Information processing method and information processing device

Abstract

An information processing method comprises: estimating the degree of risk that a message to be presented to a user poses to the user; determining a display mode of a warning pertaining to the accuracy of the message, on the basis of the degree of risk; and displaying the message on a display and also displaying the warning on the display according to the determined display mode.

Description

Information processing method and information processing device

[0001] The present disclosure relates to techniques for presenting messages to users.

[0002] For example, Patent Document 1 discloses that a terminal communicating with a server displays, on a display unit of the terminal, a chat room including a user of a first terminal and a user of a second terminal different from the first terminal; receives the first content from the server by a communication unit of the terminal based on the first content transmitted by the first terminal being received by the server; receives second information based on the first information and the second content from the server by a communication unit of the terminal based on the first information regarding a location and the second content transmitted from the second terminal being received by the server; and displays, in the chat room, the first content and the second content that is displayed in a different manner from the first content based on the second information.

[0003] However, the above-mentioned conventional techniques do not take into consideration the degree of risk that messages pose to users, and further improvements are needed.

[0004] Japanese Patent Application Laid-Open No. 2021-108009

[0005] The present disclosure has been made to solve the above problems, and aims to provide technology that can prevent users from being deceived by false information while maintaining users' willingness to use the system.

[0006] An information processing method according to one aspect of the present disclosure is an information processing method executed by a computer, and includes estimating a level of risk that a message presented to a user poses to the user, determining a display mode of a warning regarding the accuracy of the message based on the level of risk, displaying the message on a display, and displaying the warning on the display in accordance with the determined display mode.

[0007] According to the present disclosure, it is possible to prevent users from being deceived by false information while maintaining users' willingness to use the system.

[0008] 10 is a diagram showing an overall configuration of an information processing system in the present embodiment 1. FIG. 11 is a diagram showing an example of the configuration of a server in the present embodiment 1. FIG. 12 is a flowchart showing an example of display processing of the server in the present embodiment 1. FIG. 13 is a diagram showing an example of a display screen displayed on the display of the information terminal when the risk level is lower than a threshold in the present embodiment 1. FIG. 14 is a diagram showing an example of a display screen displayed on the display of the information terminal when the risk level is equal to or higher than a threshold in the present embodiment 1. FIG. 15 is a diagram showing an example of the configuration of a server in the present embodiment 2. FIG. 16 is a diagram showing an example of a prompt input to the language generation AI in the present embodiment 2. FIG. 17 is a diagram showing another example of a prompt input to the language generation AI in the present embodiment 2. FIG. 18 is a diagram for explaining a method of calculating a risk level in the present embodiment 2. FIG. 19 is a flowchart showing an example of display processing of the server in the present embodiment 2. FIG. 19 is a flowchart showing an example of answer generation processing by the generation unit in step S14 of FIG. 10. FIG. 19 is a diagram showing an example of a display screen displayed on the display of the information terminal when the risk level is equal to or higher than a first threshold and lower than a second threshold in the present embodiment 2. FIG. 19 is a diagram showing an example of the configuration of a server in the present embodiment 3. FIG. 19 is a diagram showing an example of a first prompt generated by the second generation unit and a second answer generated by the language generation AI in the present embodiment 3. 10 is a diagram showing an example of a second prompt generated by the second generation unit and a second answer generated by the language generation AI in the third embodiment. FIG. 11 is a diagram showing an example of a third prompt generated by the second generation unit and a second answer generated by the language generation AI in the third embodiment. FIG. 12 is a flowchart showing an example of a display process of the server in the third embodiment. FIG. 13 is a diagram showing an example of a display screen displayed on the display of the information terminal when the risk level is equal to or greater than the first threshold and less than the second threshold in the third embodiment. FIG. 14 is a diagram showing an example of a display screen displayed on the display of the information terminal when the risk level is equal to or greater than the second threshold in the third embodiment. FIG. 15 is a diagram showing an example of the configuration of the server in the fourth embodiment. FIG. 16 is a flowchart showing an example of the display process of the server in the fourth embodiment. FIG. 17 is a diagram showing an example of the configuration of the server in the fifth embodiment.FIG. 10 is a diagram showing an example of a second prompt generated by the generation unit and an answer generated by the language generation AI in the present embodiment 5. FIG. 11 is a diagram showing an example of a third prompt generated by the generation unit and an answer generated by the language generation AI in the present embodiment 5. FIG. 12 is a flowchart showing an example of a display process of the server in the present embodiment 5.

[0009] (Knowledge Forming the Basis of the Present Disclosure) Services known as messaging applications, which allow users to exchange messages one-on-one or within a group, are widely used. Services known as chatbots, which have a similar UI but in which the other party is not a human but an information processing system such as AI (Artificial Intelligence), are also becoming increasingly popular. In messaging exchanges using these services, whether each message is trustworthy is an important factor. The rampant use of fraud in messaging applications, which takes advantage of the inability to see the other party's face, has become a social problem. Furthermore, because the AI ​​in chatbots does not always output correct information, users may be deceived by the AI ​​by trusting its messages, resulting in losses similar to those suffered by fraud.

[0010] In the above-mentioned conventional technology, whether or not a message sent from a terminal is highly credible is estimated based on the terminal's location information sent from the terminal along with the message, but the risk level corresponding to the degree of uncertainty of the message is not taken into consideration. Therefore, the above-mentioned conventional technology does not disclose determining whether or not to display a warning about the accuracy of the message on a display based on the risk level.

[0011] In order to solve the above problems, the following techniques are disclosed.

[0012] (1) An information processing method according to one aspect of the present disclosure is an information processing method executed by a computer, and includes estimating a level of risk that a message presented to a user poses to the user, determining a display mode of a warning regarding the accuracy of the message based on the level of risk, displaying the message on a display, and displaying the warning on the display in accordance with the determined display mode.

[0013] According to this configuration, even if the message contains false information, a warning regarding the accuracy of the message is displayed on the display in accordance with the display mode determined based on the degree of risk the message poses to the user, thereby preventing users from being deceived by false information while maintaining their willingness to use the system.

[0014] (2) In the information processing method described in (1) above, the determination of the display mode may include increasing the degree of emphasis of the warning as the risk level increases.

[0015] According to this configuration, the higher the risk level, the more emphasis is placed on the warning, so that a user at high risk can be warned more emphatically.

[0016] (3) In the information processing method described in (1) or (2) above, the method may further include acquiring characteristic information of the user, and estimating the risk level may include estimating the user's trustworthiness regarding the message as the risk level based on the characteristic information.

[0017] According to this configuration, a warning regarding the accuracy of the message is displayed on the display for users who tend to believe the message, thereby preventing users from being deceived by false information.

[0018] (4) In the information processing method described in (3) above, the characteristic information may include at least one of a diagnostic result regarding the user's personality and a diagnostic result regarding the user's cognitive function.

[0019] According to this configuration, it is possible to more accurately estimate the reliability of a user based on at least one of the diagnostic results regarding the user's personality and the diagnostic results regarding the user's cognitive function.

[0020] (5) In the information processing method described in (4) above, the characteristic information may include the results of any one of an Enneagram diagnosis, an MBTI (Myers-Briggs Type Indicator) diagnosis, a coaching typing diagnosis, and a Big Five personality diagnosis.

[0021] According to this configuration, the reliability of the user can be estimated more accurately based on the diagnostic results of any one of the Enneagram diagnosis, MBTI diagnosis, coaching typing diagnosis, and Big Five personality diagnosis.

[0022] (6) In the information processing method described in (4) above, the characteristic information may include test results of either an emotional intelligence (EQ) test or a cognitive ability test.

[0023] With this configuration, the trustworthiness of a user can be more accurately estimated based on the test results of either the emotional intelligence test or the cognitive ability test.

[0024] (7) In the information processing method described in (3) above, the characteristic information may include an index representing any one of cooperation ability, analytical thinking ability, and self-evaluation.

[0025] According to this configuration, it is possible to more accurately estimate the trustworthiness of a user based on an index that indicates any one of cooperation ability, analytical thinking ability, and self-evaluation.

[0026] (8) In the information processing method described in (1) or (2) above, estimating the risk level may include estimating the degree of uncertainty of the message as the risk level depending on the source of the message.

[0027] According to this configuration, when a message that is likely to contain false information is displayed, a warning regarding the accuracy of the message is displayed on the display, thereby preventing the user from being deceived by false information.

