in the following conversation. When speaking, please follow the {Discussion Rules}. #Topic {Topic} #Knowledge Data {Search Results for Knowledge Data} #Your Perspective on the Topic {Perspective Text} #Previous Conversation {Agent A1}: {Statement Text 1} {Agent A2}: {Statement Text 2} {Agent A3}: {Statement Text 3} {Agent A4}: {Statement Text 4} ..." "Previous Conversation" is a history of statements generated based on the roles of other agents belonging to the same group. Note that when the discussion generation unit 107 generates the first statement, "Previous Conversation" may be blank or a text indicating that there is no statement. The perspective text and statement text generated based on the topic are both examples of topic-related data concerning the topic.

[0032] By inputting the above prompts into the LLM for each role, the speech text of each agent can be obtained. In addition, since the "previous conversation" includes speech texts of agents belonging to the same group, the discussion generation unit 107 can generate speech texts that simulate a discussion between agents belonging to the same group. Note that since the speech texts are generated based on the "previous conversation," the content of the speech texts changes depending on the order of the roles for which prompts are generated, i.e., the order in which the agents speak. The discussion generation unit 107 may determine the order of the roles using a round-robin method. Alternatively, the discussion generation unit 107 may let the LLM determine the order of the roles.

[0033] Below is an example of a prompt for determining the order of roles: "#Instructions You are the moderator of the discussion. Please select the person best suited to continue speaking in the following conversation on the topic of {Agenda} from the following candidates. #Conversation {Agent A1}: {Statement 1} {Agent A2}: {Statement 2} {Agent A3}: {Statement 3} {Agent A4}: {Statement 4} ... #Candidates {Agent A5}: Occupation {Job Type 5}, Persona 5. Holds the viewpoint {Perspective 5} on the agenda. {Agent A6}: Occupation {Job Type 6}, Persona 6. Holds the viewpoint {Perspective 6} on the agenda. ... #Output Format Next person to speak: <Person>" Candidates are roles that have not yet been selected for the creation of the prompt.

[0034] The extraction unit 108 extracts opinions raised in the discussion from multiple statement texts generated by the discussion generation unit 107. The upper limit of the number of opinions to be extracted may be predetermined. The extraction unit 108 generates a statement text (opinion text) representing the opinions raised in the group discussion by inputting a prompt to the LLM that includes the statement texts for each group and the upper limit of the number of opinions.

[0035] Below is an example of a prompt for extracting opinions in a discussion: "#Instructions You are the discussion moderator. Extract up to {maximum number} opinions from the following conversation on the topic {agenda}. #Conversation {Agent A1}: {Statement 1} {Agent A2}: {Statement 2} {Agent A3}: {Statement 3} {Agent A4}: {Statement 4} ... #Output format Opinion 1: <Extracted opinion> Opinion 2: <Extracted opinion> ..."

[0036] The extraction unit 108 determines the display priority for each of the multiple opinion sentences obtained from LLM. The display priority is higher the more agents (number of supporters) take a stance in favor of the opinion expressed in the opinion sentence. In other words, the extraction unit 108 determines the priority based on the number of supporters for the opinion sentence.

[0037] Below is an example of a prompt to determine the priority of opinions: "#Instructions Please tell us the number of people who agree with the opinion "{Opinion Text}" from the following conversation. #Conversation {Agent A1}: {Statement Text 1} {Agent A2}: {Statement Text 2} {Agent A3}: {Statement Text 3} {Agent A4}: {Statement Text 4} ... #Output Format Number of Agreeers: <Number>"

[0038] The output unit 109 generates a display screen showing the status of the discussion among multiple agents and outputs it to the terminal device 2.

[0039] Figure 4 is a flowchart showing the operation of the response system 1 in the first embodiment. When a user accesses the response system 1 via the terminal device 2, the agenda determination unit 102 generates an agenda input screen G1 for the terminal device 2 to input an agenda item and outputs it to the terminal device 2 (step S1).

[0040] Figure 5 shows an example of the agenda input screen G1 according to the first embodiment. The agenda input screen G1 has a first area P1 for displaying the agenda, a second area P2 for displaying the agent's status, a third area P3 for displaying the content of the discussion, and a fourth area P4 for displaying a summary of the discussion. The first area P1 includes an input form P11 for receiving the agenda input. The second area P2 displays a plurality of symbols P21 representing agents. In the first embodiment, the symbols P21 are icon images. In other embodiments, the symbols P21 may be other symbols representing agents, such as avatars, moving images, or strings of text. Each symbol P21 is associated with an agent name. In the agenda input screen G1, the plurality of symbols P21 are arranged randomly. That is, before the agenda is entered into the input form P11, the plurality of symbols P21 are arranged without distinction. When a symbol P21 is selected, a balloon P22 presenting the agent's attribute information (occupation, persona, type of LLM used for simulation, etc.) is displayed near the symbol P21. Since the discussion has not yet begun at this point, the third area P3 and the fourth area P4 on the agenda input screen G1 are blank.

[0041] When the input unit 101 receives input from the user into the input form P11, the agenda determination unit 102 determines the string entered by the user as the agenda (step S2). Next, the viewpoint acquisition unit 104 obtains viewpoint sentences representing the viewpoints of each agent by providing the LLM with a prompt that includes the agenda determined in step S2 and the role of each agent (step S3). The selection unit 105 calculates a relevance score that represents the relationship between each agent's viewpoint sentence and their role, and selects a predetermined number of roles with high relevance scores as roles to be used for simulating the discussion (step S4).

[0042] The grouping unit 106 classifies the roles selected in step S4 into groups based on the similarity between the perspective texts (step S5). Once the roles are classified into groups, the output unit 109 generates a discussion preparation screen G2 and outputs it to the terminal device 2 (step S6).

[0043] FIG. 6 is a diagram showing an example of a discussion preparation screen G2 according to the first embodiment. The discussion preparation screen G2 includes a first area P1 for displaying a topic, a second area P2 for displaying the state of agents, a third area P3 for displaying the content of the discussion, and a fourth area P4 for displaying opinions (opinion texts) raised in the discussion for each group.

[0044] The first area P1 includes a topic label P12 for displaying the input topic. Unlike the input form P11, the character string representing the topic displayed on the topic label P12 cannot be edited. In the second area P2, a plurality of symbols P21 representing agents are displayed. In the second area P2, symbols P21 representing agents belonging to the same group are arranged in the vicinity of each other. That is, after a topic is input into the input form P11, the plurality of symbols P21 are arranged separately for each group to which the corresponding agents belong. In the second area P2, an oval group frame P23 representing a group is arranged, and symbols P21 representing agents belonging to the group represented by the group frame P23 are arranged overlapping the group frame P23. A group name is displayed on the group frame P23. Since the group name is a character string representing the common point of the viewpoints related to the roles belonging to the same group, it is an example of a character string representing the common point of the roles. When the symbol P21 is selected, a balloon P22 is displayed in the vicinity of the symbol P21. An opinion text of the agent represented by the symbol P21 may be displayed in the balloon P22.

[0045] Since the discussion has not started at this point, the third area P3 of the discussion preparation screen G2 is blank. In the fourth area P4, spaces for displaying the opinion texts of each group are arranged side by side. The group name of the corresponding group is displayed above each space. Since the discussion has not started at this point, no opinion text is arranged in the fourth area P4 of the discussion preparation screen G2.