[0028] (9) In the information processing method described in (8) above, the generator may be either a generation AI or a database in which all or part of the message is stored in advance, and the degree of uncertainty when the generator is the generation AI may be higher than the degree of uncertainty when the generator is the database.

[0029] According to this configuration, when a message generated by the generation AI is displayed, a warning regarding the accuracy of the message is displayed on the screen, thereby preventing users from being deceived by false information.

[0030] (10) In the information processing method described in (1) or (2) above, estimating the risk level may include estimating an impact level indicating the magnitude of the impact that the message has on the user as the risk level.

[0031] According to this configuration, when a message that has a large impact on the user is displayed, a warning regarding the accuracy of the message is displayed on the screen, thereby preventing the user from being deceived by false information and suffering major damage.

[0032] (11) In the information processing method described in (10) above, estimating the risk level may include determining a topic addressed in the message and estimating the impact level based on the determined topic.

[0033] According to this configuration, if a message deals with a topic that has a large impact on users, a warning about the accuracy of the message is displayed on the screen, thereby preventing users from being deceived by false information and suffering major damage.

[0034] (12) In the information processing method described in (11) above, the topic may be any one of finance, health, medicine, hobbies, preferences, and comfort, and the influence level when the topic is any one of finance, health, and medicine may be higher than the influence level when the topic is any one of hobbies, preferences, and comfort.

[0035] According to this configuration, when a message relating to finance, health, or medical care is displayed, a warning regarding the accuracy of the message is displayed on the display, thereby preventing the user from being deceived by false information relating to finance, health, or medical care.

[0036] (13) In the information processing method described in (1) or (2) above, the method may further include acquiring characteristic information of the user, and estimating the risk level may include estimating the user's reliability of the message based on the characteristic information, estimating the degree of uncertainty of the message depending on the source of the message, estimating an impact level indicating the magnitude of the impact the message has on the user, and calculating the risk level based on the reliability, the degree of uncertainty, and the impact level.

[0037] According to this configuration, the risk level is calculated based on the user's trust in the message, the degree of uncertainty of the message depending on the source of the message, and the impact level indicating the extent of the impact the message has on the user, thereby more reliably preventing users from being deceived by false information and suffering damage.

[0038] (14) In the information processing method described in any one of (1) to (13) above, the message may be a response to a question received from the user.

[0039] With this configuration, even if the answer to a user's question contains incorrect information, a warning regarding the accuracy of the answer is displayed on the display, thereby preventing the user from being deceived by incorrect information.

[0040] (15) In the information processing method described in any one of (1) to (14) above, generating the message may include inputting a question received from the user into a generation AI and obtaining an answer to the question from the generation AI.

[0041] With this configuration, even if the answer generated by the generation AI contains false information, a warning about the accuracy of the answer will be displayed on the display, thereby preventing users from being deceived by false information.

[0042] Furthermore, the present disclosure can be realized not only as an information processing method that executes the characteristic processes described above, but also as an information processing device having a characteristic configuration corresponding to the characteristic processes executed by the information processing method. Furthermore, the present disclosure can also be realized as a computer program that causes a computer to execute the characteristic processes included in such an information processing method. Therefore, the same effects as those of the above information processing method can also be achieved in the following other aspects.

[0043] (16) Another aspect of the present disclosure is an information processing device that includes a processor, which estimates the level of risk that a message presented to a user poses to the user, determines a display mode of a warning regarding the accuracy of the message based on the level of risk, displays the message on a display, and displays the warning on the display in accordance with the determined display mode.

[0044] An information processing program according to another aspect of the present disclosure causes a computer to function in such a manner that it estimates the degree of risk that a message presented to a user poses to the user, determines a display mode of a warning regarding the accuracy of the message based on the degree of risk, displays the message on a display, and displays the warning on the display in accordance with the determined display mode.

[0045] A non-transitory computer-readable recording medium according to another aspect of the present disclosure records the above-described information processing program.

[0046] (17) Another aspect of the present disclosure is an information processing method executed by a computer, which includes estimating the level of risk that an answer to a question received from a user poses to the user, inputting a prompt including the question into a generation AI, obtaining text including the answer output from the generation AI, and displaying a display screen including the text on a display, wherein the content of the display screen varies depending on the level of risk.

[0047] With this configuration, even if the answer contains false information, the content displayed on the display screen will differ depending on the degree of risk the answer poses to the user, thereby preventing users from being deceived by false information while maintaining their desire to use the system.

[0048] (18) In the information processing method described in (17) above, acquiring the text may include inputting the prompt edited according to the risk level into the generation AI and acquiring the text output from the generation AI, and displaying the display screen may include displaying on the display the display screen in which the expression of the answer contained in the text differs depending on the risk level.

[0049] According to this configuration, for high-risk users, answers are displayed in a manner that calls their attention to the accuracy of the answer, thereby preventing users from being deceived by false information.

[0050] (19) In the information processing method described in (17) above, the display of the display screen may include displaying on the display a display screen in which the display mode of a warning regarding the accuracy of the answer contained in the text changes depending on the risk level.

[0051] According to this configuration, the display mode of the warning regarding the accuracy of the answer contained in the text changes depending on the degree of risk that the answer poses to the user, thereby preventing users from being deceived by false information while maintaining users' motivation to use the system.

[0052] (20) An information processing method according to another aspect of the present disclosure may be an information processing method executed by a computer, and may include estimating the level of risk that an answer to a question received from a user poses to the user, generating a prompt for rephrasing the answer based on the level of risk, inputting the prompt to a generation AI, obtaining a message corresponding to the answer output from the generation AI, and displaying the message on a display.

[0053] According to this configuration, even if the answer contains false information, the answer is rephrased according to the level of risk, and a message corresponding to the rephrased answer is shown on the display, thereby maintaining the user's willingness to use the system while preventing the user from being deceived by false information.

[0054] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. Note that each of the embodiments described below represents a specific example of the present disclosure. The numerical values, shapes, components, steps, and order of steps shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components that are not described in the independent claims that represent the highest concept are described as optional components. Furthermore, in all embodiments, the respective contents can be combined.

[0055] (Embodiment 1) Fig. 1 is a diagram showing the overall configuration of an information processing system in Embodiment 1. The information processing system shown in Fig. 1 includes a server 1, an information terminal 2, and a language generation AI (Artificial Intelligence) server 3.

[0056] The information terminal 2 is operated by, for example, a user. The information terminal 2 may be configured as a portable computer such as a smartphone or a tablet computer, or may be configured as a stationary computer. Although one information terminal 2 is illustrated in the example of FIG. 1 , multiple information terminals 2 may be connected to the server 1 via the network 4. The information terminal 2 accepts information input by the user and transmits the accepted information to the server 1. The information terminal 2 also receives information transmitted by the server 1 and displays the received information.

[0057] The information terminal 2 accepts questions from users and transmits text information indicating the accepted questions to the server 1. The information terminal 2 receives text information indicating answers to the questions transmitted by the server 1 and displays the received text information.

[0058] The language generation AI server 3 is, for example, an on-premise server, and is communicatively connected to the server 1. The language generation AI server 3 may be a cloud server, and may be communicatively connected to the server 1 via the network 4. The server 1 or the information terminal 2 may also have the functionality of the language generation AI server 3.

[0059] The language generation AI server 3 has the function of language generation AI. The language generation AI server 3 generates text as a response to a given prompt. The language generation AI is, for example, Chat-GPT, but is not limited to this. The language generation AI server 3 inputs information received from the server 1 into the language generation AI, and transmits information generated by the language generation AI to the server 1.

[0060] Language generation AI may generate erroneous information that is not based on facts. This phenomenon is called hallucination. Because the language model of language generation AI learns using large amounts of data collected from the Internet and other sources, hallucination may occur if the collected data contains erroneous information. Therefore, when presenting information generated by the language generation AI to the user, the server 1 displays a warning regarding the accuracy of the information.

[0061] The server 1 is an example of an information processing device and a computer. The server 1 and the information terminal 2 are connected to each other so that they can communicate with each other via a network 4. An example of the network 4 is the Internet. The server 1 is, for example, a cloud server made up of one or more computers. However, this is just one example, and the server 1 may be made up of an edge server or may be implemented in the information terminal 2. The aspect in which the server 1 is implemented in the information terminal 2 is an example of the aspect in which the information terminal 2 is configured as an information processing device.

[0062] FIG. 2 is a diagram showing an example of the configuration of the server 1 according to the first embodiment.

[0063] The server 1 includes a communication unit 11 , a processor 12 , and a memory 13 .

[0064] The communication unit 11 is a communication interface that connects the server 1 to the network 4. The communication unit 11 receives text information indicating a question entered by a user from the information terminal 2. The communication unit 11 transmits text information indicating an answer generated in response to the question to the information terminal 2.

[0065] The processor 12 is configured, for example, by a central processing unit (CPU). The processor 12 includes a characteristic information acquisition unit 121, a risk level estimation unit 122, a question acquisition unit 123, a generation unit 124, a determination unit 125, and a display control unit 126. The characteristic information acquisition unit 121, the risk level estimation unit 122, the question acquisition unit 123, the generation unit 124, the determination unit 125, and the display control unit 126 may be realized by the processor 12 executing an information processing program, or may be configured by a dedicated hardware circuit such as an ASIC. The information processing program may be recorded on a non-transitory computer-readable recording medium.

[0066] The memory 13 is configured by a non-volatile rewritable storage device such as a hard disk drive or a solid state drive, etc. The memory 13 includes a user information database (DB) 131.

[0067] The characteristic information acquisition unit 121 acquires characteristic information of a user. The user information DB 131 stores characteristic information for each user in advance. The characteristic information acquisition unit 121 reads out the user's characteristic information from the user information DB 131. The characteristic information is information indicating the user's personality. The characteristic information includes at least one of a diagnosis result regarding the user's personality and a diagnosis result regarding the user's cognitive function. The characteristic information includes, for example, a diagnosis result of any of an Enneagram diagnosis, an MBTI (Myers-Briggs Type Indicator) diagnosis, a coaching typing diagnosis, and a Big Five personality diagnosis.

[0068] The server 1 acquires characteristic information of the user in advance. The server 1 transmits a plurality of questions for diagnosing the personality to the information terminal 2. The information terminal 2 displays the received plurality of questions and accepts input of a plurality of answers to the plurality of questions by the user. The information terminal 2 transmits the plurality of answers input by the user to the server 1. The server 1 diagnoses the user's personality based on the received plurality of answers and stores the diagnosis result as characteristic information in the user information DB 131.

[0069] For example, the Enneagram diagnosis classifies a user's personality into nine types by having the user answer multiple questions. Also, for example, the MBTI diagnosis classifies a user's personality into two or 16 types by having the user answer multiple questions. Also, for example, the Coaching Type Classification diagnosis classifies a user's personality into four types by having the user answer multiple questions. Also, for example, the Big Five personality diagnosis classifies a user's personality into five types by having the user answer multiple questions.

[0070] The characteristic information may include the test results of either an Emotional Intelligence Quotient (EQ) test or a cognitive ability test. The characteristic information may also include an index representing either agreeableness, analytical thinking ability, or self-esteem. A person with high agreeableness tends to be more receptive to the opinions of others, and it is believed that there is a high correlation between agreeableness and the degree of belief. Therefore, by using agreeableness as characteristic information, the accuracy of risk assessment can be further improved.

[0071] The risk level estimation unit 122 estimates the level of risk that a message presented to a user poses to the user. The risk level represents the user's susceptibility to being gullible. The message is an answer to a question received from the user. In other words, the risk level estimation unit 122 estimates the level of risk that an answer to a question received from the user poses to the user. The risk level estimation unit 122 includes a reliability estimation unit 141.

[0072] The reliability estimation unit 141 estimates the user's reliability for the message as a risk level based on the characteristic information acquired by the characteristic information acquisition unit 121. The reliability represents the degree to which the user believes in the answer of the language generation AI. The reliability is expressed as two values, for example, "high" and "low." "High" is, for example, 0.9, and "low" is, for example, 0.1. If the reliability is high, the user is more likely to believe the answer of the language generation AI, and if the reliability is low, the user is less likely to believe the answer of the language generation AI.

[0073] The memory 13 stores in advance a table associating the user's personality type with a reliability. For example, if the characteristic information is the result of an Enneagram diagnosis, a reliability is associated with each of the nine types of the Enneagram diagnosis. For example, Types 1, 3, 5, and 8 are associated with a reliability of "low," and Types 2, 4, 6, 7, and 9 are associated with a reliability of "high."

[0074] The reliability estimation unit 141 refers to the table and estimates the reliability associated with the characteristic information acquired by the characteristic information acquisition unit 121 as a risk level.

[0075] Furthermore, the reliability estimation unit 141 may lower the reliability as the emotional intelligence value and the cognitive ability value are higher, and may increase the reliability as the emotional intelligence value and the cognitive ability value are lower. That is, the reliability estimation unit 141 may estimate the reliability to be "low" when the emotional intelligence value and the cognitive ability value are equal to or higher than a threshold, and may estimate the reliability to be "high" when the emotional intelligence value and the cognitive ability value are lower than the threshold.

[0076] The question acquisition unit 123 acquires a question received from a user. The information terminal 2 transmits the question input by the user to the server 1. The communication unit 11 receives the question transmitted by the information terminal 2 and outputs it to the question acquisition unit 123.

[0077] The generation unit 124 generates a message to be presented to the user. As described above, the message is an answer to a question received from the user. The generation unit 124 inputs the question received from the user to the language generation AI and obtains an answer to the question from the language generation AI. The generation unit 124 transmits the question acquired by the question acquisition unit 123 to the language generation AI server 3 via the communication unit 11. The language generation AI server 3 inputs the received question to the language generation AI and obtains an answer to the question generated by the language generation AI. The language generation AI server 3 transmits the answer generated by the language generation AI to the server 1. The communication unit 11 receives the answer sent by the language generation AI server 3 and outputs it to the generation unit 124.

[0078] The generation unit 124 may generate a prompt that includes a question received from the user. The generation unit 124 may generate a prompt that includes only the question received from the user, or may generate a prompt that includes the question received from the user and a constraint such as the number of characters.

[0079] The determination unit 125 determines a display mode of a warning regarding the accuracy of the message based on the risk level estimated by the risk level estimation unit 122. The determination unit 125 determines whether the risk level is smaller than a threshold value. The threshold value is, for example, 0.5. If the determination unit 125 determines that the risk level is smaller than the threshold value, it determines not to display the warning on the display of the information terminal 2. On the other hand, if the determination unit 125 determines that the risk level is equal to or greater than the threshold value, it determines to display the warning on the display of the information terminal 2.

[0080] The display control unit 126 displays the answer (message) on the display and also displays a warning on the display according to the display mode determined by the determination unit 125. The display control unit 126 displays a display screen on the display in which the display mode of a warning regarding the accuracy of the answer included in the text changes depending on the risk level. That is, when the determination unit 125 determines that the risk level is equal to or higher than a threshold, the display control unit 126 displays the answer (message) on the display and also displays a warning on the display. The display control unit 126 transmits text information of the answer (message) and the warning to the information terminal 2 via the communication unit 11. The communication unit of the information terminal 2 receives the text information transmitted by the server 1. The display of the information terminal 2 displays the answer (message) and the warning received by the communication unit.

[0081] Furthermore, when the determination unit 125 determines that the risk level is smaller than the threshold, the display control unit 126 displays the answer (message) on the display but does not display the warning on the display. The display control unit 126 transmits text information of the answer (message) to the information terminal 2 via the communication unit 11. The communication unit of the information terminal 2 receives the text information transmitted by the server 1. The display of the information terminal 2 displays the answer (message) received by the communication unit.

[0082] Next, the display process of the server 1 in the first embodiment will be described.

[0083] FIG. 3 is a flowchart showing an example of the display process of the server 1 in the first embodiment.

[0084] First, in step S1 , the characteristic information acquisition unit 121 acquires the user's characteristic information from the user information DB 131 .

[0085] Next, in step S2, the reliability estimation unit 141 estimates the reliability of the user with respect to the answer as a risk level based on the characteristic information acquired by the characteristic information acquisition unit 121.

[0086] Next, in step S3, the question acquisition unit 123 acquires a question received from the user.

[0087] Next, in step S4, the generation unit 124 generates an answer to the question acquired by the question acquisition unit 123. The generation unit 124 inputs the question to the language generation AI and acquires the answer to the question from the language generation AI.

[0088] Next, in step S5, the determining unit 125 determines whether the risk level estimated by the reliability estimating unit 141 is smaller than a threshold value.