[0046] For each group divided by the grouping unit 106, the response system 1 performs the processing from step S7 to step S10 below. Hereinafter, the processing for one group (target group) among each group will be described. The discussion generation unit 107 selects one role that has not yet been selected for creating a prompt for obtaining a speech text from the target group (step S7). That is, the discussion generation unit 107 selects one agent that has not spoken in the target group. The discussion generation unit 107 selects a role using a round-robin method, a method using an LLM, or the like. The discussion generation unit 107 inputs a prompt including the topic determined in step S2, the role selected in step S7, the perspective text obtained in step S3 for the role, and the speech text previously generated in the target group into the LLM to obtain the speech text (step S8).

[0047] The output unit 109 generates a discussion progress screen G3 and outputs it to the terminal device 2 (step S9). FIG. 7 is a diagram showing an example of the discussion progress screen G3 according to the first embodiment. The discussion progress screen G3 has a first area P1 for displaying a topic, a second area P2 for displaying the state of agents, a third area P3 for displaying the content of the discussion, and a fourth area P4 for displaying opinions (opinion texts) raised in the discussion for each group.

[0048] The first area P1 includes a topic label P12 for displaying the input topic, similar to the discussion preparation screen G2. In the second area P2, a plurality of symbols P21 representing agents are arranged separately for each group, similar to the discussion preparation screen G2.

[0049] In the third area P3, the utterance texts P31 from one of several groups and the symbols P32 representing the agent who is making the utterance are arranged in chronological order. In other words, the utterance texts P31 are arranged in the order in which they were generated. The utterance texts P31 are displayed dynamically according to the passage of time. For example, the utterance texts P31 may be strings output from the LLM and placed in real time. At this time, in the second area P2, a speech bubble P24 indicating that an utterance is in progress is displayed near the symbol P21 corresponding to the role used to create the displayed utterance text P31. In other words, a speech bubble P24 is displayed near the symbol P21 corresponding to the role selected in step S7. The presence or absence of a speech bubble P24 is just one example of the form of the symbol P21. In other embodiments, for example, the symbol P21 making an utterance may be enlarged and displayed, or its form may be made different from other symbols P21 by means of methods other than a speech bubble P24, such as adding an animation of the mouth moving. At this point, the discussion has not yet concluded, so no opinion text is placed in the fourth area P4 of the discussion progress screen G3.

[0050] The discussion generation unit 107 determines whether there are any roles in the target group that have not yet been selected for creating prompts to obtain spoken text (step S10). If there are unselected roles (step S10: YES), the discussion generation unit 107 returns to step S7 and continues generating the next spoken text. On the other hand, if there are no unselected roles (step S10: NO), the discussion generation unit 107 finishes generating spoken text in the target group.

[0051] Once the generation of speech texts for each group is complete, the extraction unit 108 obtains opinion texts for each group by inputting a prompt containing the multiple generated speech texts into the LLM (step S11). The extraction unit 108 then determines the display priority for each of the multiple opinion texts (step S12).

[0052] The output unit 109 generates a discussion summary screen G4 and outputs it to the terminal device 2 (step S13). Figure 8 shows an example of a discussion summary screen G4 according to the first embodiment. The discussion summary screen G4 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, and a fourth area P4 that displays the opinions (opinion texts) raised in the discussion for each group.

[0053] The first area P1 contains a topic label P12 that displays the entered topic, similar to the discussion progress screen G3. The second area P2 contains multiple symbols P21 representing agents, arranged separately for each group, similar to the discussion progress screen G3. The third area P3 contains the spoken text P31 from one of the multiple groups, and the symbol P32 representing the agent who made the spoken word, arranged in chronological order. At the end of the spoken text P31 arranged in the third area P3, a summary status display P33 such as "The discussion has ended, and N opinions have been extracted." is placed. The fourth area P4 contains the opinion texts P41 from each group extracted in step S11, arranged in order of priority determined in step S12.

[0054] (Effects of the First Embodiment) As described above, the response system 1 according to the first embodiment performs the following processing. The viewpoint acquisition unit 104 inputs a prompt (second input data) containing the setting information and agenda for each of the multiple roles given to the generation model to acquire viewpoint text (second output data). The grouping unit 106 divides the multiple roles into multiple groups based on the viewpoint text for each role. The discussion generation unit 107 inputs a prompt (first input data) containing the spoken text generated based on other roles belonging to the same group and the setting information to the generation model to acquire spoken text (first output data). By grouping roles based on the viewpoint text for each role, the response system 1 can simulate discussions among roles that have common viewpoints. As a result, the response system 1 can prevent the direction of the discussion from diverging and simulate discussions that are deepened according to common viewpoints.

[0055] In the first embodiment, the extraction unit 108 extracts one or more opinions on a topic from the spoken texts related to the roles belonging to each group. This allows users to easily recognize opinions obtained as a result of discussions on a set topic. The extraction unit 108 also assigns a priority to each opinion based on the spoken texts for each role. This allows users to use the priority as material when considering multiple opinions.

[0056] In the first embodiment, the selection unit 105 selects some of the roles from among several roles to be used in the discussion based on the perspective text for each role. This allows the response system 1 to simulate a discussion that is deepened according to a common perspective by eliminating roles that would lead to a divergence in the discussion. In particular, the selection unit 105 in the first embodiment selects some roles based on a relevance score that indicates the degree of relevance between the agenda (perspective text obtained from the agenda) and the roles. This makes it possible to exclude roles that are not very relevant to the agenda from the discussion.

[0057] In the first embodiment, the role storage unit 103 can specify the type of LLM for each role. If the LLM associated with role R1 and the LLM associated with role R2 are different, the discussion generation unit 107 will output the utterance text related to role R1 and the utterance text related to role R2 to different LLMs. As a result, the response system 1 can simulate agents with different personalities not only based on the role but also on the differences in the characteristics of the LLMs.

[0058] The screen displayed by the response system 1 according to the first embodiment has the following features. The first area P1 of the screen contains the agenda set to obtain the spoken text (output data) generated by the generation model. The second area P2 contains multiple symbols P21 representing multiple roles assigned to the generation model. The third area P3 contains multiple spoken texts generated by the generation model based on each of the multiple roles and the agenda. This allows the response system 1 to visually present to the user the progress of a discussion among multiple agents A.

[0059] (Second Embodiment) The response system 1 according to the second embodiment outputs a response depending on whether the user of the terminal device 2 has met the conditions regarding compensation (hereinafter referred to as "compensation conditions"). Specifically, the compensation conditions may be defined as the compensation that can be obtained from the user of the terminal device 2 being sufficient compared to the compensation that would normally be required for the user's request. In this case, if the compensation is insufficient, the response system 1 may perform a downgrade process so that the output content fits within the compensation conditions that the user can satisfy. "Fits within the compensation conditions" means, for example, in the case of text, the output may be stopped midway or summarization may be performed so that the number of words output is within the range that satisfies the compensation conditions. The same applies to moving images. If the compensation is insufficient, the response system 1 may perform a downgrade process so that the amount of information contained in the output content fits within the compensation conditions that the user can satisfy and approaches the amount when the compensation conditions are satisfied. "Amount of information" refers to the amount of information that the user can obtain by accessing that information. "Information volume" could be, for example, the number of words, the data size, the structure such as chapters, the length of the audio or video, or the number of spoken characters in the audio. "Approaching the conditions for compensation" means, for example, in the case of text, the output may be stopped midway or summarization may be performed so that the number of words output is within the range that satisfies the conditions for compensation. The same applies to videos.