[0089] If it is determined that the risk level is less than the threshold value (YES in step S5), the display control unit 126 causes the answer generated by the generation unit 124 to be displayed on the display of the information terminal 2 in step S6.

[0090] On the other hand, if it is determined that the risk level is above the threshold (NO in step S5), in step S7, the display control unit 126 displays the answer generated by the generation unit 124 on the display of the information terminal 2, and also displays a warning regarding the accuracy of the answer on the display of the information terminal 2.

[0091] Next, in step S8, the display control unit 126 determines whether or not to end the display process. If it is determined that the display process should be ended (YES in step S8), the display process ends. The display of the information terminal 2 displays an end button. If the end button is pressed by the user, the display control unit 126 determines that the display process should be ended.

[0092] On the other hand, if it is determined that the display process should not be ended (NO in step S8), the process returns to step S3.

[0093] In this way, a dialogue between the user and the language generation AI progresses as the user inputs a question and the language generation AI generates an answer to the question.

[0094] According to this first embodiment, even if a message contains false information, a warning regarding the accuracy of the message is displayed on the display in accordance with a display mode determined based on the degree of risk that the message poses to the user, thereby preventing users from being deceived by false information while maintaining users' willingness to use the system.

[0095] FIG. 4 is a diagram showing an example of a display screen 200 that is displayed on the display of the information terminal 2 when the risk level is smaller than the threshold value in the first embodiment.

[0096] The display screen 200 displays a question 201 entered by a user and an answer 202 to the question 201. In Fig. 4, the answer 202 displayed in response to the question 201, "Please tell me how to improve high blood pressure," is, "People with high blood pressure should drink as much water as possible." Here, the answer 202 generated by the language generation AI may be false information.

[0097] If the risk level is less than the threshold, i.e., if the user's trust in the language generation AI is less than the threshold, the display screen 200 displays the answer 202 generated by the language generation AI, but does not display a warning regarding the accuracy of the answer 202. Users with low trust in the language generation AI tend to have difficulty trusting answers generated by the language generation AI. Therefore, even if the language generation AI generates an incorrect answer, the user is unlikely to be fooled by the incorrect answer, so there is no need to display a warning. In the first embodiment, since no warning is displayed, the display screen 200 is easy to see, and the user's motivation to use the service can be maintained.

[0098] FIG. 5 is a diagram showing an example of a display screen 220 that is displayed on the display of the information terminal 2 when the risk level is equal to or greater than the threshold value in the first embodiment.

[0099] The display screen 220 displays a question 201 entered by a user, an answer 202 to the question 201, and a warning 223 regarding the accuracy of the answer. In Fig. 5, the warning 223 reads, "Warning! Answers provided by AI may not be accurate."

[0100] If the risk level is equal to or greater than a threshold, i.e., if the user's trust in the language generation AI is equal to or greater than a threshold, the display screen 220 displays a warning 223 regarding the accuracy of the answer along with the answer 202 generated by the language generation AI. Users who have a high level of trust in the language generation AI tend to be more likely to believe the answers generated by the language generation AI. Therefore, if the language generation AI generates an incorrect answer, the user may be deceived by the incorrect answer, so it is necessary to warn the user about the accuracy of the answer. In the first embodiment, a warning is displayed along with the answer to the question, thereby reducing the risk that the user will be deceived by an incorrect answer.

[0101] In the first embodiment, the display control unit 126 does not display a warning when the risk level is lower than the threshold, but the present disclosure is not particularly limited to this. When the risk level is lower than the threshold, the display control unit 126 may display a warning. In this case, the display control unit 126 differentiates the display mode of the warning when the risk level is lower than the threshold from the display mode of the warning when the risk level is equal to or higher than the threshold. The higher the risk level, the higher the degree of emphasis of the warning. In other words, the display control unit 126 increases the degree of emphasis of the warning when the risk level is equal to or higher than the threshold compared to the degree of emphasis of the warning when the risk level is lower than the threshold.

[0102] For example, the display control unit 126 may display a warning in red when the risk level is equal to or greater than the threshold, and may display a warning in a color other than red when the risk level is less than the threshold. Furthermore, the display control unit 126 may use larger text in a warning when the risk level is equal to or greater than the threshold than in a warning when the risk level is less than the threshold. Furthermore, the display control unit 126 may attach an exclamation mark icon to a warning when the risk level is equal to or greater than the threshold, but may not attach an exclamation mark icon to a warning when the risk level is less than the threshold. Furthermore, the display control unit 126 may flash a warning when the risk level is equal to or greater than the threshold, but may not flash a warning when the risk level is less than the threshold.

[0103] Furthermore, in the first embodiment, a question is received from the user and an answer to the question received from the user is generated, but the present disclosure is not particularly limited to this, and a message may be generated without receiving a question from the user. In this case, the generation unit 124 may create a question (prompt) to cause the language generation AI to generate advice for the user and input the created question to the language generation AI. Then, the generation unit 124 may obtain an answer to the question from the language generation AI. For example, the generation unit 124 may create a question such as, "Please give me some health advice for the user."

[0104] (Embodiment 2) In embodiment 1, a language generation AI generates an answer. In contrast, in embodiment 2, if a database that previously stores questions and answers in association with each other does not contain an answer corresponding to a question received from a user, the language generation AI generates an answer. Also, in embodiment 1, the user's reliability for a message is estimated based on user characteristic information. In contrast, in embodiment 2, the degree of uncertainty of a message is further estimated depending on the source of the message, and an influence indicating the magnitude of the impact the message will have on the user is estimated.

[0105] 6 is a diagram showing an example of the configuration of the server 1A in the present embodiment 2. Note that the overall configuration of the information processing system in the present embodiment 2 is the same as the overall configuration of the information processing system in the embodiment 1, and therefore a description thereof will be omitted.

[0106] The server 1A includes a communication unit 11, a processor 12A, and a memory 13A. In the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0107] The processor 12A includes a characteristic information acquisition unit 121, a risk level estimation unit 122A, a question acquisition unit 123, a generation unit 124A, a determination unit 125, and a display control unit 126. The characteristic information acquisition unit 121, the risk level estimation unit 122A, the question acquisition unit 123, the generation unit 124A, the determination unit 125, and the display control unit 126 may be realized by the processor 12A executing an information processing program, or may be configured by a dedicated hardware circuit such as an ASIC. The information processing program may be recorded on a non-transitory computer-readable recording medium.

[0108] The memory 13A includes a user information database (DB) 131 and an expert knowledge database (DB) 132 .

[0109] The specialized knowledge database (DB) 132 stores questions and answers to the questions in association with each other in advance. The answers stored in the specialized knowledge DB 132 are more accurate than the answers generated by the language generation AI. The specialized knowledge DB 132 may store questions in association with all of the answers to the questions in advance, or may store questions in association with only a portion of the answers to the questions. In other words, the specialized knowledge DB 132 may store all of the sentences to be presented to the user as answers, or may store only a portion of the sentences to be presented to the user as answers.

[0110] The generation unit 124A refers to the specialized knowledge DB 132 and extracts from the specialized knowledge DB 132 an answer associated with the question received from the user. The generation unit 124A may extract all of the answers associated with the question received from the user from the specialized knowledge DB 132. The generation unit 124A may also extract some of the wording of the answer associated with the question received from the user from the specialized knowledge DB 132 and generate the entire answer by applying the extracted some of the wording of the answer to a predetermined sentence. For example, in response to the question "Please tell me how to improve high blood pressure," the generation unit 124A may extract from the specialized knowledge DB 132 the sentence "People with high blood pressure should limit their salt intake." Furthermore, for example, in response to the question "Please tell me how to improve high blood pressure," the generation unit 124A may extract the words "high blood pressure" and "limit salt intake" from the specialized knowledge DB 132, and apply the extracted words to "XXX" and "YYY" in the sentence "For people with XXX, YYY is a good idea."

[0111] Furthermore, the generation unit 124A inputs a question received from a user into the language generation AI and obtains an answer to the question from the language generation AI. If the answer is extracted from the specialized knowledge DB 132, the generation unit 124A selects the answer extracted from the specialized knowledge DB 132. On the other hand, if the answer is not extracted from the specialized knowledge DB 132, the generation unit 124A selects the answer obtained from the language generation AI.

[0112] The risk level estimation unit 122A estimates the level of risk that a message presented to a user poses to the user. The risk level estimation unit 122A includes a reliability level estimation unit 141, an uncertainty level estimation unit 142, an influence level estimation unit 143, and a risk level calculation unit.

[0113] The reliability estimation unit 141 estimates the user's reliability for the message based on the characteristic information acquired by the characteristic information acquisition unit 121. The reliability estimation method is the same as that in the first embodiment.

[0114] The uncertainty estimation unit 142 estimates the degree of uncertainty of a message depending on the source of the message. The source of the message is either a language generation AI or a specialized knowledge DB 132 in which all or part of the response (message) is stored in advance. The degree of uncertainty when the source of the message is a language generation AI is higher than the degree of uncertainty when the source of the message is the specialized knowledge DB 132. The degree of uncertainty is expressed, for example, as two values, "high" and "low." "High" is, for example, 0.9, and "low" is, for example, 0.1. When the source of the message is a language generation AI, the uncertainty estimation unit 142 estimates the degree of uncertainty as "high," and when the source of the message is the specialized knowledge DB 132, the uncertainty estimation unit 142 estimates the degree of uncertainty as "low." Note that the degree of uncertainty is expressed as two values, "high" and "low." "High" is, for example, 0.9, and "low" is, for example, 0.1. A high degree of uncertainty means that the answer is less reliable, and a low degree of uncertainty means that the answer is more reliable.

[0115] The influence estimation unit 143 estimates an influence indicating the magnitude of the influence that a message has on a user. The influence estimation unit 143 determines a topic addressed in the message. The influence estimation unit 143 estimates an influence according to the determined topic. The topic is, for example, any one of finance, health, medicine, hobbies, preferences, and comfort. The influence when the topic is any one of finance, health, and medicine is higher than the influence when the topic is any one of hobbies, preferences, and comfort.

[0116] The influence estimation unit 143 creates a prompt for determining the topic and inputs the created prompt to the language generation AI.

[0117] FIG. 7 is a diagram showing an example of a prompt to be input to the language generation AI in the second embodiment.

[0118] The language generation AI determines the topics covered in a message through few-shot learning, which solves tasks by presenting a small number of example sentences in context as training data.

[0119] The prompt 300 includes a plurality of example sentences 301, labels 302 for each of the plurality of example sentences, and an answer 303 generated by the generation unit 124A. The labels indicate the topics addressed in the example sentences. When the prompt 300 shown in FIG. 7 is input, the language generation AI outputs the label of the answer 303 by few-shot learning. Here, the language generation AI determines that the topic of the answer 303 generated by the generation unit 124A is "health."

[0120] The memory 13A stores in advance a table associating topic types with influence levels. For example, topics such as health, medicine, and finance are associated with a "high" influence level, and topics such as hobbies, preferences, and comfort are associated with a "low" influence level.

[0121] The influence estimation unit 143 refers to the table and estimates the influence associated with the topic determined by the language generation AI. The influence determined by the language generation AI is expressed as two values: "high" and "low." "High" is, for example, 0.9, and "low" is, for example, 0.1. If the influence is high, the answer has a large influence on the user, and if the influence is low, the answer has a small influence on the user.

[0122] In addition, the influence estimation unit 143 may create a prompt for determining the influence of the answer (message) generated by the generation unit 124A, and input the created prompt to the language generation AI.

[0123] FIG. 8 is a diagram showing another example of a prompt to be input to the language generation AI in the second embodiment.

[0124] The prompt 310 includes a plurality of example sentences 311, labels 312 for each of the plurality of example sentences, a task 313 to be solved by the language generation AI, and an answer 314 generated by the generation unit 124A. The labels indicate the influence of the topic addressed in the example sentence. When the prompt 310 shown in FIG. 8 is input, the language generation AI outputs the influence of the answer 314 by few-shot learning. Here, the language generation AI determines that the influence of the answer 314 generated by the generation unit 124A is "medium." Note that the influence determined by the language generation AI may be expressed as three values: "high," "medium," and "low," rather than as two values: "high" and "low." In this case, "high" is, for example, 0.9, "medium" is, for example, 0.5, and "low" is, for example, 0.1.

[0125] In this way, the influence estimation unit 143 may directly estimate the influence of the answer using language generation AI.

[0126] The risk level calculation unit 144 calculates the risk level R based on the reliability UR estimated by the reliability estimation unit 141, the degree of uncertainty CR estimated by the uncertainty estimation unit 142, and the impact IR estimated by the impact estimation unit 143. More specifically, the risk level calculation unit 144 calculates the risk level R by multiplying the reliability UR, the degree of uncertainty CR, and the impact IR.

[0127] FIG. 9 is a diagram for explaining a method for calculating a risk level in the second embodiment.

[0128] 9, the chat number is a number for identifying the generated reply (message). For example, the risk level R of a reply with chat number "1" is calculated as 0.081 (= 0.9 (reliability UR) * 0.1 (degree of uncertainty CR) * 0.9 (influence IR)).

[0129] In this embodiment 2, the risk level calculation unit 144 calculates the risk level R by multiplying the reliability level UR, the degree of uncertainty CR, and the impact level IR, but the present disclosure is not limited to this, and the risk level R may also be calculated by adding the reliability level UR, the degree of uncertainty CR, and the impact level IR.

[0130] In addition, the risk level calculation unit 144 calculates the risk level R based on the reliability level UR, the degree of uncertainty CR, and the impact level IR, but the present disclosure is not limited to this, and the risk level R may be calculated based on any two of the reliability level UR, the degree of uncertainty CR, and the impact level IR.

[0131] Next, the display process of the server 1A in the second embodiment will be described.

[0132] FIG. 10 is a flowchart showing an example of a display process of the server 1A in the second embodiment.

[0133] First, in step S11 , the characteristic information acquisition unit 121 acquires the user's characteristic information from the user information DB 131 .

[0134] Next, in step S12 , the reliability estimation unit 141 estimates the reliability of the user for the answer based on the characteristic information acquired by the characteristic information acquisition unit 121 .

[0135] Next, in step S13, the question acquisition unit 123 acquires a question received from the user.

[0136] Next, in step S14 , the generation unit 124A generates an answer to the question acquired by the question acquisition unit 123 .

[0137] FIG. 11 is a flowchart showing an example of the response generation process by the generation unit 124A in step S14 of FIG.

[0138] First, in step S31, the generation unit 124A refers to the specialized knowledge DB 132 and extracts an answer associated with the question received from the user from the specialized knowledge DB 132. Note that the answer to the question is not necessarily stored in the specialized knowledge DB 132. Therefore, if the answer to the question is not stored in the specialized knowledge DB 132, the generation unit 124A does not extract the answer from the specialized knowledge DB 132.

[0139] Next, in step S32, the generation unit 124A inputs the question received from the user to the language generation AI, which generates an answer to the input question.

[0140] Next, in step S33, the generation unit 124A acquires the answer generated by the language generation AI.

[0141] Next, in step S34, the generation unit 124A determines whether or not an answer has been extracted from the specialized knowledge DB 132. If it is determined that the answer has been extracted from the specialized knowledge DB 132 (YES in step S34), the generation unit 124A selects the answer extracted from the specialized knowledge DB 132 in step S35.

[0142] On the other hand, if it is determined that the answer has not been extracted from the specialized knowledge DB 132 (NO in step S34), in step S36, the generation unit 124A selects the answer generated by the language generation AI.

[0143] The generation unit 124A extracts an answer to the question from the specialized knowledge DB 132 and inputs the question to the language generation AI. The reason for both extracting an answer from the specialized knowledge DB 132 and inputting a question to the language generation AI is that it takes time for the language generation AI to create an answer.

[0144] Furthermore, the determination of step S34 may be performed after the process of step S31. Then, if it is determined that the answer has not been extracted from the specialized knowledge DB 132 (NO in step S34), the processes of steps S32 and S33 may be performed.

[0145] 10 , next, in step S15, the uncertainty estimation unit 142 estimates the degree of uncertainty of the answer depending on the source of the answer generated by the generation unit 124A. If the generated answer is an answer extracted from the specialized knowledge DB 132, the uncertainty estimation unit 142 estimates the degree of uncertainty as “low.” On the other hand, if the generated answer is an answer generated by a language generation AI, the uncertainty estimation unit 142 estimates the degree of uncertainty as “high.”

[0146] Next, in step S16, the influence estimation unit 143 estimates the influence of the answer. The influence estimation unit 143 determines the topic dealt with in the answer and estimates the influence corresponding to the determined topic.