[0060] The amount of information contained in the output may be implemented in such a way that the semantic similarity between the two outputs increases depending on whether the user is satisfied with the price conditions or not. For example, if the output is text information, the text can be vectorized (either as word vectors like Gag-of-words or embedding vectors used in NNs), and the cosine similarity can be calculated. Such downgrading can be achieved, for example, by using an LLM to summarize the text so that it fits within a predetermined number of characters while preserving as much meaning as possible.

[0061] One specific example of such processing is an output that differs from the output when the compensation conditions are met, but is closer to the output when the compensation conditions are met. For example, part or all of the response (for example, part or all of the first type response including the response) is output through downgrade processing. Downgrade processing is a process in which the response itself is generated, but the grade of the output is lowered. One specific example of output due to downgrade processing is a manner in which the resolution of the output characters or images is lowered compared to normal output, the number of colors used is reduced, or the color difference between the output characters or images and the surrounding image is reduced. One specific example of output due to downgrade processing is a manner in which part or all of the opinions and conclusions that should be output are not displayed, or other characters or images are displayed in place of part or all of the opinions and conclusions that should be output. Such output may be achieved by performing all the processing in response to the request, but not outputting part or all of the opinions and conclusions obtained by that processing.

[0062] The following are specific examples of downgrade processing. First, let's consider the case where the response output is text. In this case, the text output may be displayed in a way that hides the text output of the response by overlaying (superimposing) other characters or images on top of it. The text may be displayed blurred. The text may be displayed with a mosaic effect. The content of the text may be displayed in a summarized form. Next, let's consider the case where the response output is a still image. In this case, the still image may be cropped and only a portion of it may be displayed. The resolution of the still image may be reduced for display. The still image may be displayed blurred. The still image may be displayed with a mosaic effect.

[0063] Next, we will describe the case where the response output is acoustic. In this case, the acoustic output may be output with a modified frequency range. Such downgrading is effective when the acoustic quality (e.g., voice quality) is important information for the user. Part or all of the acoustic output may be drowned out by other acoustic sounds. Specific examples of other acoustic sounds include voices or simple sounds (such as beeps). Part or all of the acoustic output may be omitted. The original string represented by the voice may be summarized and output.

[0064] Next, we will explain the case where the response output is a video. In this case, the video may be output as a digest version summarizing the content of the video. The range of sounds contained in the video may be changed before output. Such downgrade processing is effective when the sound quality of the sounds contained in the video (e.g., voice quality) is information necessary for the user. Some or all of the sounds contained in the video may be drowned out by other sounds during output. Specific examples of other sounds include voices and simple sounds (such as beeps). The video may be cropped and only a portion of it may be displayed. The video may be displayed with reduced resolution. The video may be displayed blurred. The video may be displayed with mosaic processing applied.

[0065] The extent of the downgrade to be implemented may be determined according to the degree to which the above-mentioned compensation conditions are not met. In other words, the extent of the downgrade to be implemented may be determined according to the degree to which the compensation obtainable from the user of terminal device 2 is insufficient compared to the compensation that would normally be required (hereinafter referred to as "shortfall"). For example, the number of opinions and conclusions subject to downgrade processing may be determined according to the shortfall. Specifically, the greater the shortfall, the more opinions or conclusions may be determined to be subject to downgrade processing. For example, the content of the downgrade processing may be determined according to the shortfall. Specifically, the greater the shortfall, the more the content of the downgrade processing may be determined to make the response output more unclear. More specifically, for example, the greater the shortfall, the lower the resolution of characters and images may be reduced. For example, the greater the shortfall, the fewer colors may be used in characters and images. For example, the greater the shortfall, the closer the color difference between the output characters and images and the surrounding images may be.

[0066] Figure 9 shows an example of the configuration of the response system 1 according to the second embodiment. The response system 1 according to the second embodiment further includes a consideration acquisition unit 110, a value information storage unit 111, a limit control unit 112, a settlement history storage unit 113, and a settlement control unit 114, in addition to the configuration of the response system 1 according to the first embodiment.

[0067] The consideration acquisition unit 110 acquires the value of the consideration necessary to perform processing in accordance with the instructions received from the terminal device 2. Such consideration may be determined according to the number of unit data (e.g., tokens) used in the processing of a generative model such as LLM. It is desirable that the rules for determining the consideration are predetermined in the response system 1. The consideration acquisition unit 110 acquires the necessary value of the consideration based on the instructions received from the terminal device 2 and the rules for determining the consideration.

[0068] The value information storage unit 111 stores value information for each user of the terminal device 2. Figure 10 shows a specific example of the value information table stored by the value information storage unit 111. The value information storage unit 111 stores user IDs in association with value information. The user ID is identification information that indicates a user of the terminal device 2. The value information stored in association with the user ID indicates the value of the value information that the user indicated by that user ID possesses.

[0069] Value information is something that can be used as consideration when obtaining output from response system 1. It is desirable that the value information be electronically settlementable via the network. Value information may be, for example, the balance of a user's bank account, a ticket purchased in advance by the user, electronic money acquired in advance by the user, points acquired by the user in exchange for using other services, or a balance of an amount that is set to be payable for a predetermined period (e.g., weekly or monthly). Value information may also be, for example, NFTs or cryptocurrencies. Value information may be composed of any other thing that can be used as consideration. In this embodiment, the act of transferring value information to the other party is called payment, and the acceptance of payment by the other party and the completion of the transaction is called settlement.

[0070] The restriction control unit 112 determines whether the user requesting the processing meets the payment conditions based on the value of the payment obtained by the payment acquisition unit 110 and the value of the user's value information stored in the value information storage unit 111. For example, the restriction control unit 112 may determine whether the user requesting the processing is able to pay the payment. If the restriction control unit 112 determines that payment is possible, it determines that the payment conditions are met and notifies the settlement control unit 114 accordingly. On the other hand, if the restriction control unit 112 determines that payment is not possible, it determines that the payment conditions are not met and performs the restriction processing. The restriction processing is the process of performing the downgrade processing described above on some or all of the opinions and conclusions. The restriction processing can be implemented in any way. For example, the restriction control unit 112 may perform the restriction processing by restricting some or all of the processing of the agenda determination unit 102, the viewpoint acquisition unit 104, the selection unit 105, the grouping unit 106, the discussion generation unit 107, and the extraction unit 108 during the execution process of processing in response to the user's request. For example, the restriction control unit 112 may perform restriction processing by restricting the output in the output unit 109 after all processing in response to the user's request has been executed. For example, the output unit 109 may be controlled so that some or all of the opinions and conclusions obtained through processing in response to the user's request are not output.

[0071] The payment history storage unit 113 stores the history of payments made in the response system 1. Figure 11 shows a specific example of the payment history table stored by the payment history storage unit 113. The payment history storage unit 113 stores the payment date and time, user ID, and value information in association with each other. The payment date and time indicates the date and time the payment was made. The user ID is the user ID of the user who made the payment. The value information stored in association with the payment date and time and user ID indicates the value of the value information used in the payment made by the user indicated by that user ID at the date and time indicated by the payment date and time.

[0072] The settlement control unit 114 settles the payment for the usage of the response system 1 with the user. The settlement control unit 114 completes the settlement by receiving payment of the payment acquired by the payment acquisition unit 110, for example, according to the determination result of the restriction control unit 112. For example, the settlement control unit 114 records the value of the payment to be made, the user ID of the user making the payment, and the date and time of payment in the settlement history storage unit 113. Furthermore, the settlement control unit 114 subtracts the value of the payment to be made from the value of the value information associated with the user ID of the user making the payment in the value information storage unit 111, and updates the record with the value after the subtraction.

[0073] Figure 12 is a specific example of a flowchart showing the flow of the restriction control process in the second embodiment. First, the consideration acquisition unit 110 acquires the value of the consideration required for the process requested by the user (step S101). The restriction control unit 112 acquires the value of the value information held by the user who requested the process by reading information from the value information storage unit 111 (step S102). Based on the acquired value of the consideration and the value of the value information, the restriction control unit 112 determines whether or not the user who requested the process can pay the consideration (step S103).

[0074] If payment of consideration is possible (step S104-YES), the restriction control unit 112 decides not to impose a restriction and notifies the settlement control unit 114 of the value of consideration obtained by the consideration acquisition unit 110. The settlement control unit 114 performs settlement processing based on the value of consideration obtained by the consideration acquisition unit 110 (step S105). On the other hand, if payment of consideration is not possible (step S104-NO), the restriction control unit 112 performs restriction processing (step S106). By performing restriction processing, some or all of the opinions and conclusions are downgraded and output. The restriction control unit 112 notifies the settlement control unit 114 that restriction processing has been performed. The settlement control unit 114 performs settlement processing for consideration that is less than the value of consideration obtained by the consideration acquisition unit 110 and is payable by the user (step S107). In such settlement processing, for example, settlement may be performed to pay the full amount the user has (all the value of the value information stored in the value information storage unit 111), or settlement may be performed to pay only a portion of it. The restriction control process described above is executed in the processing flow of the response system 1 after the agenda has been decided (step S2 in Figure 4) and before all opinions and conclusions are output (step S13).

[0075] Next, a specific example of the output of the response system 1 according to the second embodiment will be described. Figure 13 is a diagram showing an example of the response screen G5 according to the second embodiment. The response screen G5 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the discussion for each group, and a fifth area P5 that displays the content of the conclusion. The first to fourth areas P1 to P4 may be the same as the discussion summary screen G4 shown in Figure 8. The fifth area P5 contains a conclusion statement P51 that shows the conclusion of the discussion held by multiple groups. The conclusion of the discussion may be obtained, for example, by the processing of the discussion generation unit 107 or by the processing of the extraction unit 108.

[0076] Below is an example of a prompt for extracting conclusions from a discussion. "#Instructions You are the decision-maker who will determine the conclusion of the discussion. Based on the decision-maker's past statements shown below, please select the one opinion that you consider most important from among the opinions raised in the discussion by multiple groups. #Agenda {Agenda} #Groups and Opinions Group 1. {Group Name} 1-1. {Opinion Text} 1-2. {Opinion Text} Group 2. {Group Name} 2-1. {Opinion Text} 2-2. {Opinion Text} ... #Decision-maker's Past Statements {Information on past statements by users and decisions made by past users} #Output Format <Opinion Number>"

[0077] On response screen G5, the restriction image from the restriction control unit 112 is not displayed. In other words, response screen G5 is one specific example of a screen displayed when the user is able to pay the required price. On response screen G5, all opinions and conclusions obtained through the processing of the response system 1 are output.

[0078] Figure 14 shows an example of the first restriction screen G6 according to the second embodiment. The first restriction screen G6 is a specific example of a screen that is displayed when restriction processing is performed. The first restriction screen G6 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the discussion for each group, and a fifth area P5 that displays the content of the conclusion. The first to third areas P1 to P3 may be the same as the discussion summary screen G4 shown in Figure 8.

[0079] In the fourth area P4, the opinion texts P41 extracted in step S11 are arranged in order of priority determined in step S12. In the fourth area P4 of the first restriction screen G6, a restriction image P42 is further superimposed on part or all of the opinion texts P41. The placement of the restriction image P42 makes part or all of the opinion texts P41 invisible to the user. The restriction image P42 may display text or images indicating that the output is restricted. In the example in Figure 14, the restriction image P42 displays the text "An additional ticket is required to view the opinions after the interruption. Purchase additional tickets here." Part or all of this text may contain a link to a website where additional tickets can be purchased. In the example in Figure 14, the three characters "here" are linked. In this example in Figure 14, tickets are used as a concrete example of value information.

[0080] The fifth area P5 contains an interruption conclusion document P52, which shows the conclusion reached when a discussion among multiple groups is interrupted midway. The fifth area P5 of the first restriction screen G6 also contains a restriction image P53. The restriction image P53 may display text or images indicating that the output is restricted. In the example in Figure 14, the restriction image P53 displays the text "Due to insufficient tickets, the discussion has been interrupted and a conclusion has been reached. ×× tickets are required to view the complete discussion conclusion."

[0081] The "××" part indicates the value of the missing value information (the number of tickets missing). This value of missing value information may be calculated, for example, as the difference between the value of the consideration obtained by the consideration acquisition unit 110 and the value of the value information held by the user who requested the processing. Also, although the restriction image P53 contains the string "The discussion was interrupted midway and a conclusion was reached," it is not necessarily required to actually interrupt the discussion midway, and the processing of the discussion in the response system 1 may be executed to the end. Then, assuming that the discussion was interrupted midway, the conclusion obtained may be obtained and the obtained conclusion may be output as the interruption conclusion text P52.

[0082] Figure 15 shows a specific example of the functional configuration of a service provision system 100 that displays a response screen G5 according to the second embodiment. In Figure 15, a topic is given by the user as a theme and a discussion takes place, and text is obtained as output. A ticket is used as a specific example of value information. The service provision system 100 includes, as functions, an account information storage unit 401, a ticket information storage unit 402, a ticket management unit 403, a token unit price storage unit 404, a ticket usage calculation unit 405, a display content control unit 406, and a discussion processing unit 409. The display content control unit 406 includes a non-display range setting unit 407 and a screen generation unit 408.

[0083] The account information storage unit 401 stores information about the user account of the terminal device 2. The value information storage unit 111 of the response system 1 may be implemented in the same way as the account information storage unit 401. The account information storage unit 401 may be provided on the external server 3. The account information may be, for example, information that associates user identification information with the number of tickets the user has.

[0084] The ticket information storage unit 402 stores information indicating the value of the ticket and the price required to purchase it. The number of tickets purchased and the price are proportional, but discounts may be applied if discount conditions such as set discounts are met. A ticket is one specific example of value information. The ticket information storage unit 402 may be provided, for example, on an external server 3.