[0147] Next, in step S17, the risk level calculation unit calculates the risk level based on the reliability estimated by the reliability estimation unit 141, the degree of uncertainty estimated by the uncertainty estimation unit 142, and the impact estimated by the impact estimation unit 143.

[0148] The processing in steps S18 to S21 is the same as the processing in steps S5 to S8 shown in FIG. 3, and therefore a description thereof will be omitted.

[0149] In the first and second embodiments, the risk level, reliability level, degree of uncertainty, and impact level are expressed as two values, "high" and "low," but the present disclosure is not particularly limited to this and may be expressed as three values, "high," "medium," and "low." "High" is, for example, 0.9, "medium" is, for example, 0.5, and "low" is, for example, 0.1.

[0150] Furthermore, in the first and second exemplary embodiments, the determination unit 125 determines whether the risk level is smaller than a threshold value, but the present disclosure is not particularly limited thereto. The determination unit 125 may determine whether the risk level is smaller than a first threshold value. The first threshold value is, for example, 0.3. If it is determined that the risk level is smaller than the first threshold value, the display control unit 126 may cause the answer to be displayed on the display of the information terminal 2. Furthermore, if it is determined that the risk level is equal to or greater than the first threshold value, the determination unit 125 may determine whether the risk level is smaller than a second threshold value that is larger than the first threshold value. The second threshold value is, for example, 0.7. If it is determined that the risk level is smaller than the second threshold value, the display control unit 126 may cause the answer to be displayed on the display of the information terminal 2, and may also cause a warning regarding the accuracy of the answer to be displayed on the display of the information terminal 2 in a first display mode. On the other hand, when it is determined that the risk level is equal to or greater than the second threshold, the display control unit 126 may display the answer on the display of the information terminal 2, and may also display a warning regarding the accuracy of the answer in a second display mode different from the first display mode on the display of the information terminal 2. The degree of emphasis in the second display mode is higher than the degree of emphasis in the first display mode.

[0151] The display screen displayed on the display of the information terminal 2 when the risk level is lower than the first threshold is the same as the display screen 200 shown in Fig. 4. Furthermore, the display screen displayed on the display of the information terminal 2 when the risk level is equal to or higher than the second threshold is the same as the display screen 220 shown in Fig. 5. The warning displayed in the second display mode is the same as the warning 223 shown in Fig. 5.

[0152] FIG. 12 is a diagram showing an example of a display screen 210 that is displayed on the display of the information terminal 2 when the risk level is equal to or greater than the first threshold value and smaller than the second threshold value in the second embodiment.

[0153] The display screen 210 displays a question 201 input by a user, an answer 202 to the question 201, and a warning 213 regarding the accuracy of the answer. In FIG. 12 , the warning 213, "Answers provided by AI may not be accurate," is displayed in the first display mode. The display control unit 126 increases the degree of emphasis of the warning as the risk level increases. That is, the display control unit 126 increases the degree of emphasis of the warning 223 in the second display mode compared to the degree of emphasis of the warning 213 in the first display mode.

[0154] For example, the display control unit 126 may display the attention warning 223 in the second display mode in red, and display the attention warning 213 in the first display mode in a color other than red. Furthermore, the display control unit 126 may make the characters of the attention warning 223 in the second display mode larger than the characters of the attention warning 213 in the first display mode. Furthermore, the display control unit 126 may add an exclamation mark icon to the attention warning 223 in the second display mode, but not add an exclamation mark icon to the attention warning 213 in the first display mode. Furthermore, the display control unit 126 may flash the attention warning 223 in the second display mode, but not flash the attention warning 213 in the first display mode.

[0155] In the first embodiment, the risk level estimation unit 122 includes the reliability estimation unit 141, but the present disclosure is not particularly limited to this. In a first modification of the first embodiment, the risk level estimation unit 122 may include an uncertainty estimation unit 142 instead of the reliability estimation unit 141. In this case, the uncertainty estimation unit 142 may estimate the degree of uncertainty of the message as the risk level depending on the message generation source. The method for estimating the degree of uncertainty is as described in the second embodiment.

[0156] In a second modification of the first embodiment, the risk level estimation unit 122 may include an influence level estimation unit 143 instead of the reliability level estimation unit 141. In this case, the influence level estimation unit 143 may estimate, as the risk level, an influence level that indicates the magnitude of the influence that the message has on the user. The method for estimating the influence level is as described in the second embodiment.

[0157] Furthermore, in the first and second embodiments, the risk level, reliability level, degree of uncertainty, and impact level may be expressed as continuous values ​​ranging from 0.0 to 1.0, for example. In this case, the display control unit 126 may change the degree of emphasis of the warning depending on the value of the risk level. For example, the display control unit 126 may make the font size of the warning larger as the value of the continuous value of the risk level increases.

[0158] (Embodiment 3) In embodiment 1, when the risk level is equal to or greater than a threshold, a warning regarding the accuracy of the answer is displayed along with the answer generated by the language generation AI. In contrast, in embodiment 3, a first answer to a question is generated by the language generation AI, a second answer is generated by the language generation AI by paraphrasing the generated first answer according to the risk level, and the generated second answer is displayed.

[0159] 13 is a diagram showing an example of the configuration of the server 1B in the present embodiment 3. Note that the overall configuration of the information processing system in the present embodiment 3 is the same as the overall configuration of the information processing system in the embodiment 1, and therefore a description thereof will be omitted.

[0160] The server 1B includes a communication unit 11, a processor 12B, and a memory 13. In the third embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0161] The processor 12B includes a characteristic information acquisition unit 121, a risk level estimation unit 122, a question acquisition unit 123, a first generation unit 127, a second generation unit 128, and a display control unit 126B. The characteristic information acquisition unit 121, the risk level estimation unit 122, the question acquisition unit 123, the first generation unit 127, the second generation unit 128, and the display control unit 126B may be realized by the processor 12B executing an information processing program, or may be configured by a dedicated hardware circuit such as an ASIC. The information processing program may be recorded on a non-transitory computer-readable recording medium.

[0162] The first generation unit 127 generates a first answer to a question received from a user. The first generation unit 127 inputs the question received from the user to the language generation AI and acquires a first answer to the question from the language generation AI. The first generation unit 127 transmits the question acquired by the question acquisition unit 123 to the language generation AI server 3 via the communication unit 11. The language generation AI server 3 inputs the received question to the language generation AI and acquires a first answer to the question generated by the language generation AI. The language generation AI server 3 transmits the first answer generated by the language generation AI to the server 1B. The communication unit 11 receives the first answer transmitted by the language generation AI server 3 and outputs it to the first generation unit 127. Note that the first answer generated by the first generation unit 127 is the same as the answer generated by the generation unit 124 in embodiment 1.

[0163] The second generation unit 128 generates a second answer by paraphrasing the expression of the first answer according to the risk level estimated by the risk level estimation unit 122. The second generation unit 128 generates a prompt for paraphrasing the expression of the first answer according to the risk level. The second generation unit 128 inputs the prompt to the language generation AI and obtains a message corresponding to the second answer output from the language generation AI.

[0164] The second generation unit 128 inputs a prompt to the language generation AI requesting that the first answer be rephrased according to the level of risk, and obtains a second answer from the language generation AI in which the first answer has been rephrased according to the level of risk. The second generation unit 128 transmits the generated prompt to the language generation AI server 3 via the communication unit 11. The language generation AI server 3 inputs the received prompt to the language generation AI and obtains a second answer to the prompt generated by the language generation AI. The language generation AI server 3 transmits the second answer generated by the language generation AI to the server 1B. The communication unit 11 receives the second answer transmitted by the language generation AI server 3 and outputs it to the second generation unit 128.

[0165] In the third embodiment, the risk level is expressed as one of three values: "high," "medium," and "low." A "high" risk level is, for example, 0.9, a "medium" risk level is, for example, 0.5, and a "low" risk level is, for example, 0.1.

[0166] The second generation unit 128 determines whether the risk level is less than a first threshold. The first threshold is, for example, 0.3. If the risk level is less than the first threshold, the second generation unit 128 generates a first prompt requesting the user to rephrase the first answer according to a risk level of “low.” If the risk level is equal to or greater than the first threshold, the second generation unit 128 determines whether the risk level is less than a second threshold that is greater than the first threshold. The second threshold is, for example, 0.7. If the risk level is less than the second threshold, the second generation unit 128 generates a second prompt requesting the user to rephrase the first answer according to a risk level of “medium.” On the other hand, if the risk level is equal to or greater than the second threshold, the second generation unit 128 generates a third prompt requesting the user to rephrase the first answer according to a risk level of “high.”