[0085] The ticket management unit 403 processes the user to purchase a new ticket by performing an online payment based on the information stored in the ticket information storage unit 402 in response to the user's instructions (instructions for purchase processing) from the terminal device 2. The number of tickets held by the user increases with the purchase. The ticket management unit 403 records the number of tickets after purchase in the account information storage unit 401. The ticket management unit 403 receives notification from the ticket usage calculation unit 405 of the number of tickets used. The ticket management unit 403 subtracts the number of tickets used from the number of tickets stored in the account information storage unit 401. The ticket management unit 403 may be provided in the response system 1 or in the external server 3.

[0086] The token unit price storage unit 404 stores the unit price of the tokens required for processing in the response system 1. The unit price of the tokens may be expressed, for example, as the number of tickets (amount of value information) required to use a unit amount of tokens (e.g., 1 opinion token). The token unit price storage unit 404 may be provided in the response system 1. More specifically, the consideration acquisition unit 110 of the response system 1 may be implemented as the token unit price storage unit 404.

[0087] The ticket usage calculation unit 405 receives notification of the number of tokens used by the discussion processing unit 409 and calculates the number of tickets used based on the number of tokens used and the token price. The ticket usage calculation unit 405 obtains the number of tickets owned by the user from the ticket management unit 403 and obtains the number of tickets that are insufficient. The ticket usage calculation unit 405 notifies the ticket management unit 403 of the number of tickets used. The payment acquisition unit 110 of the response system 1 may be implemented in the same way as the ticket usage calculation unit 405.

[0088] The display content control unit 406 controls the generation of screens according to the discussion and the number of tickets owned by the user. For example, the response screen G5 shown in Figure 13 is generated by the display content control unit 406. The output unit 109 and the limit control unit 112 of the response system 1 may be implemented in the same way as the display content control unit 406.

[0089] The non-display range setting unit 407 sets the range of discussion results (responses) acquired by the response system 1 to be displayed and not displayed. The limit control unit 112 of the response system 1 may be implemented in the same way as the non-display range setting unit 407.

[0090] The screen generation unit 408 generates a screen that displays the discussion results (responses) obtained by the response system 1. The output unit 109 of the response system 1 may be implemented in the same way as the screen generation unit 408.

[0091] The discussion processing unit 409 receives a theme input from the user's terminal device 2 and performs processing to obtain a discussion result (response) corresponding to the theme. The agenda determination unit 102, viewpoint acquisition unit 104, selection unit 105, grouping unit 106, discussion generation unit 107, and extraction unit 108 of the response system 1 may be implemented in the same way as the discussion processing unit 409. The discussion processing unit 409 notifies the ticket usage calculation unit 405 of the number of tokens required for executing the processing to obtain the discussion result (response). The discussion processing unit 409 outputs the content of the discussion result (response) obtained by the processing of the discussion processing unit 409 (e.g., text data) to the display content control unit 406.

[0092] Next, the processing flow of each function shown in Figure 15 will be explained. The ticket management unit 403 reads ticket information and purchases tickets. The ticket management unit 403 updates the number of purchased tickets in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The discussion processing unit 409 conducts the discussion and obtains responses (opinions and conclusions). The discussion processing unit 409 outputs the total number of tokens of the responses (opinions and conclusions) obtained as a result of the discussion to the ticket usage calculation unit 405. The discussion processing unit 409 outputs the text of the discussion results to the display content control unit 406. The ticket usage calculation unit 405 calculates the number of tokens that can be output based on the number of tickets owned by the user and the token unit price. The ticket usage calculation unit 405 calculates the number of missing tokens based on the number of outputtable tokens and the number of tokens of the discussion results from the discussion processing unit 409. Based on the calculation result of the number of missing tokens, the ticket usage calculation unit 405 outputs the number of tokens to be displayed and the number of tokens not to be displayed from the discussion results.

[0093] The hidden range setting unit 407 determines the range to be hidden from the agenda results according to the number of insufficient tickets and outputs it to the screen generation unit 408. The screen generation unit 408 displays the discussion results, including the hidden portion. The user checks the displayed discussion results and instructs to purchase additional tickets to view the entire discussion result. In response to this instruction, the ticket management unit 403 calculates the total number of tickets by adding the number of tickets the user has purchased and the number of tickets the user already possesses, and records it in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The ticket usage calculation unit 405 subtracts the insufficient number of tickets from the total number of tickets and the number of insufficient tickets. The ticket usage calculation unit 405 outputs the result that the number of insufficient tickets has become zero. The screen generation unit 408 unhides the hidden portion and displays the full text of the discussion result. As a result, the user can view the entire text, including the previously hidden portion.

[0094] Figure 16 shows an example of a second restriction screen G7 according to the second embodiment. The second restriction screen G7 is a specific example of a screen displayed when restriction processing is performed. The second restriction screen G7 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the discussion for each group, and a fifth area P5 that displays the content of the conclusion. The first to third areas P1 to P3 may be the same as the discussion summary screen G4 shown in Figure 8. The fifth area P5 may be the same as the response screen G5 shown in Figure 13. In the example of the second restriction screen G7 shown in Figure 16, the conclusion (second type response) is output preferentially, and the opinions (first type response) are subjected to downgrade processing first.

[0095] In the fourth area P4, the opinion texts P41 from each group extracted in step S11 are arranged in order of priority determined in step S12. In the fourth area P4 of the second restriction screen G7, one or more restriction images P43 are further superimposed on a portion of the opinion texts P41. Each restriction image P43 may be superimposed on, for example, individual opinion texts P41. By placing such restriction images P42, a portion of the opinion texts P41 becomes invisible to the user. The restriction image P42 may display text or images indicating that the output is restricted. In the example in Figure 16, the text "An additional ticket is required to view this opinion." is displayed on the restriction image P42. In this example in Figure 16, a ticket is used as a concrete example of value information.

[0096] The opinion texts P41 to which the restriction image P42 is displayed may be selected in any way depending on their number. For example, from all opinion texts P41 that would normally be displayed without the restriction image P42, opinion texts P41 to be subject to the restriction image P42 may be randomly selected. The selection is made by the restriction control unit 112. Opinion texts P41 to be subject to the restriction image P42 may also be selected based on other criteria. For example, the restriction control unit 112 may select opinion texts P41 according to the importance of their opinions. Importance may be determined based on, for example, their importance in reaching a conclusion, the user's subjective importance (e.g., whether or not it aligns with the user's interests), the degree to which agreement was reached with other groups during the discussion, whether or not it was a turning point in the discussion, or whether or not it was a majority opinion. The above-mentioned criteria for importance are merely examples, and other criteria may be used.

[0097] Figure 17 shows an example of the third restriction screen G8 according to the second embodiment. The third restriction screen G8 is a specific example of a screen displayed when restriction processing is performed. In the example of the third restriction screen G8 shown in Figure 17, the conclusion (second type response) is output preferentially, and the opinion (first type response) is subjected to downgrade processing first. The difference between the third restriction screen G8 and the second restriction screen G7 is that the restriction image P43 is superimposed on all opinion sentences P41, not just some of the opinion sentences P41. The third restriction image G8 will be explained below for clarification.

[0098] The third restriction screen G8 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the group discussion, and a fifth area P5 that displays the content of the conclusion. The first to third areas P1 to P3 may be the same as the discussion summary screen G4 shown in Figure 8. The fifth area P5 may be the same as the response screen G5 shown in Figure 13.