[0167] FIG. 14 is a diagram showing an example of a first prompt 401 generated by the second generation unit 128 and a second answer 402 generated by the language generation AI in the third embodiment.

[0168] The first prompt 401 includes a request statement saying, "Please convey the following health advice to the user as is," and an advice statement saying, "People with high blood pressure should drink as much water as possible." The advice statement is a first answer generated by the first generation unit 127. The request statement is predetermined according to the risk level.

[0169] When the first prompt 401 is input to the language generation AI, the language generation AI generates a second answer 402. The second answer 402 includes the sentence, "People with high blood pressure should drink as much water as possible." If the risk level is "low," the first answer is generated as the second answer.

[0170] FIG. 15 is a diagram showing an example of a second prompt 411 generated by the second generation unit 128 and a second answer 412 generated by the language generation AI in the third embodiment.

[0171] The second prompt 411 includes a request sentence saying, "Please rephrase the following health advice for a user who tends to be somewhat trusting of information," and an advice sentence saying, "People with high blood pressure should drink as much water as possible." The advice sentence is the first answer generated by the first generation unit 127.

[0172] When the second prompt 411 is input to the language generation AI, the language generation AI generates a second answer 412. The second answer 412 includes the sentence, "It is said that it is important for people with high blood pressure to stay hydrated. Drinking plenty of water may help keep your body in good condition." If the risk level is "medium," the second answer 412 is rephrased as a speculative expression rather than an assertive expression.

[0173] FIG. 16 is a diagram showing an example of a third prompt 421 generated by the second generation unit 128 and a second answer 422 generated by the language generation AI in the third embodiment.

[0174] The third prompt 421 includes a request statement saying, "Please rephrase the following health advice for users who are at high risk of accepting the information at face value and acting on it without questioning it," and an advice statement saying, "People with high blood pressure should drink as much water as possible." The advice statement is the first answer generated by the first generation unit 127.

[0175] When the third prompt 421 is input to the language generation AI, the language generation AI generates a second answer 422. The second answer 422 includes the sentence, "It is said that it is important for people with high blood pressure to drink plenty of fluids, but be sure to consult a doctor or specialist to confirm the appropriate amount of fluid intake for you. It is important to receive advice based on your physical condition or situation." If the risk level is "high," the second answer 422 is rephrased as a speculative expression and as a expression urging the person to check with a specialist or the like.

[0176] The display control unit 126B displays the second answer (message) acquired by the second generation unit 128 on the display. The display control unit 126B transmits text information of the second answer (message) to the information terminal 2 via the communication unit 11. The communication unit of the information terminal 2 receives the text information transmitted by the server 1B. The display of the information terminal 2 displays the second answer (message) received by the communication unit.

[0177] Next, the display process of the server 1B in the third embodiment will be described.

[0178] FIG. 17 is a flowchart showing an example of a display process of the server 1B in the third embodiment.

[0179] The processing in steps S41 to S43 is the same as the processing in steps S1 to S3 shown in FIG. 3, and therefore a description thereof will be omitted.

[0180] Next, in step S44, the first generation unit 127 generates a first answer to the question acquired by the question acquisition unit 123. The first generation unit 127 inputs the question to the language generation AI and acquires the first answer to the question from the language generation AI.

[0181] Next, in step S45, the second generation unit 128 generates a second answer by rephrasing the expression of the first answer according to the risk level estimated by the risk level estimation unit 122. The second generation unit 128 generates a prompt for rephrasing the expression of the first answer according to the risk level. The second generation unit 128 inputs the prompt to the language generation AI and obtains from the language generation AI a second answer by rephrasing the expression of the first answer according to the risk level.

[0182] Next, in step S46 , the display control unit 126B causes the second answer generated by the second generation unit 128 to be displayed on the display of the information terminal 2 .

[0183] The process of step S47 is the same as the process of step S8 shown in FIG. 3, and therefore a description thereof will be omitted.

[0184] According to this third embodiment, even if a message contains false information, a message including a warning about the accuracy of the message is generated based on the degree of risk that the message poses to the user, and the generated message is displayed on the display, thereby maintaining users' willingness to use the system while preventing them from being deceived by false information.

[0185] In the third embodiment, the display screen displayed on the display of the information terminal 2 when the risk level is smaller than the first threshold value is the same as the display screen 200 shown in FIG.

[0186] The display screen 200 displays a question 201 entered by a user and an answer 202 to the question 201. The answer 202 is a second answer, but is the same as the first answer. In FIG. 4, the answer 202 displayed in response to the question 201, "Please tell me how to improve high blood pressure," is, "People with high blood pressure should drink as much water as possible." Here, the first answer generated by the language generation AI may be incorrect information.

[0187] If the risk level is less than the first threshold, i.e., if the user's trust in the language generation AI is less than the first threshold, the display screen 200 displays the first answer (answer 202 shown in FIG. 4) generated by the language generation AI without paraphrasing. Users with low trust in the language generation AI tend to have difficulty trusting answers generated by the language generation AI. Therefore, even if the language generation AI generates an incorrect answer, the user is unlikely to be fooled by the incorrect answer, so there is no need to paraphrase the expression.

[0188] FIG. 18 is a diagram showing an example of a display screen 230 that is displayed on the display of the information terminal 2 when the risk level is equal to or greater than the first threshold value and smaller than the second threshold value in the third embodiment.

[0189] Display screen 230 displays question 201 entered by the user and second answer 203 to question 201. In Fig. 18, in response to question 201, "Please tell me how to improve high blood pressure," second answer 203 is displayed, "It is said that it is important for people with high blood pressure to drink plenty of fluids. Drinking plenty of water may help keep your body in good condition."

[0190] If the risk level is greater than or equal to the first threshold and less than the second threshold, i.e., if the user's trust in the language generation AI is greater than or equal to the first threshold and less than the second threshold, the display screen 230 displays a second answer 203 that paraphrases the first answer generated by the language generation AI. A user with a somewhat high level of trust in the language generation AI is likely to believe the answer generated by the language generation AI. Therefore, if the language generation AI generates an incorrect first answer, the user may be inclined to believe the incorrect first answer. Therefore, the first answer needs to be paraphrased for users who tend to be somewhat trusting of information. In the third embodiment, the first answer to the question is paraphrased into a second answer that is expressed in a speculative manner, and the second answer is displayed, thereby reducing the risk that the user will be deceived by an incorrect answer.

[0191] FIG. 19 is a diagram showing an example of a display screen 240 that is displayed on the display of the information terminal 2 when the risk level is equal to or greater than the second threshold value in the third embodiment.

[0192] Display screen 240 displays question 201 entered by the user and second answer 204 to question 201. In Fig. 19, in response to question 201, "Please tell me how to improve high blood pressure," second answer 204 is displayed, which reads, "It is said that it is important for people with high blood pressure to drink plenty of fluids, but be sure to consult a doctor or specialist to find out the appropriate amount of fluid intake for you. It is important to receive advice based on your physical condition or situation."

[0193] If the risk level is equal to or greater than the second threshold, i.e., if the user's trust in the language generation AI is equal to or greater than the second threshold, the display screen 240 displays a second answer 204 that paraphrases the first answer generated by the language generation AI. Users who have a high level of trust in the language generation AI tend to believe the answers generated by the language generation AI. Therefore, if the language generation AI generates an incorrect first answer, the user may believe the incorrect first answer. Therefore, the first answer needs to be paraphrased for users who tend to accept information as is and act accordingly without questioning it. In the third embodiment, the first answer to the question is paraphrased into a second answer that uses wording that warns users not to believe the answer as is, and the second answer is displayed, thereby reducing the risk of the user being deceived by an incorrect answer.

[0194] In the third embodiment, the second answer is generated after the first answer is generated, but the present disclosure is not particularly limited to this. After the first answer is generated, the second generation unit 128 may determine whether the risk level estimated by the reliability estimation unit 141 is smaller than a first threshold. If it is determined that the risk level is smaller than the first threshold, the display control unit 126B may display the first answer generated by the first generation unit 127 on the display of the information terminal 2. On the other hand, if it is determined that the risk level is equal to or greater than the first threshold, the second generation unit 128 may generate a second answer by rephrasing the expression of the first answer according to the risk level.