[0099] In the fourth area P4, the opinion texts P41 from each group extracted in step S11 are arranged in order of priority determined in step S12. In the fourth area P4 of the second restriction screen G7, one or more restriction images P43 are further superimposed on all of the opinion texts P41. Each restriction image P43 may be superimposed on, for example, each individual opinion text P41. By placing such restriction images P42, the entirety of the opinion texts P41 becomes invisible to the user. The restriction image P42 may display text or images indicating that the output is restricted. In the example in Figure 16, the text "An additional ticket is required to view this opinion." is displayed on the restriction image P42. In this example in Figure 16, a ticket is used as a concrete example of value information.

[0100] Figure 18 is a diagram showing a specific example of the functional configuration of a service provision system 100 that displays a response screen G7 or G8 according to the second embodiment. The functional configuration shown in Figure 18 differs from that in Figure 15 in that it further includes a response output ticket count storage unit 410. The response output ticket count storage unit 410 stores the value of the response output ticket count. The response output ticket count indicates the number of tickets required to output all responses obtained by the discussion processing unit 409. The response output ticket count may be determined in advance as a fixed value, rather than being determined according to the number of tokens or content of the response.

[0101] Next, the processing flow of each function shown in Figure 18 will be explained. The ticket management unit 403 reads ticket information and purchases tickets. The ticket management unit 403 updates the number of purchased tickets in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The discussion processing unit 409 conducts the discussion and obtains responses (opinions and conclusions). The discussion processing unit 409 outputs the total number of opinion tokens from the responses obtained as a result of the discussion to the ticket usage calculation unit 405. The discussion processing unit 409 outputs the text of the discussion results to the display content control unit 406. The ticket usage calculation unit 405 reads the number of response output tickets from the response output ticket number storage unit 410, obtains the number of tickets required to display the full text of the response from the number of response output tickets, and converts it into the number of tokens. The ticket usage calculation unit 405 adds the number of tokens obtained by the conversion and the total number of opinion tokens, and calculates the number of tokens that can be output from the number of tickets owned by the user and the token unit price. The ticket usage calculation unit 405 calculates the number of missing tokens based on the number of tokens that can be output and the number of tokens in the discussion results from the discussion processing unit 409. Based on the calculation result of the number of missing tokens, the ticket usage calculation unit 405 outputs the number of tokens to display and the number of tokens not to display from the opinions shown in the discussion results.

[0102] The hidden range setting unit 407 determines the range of opinions to be hidden according to the number of insufficient tickets and outputs it to the screen generation unit 408. At this time, the hidden range setting unit 407 may randomly determine which opinions to hide according to the number of insufficient tickets. The screen generation unit 408 displays the opinions, including the hidden parts. The user checks the displayed discussion results and instructs to purchase additional tickets to view all of the opinions. In response to this instruction, the ticket management unit 403 calculates the total number of tickets by adding the number of tickets the user has already purchased and the number of tickets the user already possesses, and records it in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The ticket usage calculation unit 405 subtracts the insufficient number of tickets from the total number of tickets and the number of insufficient tickets. The ticket usage calculation unit 405 outputs the result that the number of insufficient tickets has become zero. The screen generation unit 408 unhides the hidden parts and displays the full text of the opinions. As a result, the user can view the full text, including the parts that were hidden.

[0103] Figure 19 is a diagram showing an example of the fourth restriction screen G9 according to the second embodiment. The fourth restriction screen G9 is a specific example of a screen that is displayed when restriction processing is performed. The fourth restriction screen G9 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, and a fourth area P4 that displays the opinions (expressed opinions) that were raised in the discussion for each group.

[0104] The first area P1 and the fourth area P4 may be the same as the discussion preparation screen G2 shown in Figure 6. The second area P2 displays the same information as the second area P2 in the discussion preparation screen G2 shown in Figure 6, plus a consumption value information image P24. The consumption value information image P24 is placed inside or near each group frame P23. The consumption value information image P24 shows the value of the value information that is consumed (required as consideration) when a discussion is conducted within that group and its opinions are displayed. The third area P3 displays a restriction image P34 and an ownership value information image P35. The restriction image P34 shows a string of characters or an image indicating the content of the restriction process. In the example in Figure 19, the restriction image P34 shows the string of characters "Some groups cannot participate in the discussion due to insufficient tickets. Please select which groups will participate from the remaining tickets." The ownership value information image P35 shows the value of the remaining value information held by the requesting user. In this example in Figure 19, tickets are used as a concrete example of value information. Furthermore, the area in which the restriction image P34 and the ownership value information image P35 are displayed does not have to be limited to the third area P3. For example, they may be displayed in the second area P2, the fourth area P4, or any other area.

[0105] In the fourth restriction screen G9 shown in Figure 19, the user selects a group to participate in the discussion within the scope of the value information they possess. If the compensation required by the group selected by the user is within the scope of the user's value information (a payable value), the response system 1 generates and outputs a response screen G5 that includes the opinions and conclusions reached in the discussion of the selected group.

[0106] Figure 20 shows an example of the fifth restriction screen G10 according to the second embodiment. The fifth restriction screen G10 is a specific example of a screen displayed when restriction processing is performed. The difference between the fifth restriction screen G10 and the second restriction screen G7 is that the restriction images P43 are not arranged for each individual opinion document P41, but rather are arranged in groups and superimposed on the opinion documents P41. The fifth restriction image G10 will be explained below for clarification.

[0107] The third restriction screen G8 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the group discussion, and a fifth area P5 that displays the content of the conclusion. The first to third areas P1 to P3 may be the same as the discussion summary screen G4 shown in Figure 8. The fifth area P5 may be the same as the response screen G5 shown in Figure 13.

[0108] In the fourth area P4, the opinion texts P41 from each group extracted in step S11 are arranged in order of priority determined in step S12. In the fourth area P4 of the second restriction screen G7, one or more restriction images P43 are further superimposed on all of the opinion texts P41 of the restricted groups. In the example in Figure 20, all of the opinion texts P41 from the legal group are not visible to the user. The restriction image P42 may show text or images indicating that the output is restricted. In the example in Figure 20, the text "An additional ticket is required to view the opinions of this group." is shown on the restriction image P43. In this example in Figure 20, a ticket is used as a concrete example of value information.

[0109] Figure 21 is a diagram showing a specific example of the functional configuration of the service provision system 100 that displays the response screen G9 or G10 according to the second embodiment. The functional configuration shown in Figure 21 differs from that in Figure 15 in that the display content control unit 406 further comprises an input reception unit 411. The input reception unit 411 sets the number of tickets required to display an opinion for each agent participating in the discussion in the discussion processing unit 409. The user selects which agent's opinion to display while looking at the tickets they possess. The input reception unit 411 outputs the agent selected by the user to the non-display range setting unit 407 and outputs the total number of tickets required for the agent selected by the user to the ticket usage calculation unit 405.

[0110] Next, the processing flow of each function shown in Figure 21 will be explained. The ticket management unit 403 reads the ticket information and purchases tickets. The ticket management unit 403 updates the number of purchased tickets in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The discussion processing unit 409 conducts the discussion and obtains responses (opinions and conclusions). The discussion processing unit 409 outputs the total number of tokens of the responses (opinions and conclusions) obtained as a result of the discussion to the ticket usage calculation unit 405. The discussion processing unit 409 outputs the text of the discussion results and the agents who participated in the discussion to the display content control unit 406. The ticket usage calculation unit 405 calculates the number of tickets required if all agents participate. The ticket usage calculation unit 405 calculates the number of tickets that are missing based on the calculated number of tickets and the number of tickets the user has.

[0111] The input reception unit 411 outputs the agent selected by the user to the hidden range setting unit 407 and outputs the total number of tickets required for the agent selected by the user to the ticket usage calculation unit 405. The hidden range setting unit 407 determines the range in which the opinions of the agent selected by the user as "hidden" will be hidden and outputs it to the screen generation unit 408. The screen generation unit 408 displays the discussion results, including the hidden parts. The user checks the displayed discussion results and instructs to purchase additional tickets to view all of the discussion results. In response to this instruction, the ticket management unit 403 calculates the total number of tickets by adding the number of tickets the user has already purchased and the number of tickets the user already possesses, and records it in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The ticket usage calculation unit 405 subtracts the shortage from the total number of tickets and the number of insufficient tickets. The ticket usage calculation unit 405 outputs the result that the number of insufficient tickets has become zero. The screen generation unit 408 unlocks the hidden portion and displays the full text of the discussion results. As a result, the user can view the entire text, including the previously hidden portion.

[0112] Figure 22 shows an example of the sixth restriction screen G11 according to the second embodiment. The sixth restriction screen G11 is a specific example of a screen that is displayed when restriction processing is performed. The sixth restriction screen G11 has a first area P1 that displays the agenda, a second area P2 that displays the status of the agents, a third area P3 that displays the content of the discussion, a fourth area P4 that displays the opinions (opinion statements) raised in the discussion for each group, and a fifth area P5 that displays the content of the conclusion. The first to third areas P1 to P3 may be the same as the discussion summary screen G4 shown in Figure 8.

[0113] In the fourth area P4, the opinion texts P41 extracted in step S11 are arranged in order of priority determined in step S12. In the fourth area P4 of the first restriction screen G6, a restriction image P42 is further superimposed on part or all of the opinion texts P41. The placement of the restriction image P42 makes part or all of the opinion texts P41 invisible to the user. The restriction image P42 may display text or images indicating that the output is restricted. In the example in Figure 22, the text "An additional ticket is required to view individual opinions." is displayed on the restriction image P42. In this example in Figure 22, a ticket is used as a concrete example of value information.

[0114] In the fifth area P5, a summary conclusion document P54 is placed, which shows a summary of the conclusions obtained from discussions conducted by multiple groups. In the fifth area P5 of the sixth restriction screen G11, a restriction image P55 is placed. The restriction image P55 may show text or images indicating that the output is restricted. In the example in Figure 22, the restriction image P55 shows the text "Due to insufficient tickets, the conclusions have been summarized. ×× tickets are required to view the complete conclusions of the discussion."

[0115] The "××" part indicates the value of the missing value information (the number of tickets missing). This value of missing value information may be calculated, for example, as the difference between the value of the consideration obtained by the consideration acquisition unit 110 and the value of the value information held by the user who requested the processing.

[0116] Figure 23 is a diagram showing a specific example of the functional configuration of a service provision system 100 that displays a response screen G11 according to the second embodiment. The functional configuration shown in Figure 23 differs from that in Figure 15 in that the display content control unit 406 includes a summary unit 412 instead of a non-display range setting unit 407. The summary unit 412 obtains a summary based on the discussion results (responses) obtained by the discussion processing unit 409 if there are insufficient tickets. When obtaining a summary, the summary unit 412 may obtain a summary according to the number of tickets the user possesses. For example, the summary unit 412 may obtain a summary that can be obtained with a number of tokens corresponding to the number of tickets the user possesses. The summary unit 412 outputs the obtained summary to the screen generation unit 408.

[0117] Next, the processing flow of each function shown in Figure 23 will be explained. The ticket management unit 403 reads ticket information and purchases tickets. The ticket management unit 403 updates the number of purchased tickets in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The discussion processing unit 409 conducts the discussion and obtains responses (opinions and conclusions). The discussion processing unit 409 outputs the total number of tokens of the responses (opinions and conclusions) obtained as a result of the discussion to the ticket usage calculation unit 405. The discussion processing unit 409 outputs the text of the discussion results to the display content control unit 406. The ticket usage calculation unit 405 calculates the number of tokens that can be output based on the number of tickets owned by the user and the token unit price. The ticket usage calculation unit 405 calculates the number of missing tokens based on the number of outputtable tokens and the number of tokens of the discussion results from the discussion processing unit 409. Based on the calculation result of the number of missing tokens, the ticket usage calculation unit 405 outputs the number of tokens to be displayed and the number of tokens not to be displayed from the discussion results.

[0118] The summarization unit 412 retrieves summaries from the agenda results according to the number of missing tickets and outputs them to the screen generation unit 408. The screen generation unit 408 displays the discussion results, including the summary and hidden portions. The user checks the displayed summary and instructs to purchase additional tickets to view the entire discussion result. In response to this instruction, the ticket management unit 403 calculates the total number of tickets by adding the number of tickets the user has already purchased and the number of tickets the user already possesses, and records it in the account information storage unit 401. The ticket management unit 403 outputs the total number of tickets after purchase to the ticket usage calculation unit 405. The ticket usage calculation unit 405 subtracts the missing tickets from the total number of tickets and the number of missing tickets. The ticket usage calculation unit 405 outputs the result that the number of missing tickets has become zero. The screen generation unit 408 displays the full text of the discussion result instead of the summary. For example, it displays the full text of the conclusion instead of the summary, and further displays the full text of the opinions by unhiding the hidden portions. This allows the user to view the entire text, including the previously hidden portions.

[0119] (Effects of the Second Embodiment) As described above, the response system 1 according to the second embodiment performs the following processes. The consideration acquisition unit 110 acquires the value of the consideration necessary to perform the process in accordance with the instructions received from the terminal device 2. The restriction control unit 112 determines whether the user of the terminal device 2 is able to pay the required consideration. If payment is possible, the restriction control unit 112 does not impose any particular restrictions. On the other hand, if payment is not possible, the restriction control unit 112 performs a downgrade process on some or all of the opinions and conclusions.

[0120] The screen displayed by the response system 1 according to the second embodiment has the following characteristics. When restriction processing is performed by the restriction control unit 112, the opinions and conclusions are displayed in a downgraded form, either partially or entirely. This allows the response system 1 to visually present to the user the progress of discussions among multiple agents A within the user's budget.

[0121] As described above, the restriction control unit 112 may release the restriction processing and output the data without downgrade processing if payment for the deficit is made after the restriction processing has been performed and output has been output. The condition for releasing such restriction processing does not have to be limited to payment for the deficit. For example, the restriction processing may be released when an advertisement corresponding to the deficit is output on the user's terminal device 2. An advertisement corresponding to the deficit may be, for example, an advertisement of a type pre-associated with the deficit, or an advertisement of a length corresponding to the deficit.

[0122] The second embodiment configured as described above provides the following benefits. When a service is provided for acquiring data using multiple response models (for example, when commercializing an AI constellation), it is difficult to predict how the discussions in the multiple response models will unfold. Therefore, it is also difficult to predict the price required to receive the service, and there may be cases where the user cannot view the desired answer with the value information such as tickets they possess. In such cases, it is necessary to devise an output that comes as close as possible to the desired range within the range of the value information the user possesses. The present invention solves these problems and makes it possible to output information that comes as close as possible to the range desired by the user within the range of the value information the user possesses.

[0123] Figure 24 is a schematic block diagram showing the configuration of a computer according to at least one embodiment. The computer 90 includes a processor 91, memory 92, storage 93, and interface 94. The response system 1 described above is implemented in the computer 90. The operation of each processing unit described above is stored in the storage 93 in the form of a program. The processor 91 reads the program from the storage 93, loads it into the memory 92, and executes the above processing according to the program. The processor 91 also allocates memory areas in the memory 92 corresponding to each of the above-mentioned storage units according to the program. Examples of the processor 91 include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and a microprocessor.

[0124] The program may be for implementing a part of the functions to be performed by the computer 90. For example, the program may perform functions in combination with other programs already stored in storage, or in combination with other programs implemented on other devices. In other embodiments, the computer 90 may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to, or instead of, the above configuration. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Array). In this case, some or all of the functions implemented by the processor 91 may be implemented by the integrated circuit. Such an integrated circuit is also included as an example of a processor. In other embodiments, the computer 90 may be virtualized on one or more computers.

[0125] Examples of storage 93 include magnetic disks, magneto-optical disks, optical disks, and semiconductor memory. Storage 93 may be an internal medium directly connected to the bus of the computer 90, or it may be an external medium connected to the computer 90 via an interface 94 or a communication line. Furthermore, if this program is delivered to the computer 90 via a communication line, the computer 90 that receives the delivery may expand the program into memory 92 and execute the above processing. In at least one embodiment, storage 93 is a tangible storage medium that is not temporary.

[0126] Furthermore, the program may be intended to implement some of the functions described above. In addition, the program may be a so-called differential file (differential program) that implements the functions described above in combination with other programs already stored in the storage 93.

[0127] The response system 1 may be implemented using one information processing device or multiple information processing devices. For example, the response system 1 may be implemented using a cloud or similar device.

[0128] Figure 25 shows one aspect of the response system 1 according to this embodiment. The response system 1 comprises a discussion processing unit 409, a response acquisition unit, a ticket usage calculation unit 405, and a display content control unit 406. The discussion processing unit 409 corresponds to the discussion generation unit 107. The ticket usage calculation unit 405 comprises a token unit price storage unit 404, an account information storage unit 401, and a consideration condition consideration unit. The account information storage unit 401 corresponds to the value information storage unit 111. The consideration condition consideration unit corresponds to the consideration acquisition unit 110. The display content control unit 406 comprises a non-display range setting unit 407 and a screen generation unit 408. The screen generation unit 408 corresponds to the output unit 109. The configuration including the display content control unit 406 and the value information storage unit 111 corresponds to the restriction control unit 112. The configuration including the display content control unit 406 and the consideration condition consideration unit (consideration acquisition unit 110) corresponds to the control unit. Information indicating whether the compensation conditions have been met is input from the compensation condition consideration unit to the non-display range setting unit. The discussion processing unit 409 receives themes from one or more users or other information processing devices via the network.

[0129] The following additional information is disclosed regarding the embodiments described above.

[0130] (Note 1) An information processing device comprising: a memory; and at least one processor connected to the memory, wherein the processor uses a plurality of response models to obtain a response indicating information relating to at least one common theme; and if a consideration condition, which is a condition relating to consideration, is not met, it performs a downgrade process on part or all of the response and outputs a response after the downgrade process. (Note 2) A non-temporary storage medium storing a program executable by a computer to perform information processing, wherein the information processing uses a plurality of response models to obtain a response indicating information relating to at least one common theme; and if a consideration condition, which is a condition relating to consideration, is not met, it performs a downgrade process on part or all of the response and outputs a response after the downgrade process.

[0131] According to the above embodiment, the response system can effectively acquire data.

[0132] 1...Response System 101...Input Unit 102...Agenda Determination Unit 103...Role Storage Unit 104...Perspective Acquisition Unit 105...Selection Unit 106...Grouping Unit 107...Discussion Generation Unit 108...Extraction Unit 109...Output Unit 110...Consideration Acquisition Unit 111...Value Information Storage Unit 112...Restriction Control Unit 113...Settlement History Storage Unit 114...Settlement Control Unit 2...Terminal Device 3...External Server 90...Computer 91...Processor 92...Memory 93...Storage 94...Interface G1...Agenda Input Screen G2...Discussion Preparation Screen G3...Discussion Progress Screen G4...Discussion Summary Screen G5...Response Screen G6...First Restriction Screen G7...Second Restriction Screen G8...Third Restriction Screen G9...Fourth Restriction Screen G10...Fifth Restriction Screen G11...Sixth Restriction Screen P1...First Area P11...Input Form P12... Agenda Label P2... Second Area P21... Symbol P22... Balloon P23... Group Frame P24... Speech Bubble P25... Related Lines P3... Third Area P31... Statement Text P32... Symbol P33... Summary Status Display P34... Restricted Image P35... Ownership Value Information Image P4... Fourth Area P41... Opinion Text P42... Restricted Image P5... Fifth Area P51... Conclusion Text P52... Interruption Conclusion Text P53... Restricted Image P54... Summary Conclusion Text P55... Restricted Image

Claims

1. An information processing device comprising: a control unit that uses multiple response models to obtain responses indicating information about at least one common theme, and performs a downgrade process on part or all of the responses if the compensation conditions, which are conditions related to compensation, are not met, and the control unit outputs the responses after the downgrade process.

2. The information processing apparatus according to claim 1, wherein the control unit performs a downgrade process so that the output content falls within the price conditions that the user can satisfy if the price conditions are not met.

3. The information processing apparatus according to claim 1, wherein the control unit outputs the response partially as a downgrade process when the consideration conditions are not met.

4. The information processing apparatus according to claim 1, wherein the control unit outputs only the responses obtained based on a portion of the response model as a downgrade process when the consideration conditions are not met.

5. The information processing apparatus according to claim 1, wherein the control unit obtains at least one second-kind response obtained based on a plurality of first-kind responses obtained using the plurality of response models, and if the consideration condition is not met, preferentially outputs the second-kind response as the downgrade process, and the response model used when obtaining the first-kind response is different from the response model used when obtaining the second-kind response.

6. The information processing apparatus according to claim 1, wherein the control unit outputs a summary of a plurality of responses obtained using the plurality of response models as a downgrade process when the consideration conditions are not met.

7. The information processing apparatus according to claim 1, wherein the control unit determines the degree of downgrade in the downgrade process according to the extent to which the consideration conditions are not met.

8. The information processing apparatus according to claim 5, wherein the control unit obtains a new first-kind response by using the first-kind response as input.

9. An output method that uses multiple response models to obtain responses indicating information about at least one common theme, and if the compensation condition is not met, performs a downgrade process on part or all of the response, and outputs the response after the downgrade process.

10. A program for causing a computer to function as an information processing device, comprising a control unit that uses multiple response models to obtain responses indicating information about at least one common theme, and performs a downgrade process on part or all of the responses if a consideration condition, which is a condition relating to consideration, is not met, wherein the control unit outputs the responses after the downgrade process.

Images