[0195] (Embodiment 4) In embodiment 4, the risk level is calculated in the same manner as in embodiment 2, the expression of the answer generated by the language generation AI is rephrased according to the risk level, and the rephrased answer is displayed.

[0196] 20 is a diagram showing an example of the configuration of a server 1C in the present embodiment 4. Note that the overall configuration of the information processing system in the present embodiment 4 is the same as the overall configuration of the information processing system in the embodiment 1, and therefore a description thereof will be omitted.

[0197] The server 1C includes a communication unit 11, a processor 12C, and a memory 13A. In the fourth embodiment, the same components as those in the first to third embodiments are denoted by the same reference numerals, and the description thereof will be omitted.

[0198] The processor 12C includes a characteristic information acquisition unit 121, a risk level estimation unit 122A, a question acquisition unit 123, a first generation unit 127C, a second generation unit 128, and a display control unit 126B. The characteristic information acquisition unit 121, the risk level estimation unit 122A, the question acquisition unit 123, the first generation unit 127C, the second generation unit 128, and the display control unit 126B may be realized by the processor 12C executing an information processing program, or may be configured by a dedicated hardware circuit such as an ASIC. The information processing program may be recorded on a non-transitory computer-readable recording medium.

[0199] The first generation unit 127C refers to the specialized knowledge DB 132 and extracts, as the first answer, an answer associated with the question received from the user from the specialized knowledge DB 132. Note that the first generation unit 127C may extract, as the first answer, all of the answers associated with the question received from the user from the specialized knowledge DB 132. Alternatively, the first generation unit 127C may extract, from the specialized knowledge DB 132, some wording of the answer associated with the question received from the user, and apply the extracted some wording of the answer to a predetermined sentence to generate all of the answer as the first answer. The method of extracting an answer by the first generation unit 127C is the same as the method of extracting an answer by the generation unit 124A in the second embodiment.

[0200] Furthermore, the first generation unit 127C inputs a question received from a user to the language generation AI and obtains a first answer to the question from the language generation AI. If the first answer is extracted from the specialized knowledge DB 132, the first generation unit 127C selects the first answer extracted from the specialized knowledge DB 132. On the other hand, if the first answer is not extracted from the specialized knowledge DB 132, the first generation unit 127C selects the answer obtained from the language generation AI.

[0201] Next, the display process of the server 1C in the fourth embodiment will be described.

[0202] FIG. 21 is a flowchart showing an example of a display process of the server 1C in the fourth embodiment.

[0203] The processing in steps S51 to S53 is the same as the processing in steps S11 to S13 shown in FIG. 10, and therefore a description thereof will be omitted.

[0204] Next, in step S54, the first generation unit 127C generates a first answer to the question acquired by the question acquisition unit 123. The first answer generation process by the first generation unit 127C in step S54 is the same as the answer generation process by the generation unit 124A shown in Fig. 11. The first answer generated by the first generation unit 127C is the same as the answer generated by the generation unit 124A.

[0205] The processes in steps S55 to S57 are the same as those in steps S15 to S17 shown in FIG. 10, and therefore will not be described here.

[0206] Furthermore, the processing in steps S58 to S60 is the same as the processing in steps S45 to S47 shown in FIG. 17, and therefore a description thereof will be omitted.

[0207] (Embodiment 5) In embodiment 3, a first answer to a question is generated by the language generation AI, and a second answer is generated by the language generation AI by paraphrasing the expression of the generated first answer according to the risk level. In contrast, in embodiment 5, an answer to a question expressed according to the risk level is generated by the language generation AI. That is, in embodiment 3, an answer to a question expressed according to the risk level is generated by two processes, whereas in embodiment 5, an answer to a question expressed according to the risk level is generated by one process.

[0208] 22 is a diagram showing an example of the configuration of the server 1D in the present embodiment 5. Note that the overall configuration of the information processing system in the present embodiment 5 is the same as the overall configuration of the information processing system in the embodiment 1, and therefore a description thereof will be omitted.

[0209] The server 1D includes a communication unit 11, a processor 12D, and a memory 13. In the fifth embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof will be omitted.

[0210] The processor 12D includes a characteristic information acquisition unit 121, a risk level estimation unit 122, a question acquisition unit 123, a generation unit 124D, and a display control unit 126D. The characteristic information acquisition unit 121, the risk level estimation unit 122, the question acquisition unit 123, the generation unit 124D, and the display control unit 126D may be realized by the processor 12D executing an information processing program, or may be configured by a dedicated hardware circuit such as an ASIC. The information processing program may be recorded on a non-transitory computer-readable recording medium.

[0211] The generation unit 124D generates an answer to a question received from a user according to the risk level estimated by the risk level estimation unit 122. The generation unit 124D generates a prompt that includes the question acquired by the question acquisition unit 123 and user characteristics according to the risk level estimated by the risk level estimation unit 122, and requests that an answer to the question be generated taking the user characteristics into consideration. The memory 13 pre-stores a table that associates risk levels with user characteristics.

[0212] In the fifth embodiment, the risk level is expressed as one of three values: "high," "medium," and "low." A "high" risk level is, for example, 0.9, a "medium" risk level is, for example, 0.5, and a "low" risk level is, for example, 0.1.

[0213] For example, a "low" risk level is associated with the user characteristic "people tend not to believe advice," a "medium" risk level is associated with the user characteristic "people tend to believe advice easily, so careful wording is required," and a "high" risk level is associated with the user characteristic "people tend to accept advice without questioning it and act on it immediately, so wording that calls for caution is required."

[0214] The generation unit 124D inputs a prompt including a question to the language generation AI and acquires text including an answer output from the language generation AI. The generation unit 124D inputs a prompt edited according to the risk level to the language generation AI and acquires text output from the language generation AI. The generation unit 124D inputs the generated prompt to the language generation AI and acquires an answer to the question from the language generation AI. The generation unit 124D transmits the generated prompt to the language generation AI server 3 via the communication unit 11. The language generation AI server 3 inputs the received prompt to the language generation AI and acquires an answer to the question generated by the language generation AI. The language generation AI server 3 transmits the answer generated by the language generation AI to the server 1D. The communication unit 11 receives the answer transmitted by the language generation AI server 3 and outputs it to the generation unit 124D.

[0215] The generation unit 124D determines whether the risk level is less than a first threshold. The first threshold is, for example, 0.3. If the risk level is less than the first threshold, the generation unit 124D acquires user characteristics corresponding to a "low" risk level from the memory 13 and generates a first prompt. If the risk level is equal to or greater than the first threshold, the generation unit 124D determines whether the risk level is less than a second threshold that is greater than the first threshold. The second threshold is, for example, 0.7. If the risk level is less than the second threshold, the generation unit 124D acquires user characteristics corresponding to a "medium" risk level from the memory 13 and generates a second prompt. On the other hand, if the risk level is equal to or greater than the second threshold, the generation unit 124D acquires user characteristics corresponding to a "high" risk level from the memory 13 and generates a third prompt.

[0216] For example, the first prompt includes a request sentence such as "Below are the user's question and user characteristics. Please generate an answer to the user's question taking the user's characteristics into consideration," a question sentence such as "What can I do to lower my blood pressure?", and a user characteristic sentence such as "I tend to not believe advice." The question sentence is a question acquired by the question acquisition unit 123. The user characteristic sentence is a user characteristic acquired according to the risk level.

[0217] When the first prompt is input to the language generation AI, the language generation AI generates an answer. For example, the answer may include the sentence, "To lower blood pressure, it is generally recommended to improve lifestyle habits such as a balanced diet, moderate exercise, stress management, and sufficient sleep." If the risk level is "low," the answer does not include any wording that calls for caution.

[0218] FIG. 23 is a diagram showing an example of a second prompt 431 generated by the generation unit 124D and an answer 432 generated by the language generation AI in the fifth embodiment.

[0219] For example, the second prompt 431 includes a request sentence saying, "Below are the user's question and user characteristics. Please generate an answer to the user's question taking the user's characteristics into consideration," a question sentence saying, "What should I do to lower my blood pressure?", and a user characteristic sentence saying, "Because I tend to believe advice easily, careful wording is required." The question sentence is a question acquired by the question acquisition unit 123. The user characteristic sentence is a user characteristic acquired according to the risk level.

[0220] When the second prompt 431 is input to the language generation AI, the language generation AI generates an answer 432. For example, the answer 432 includes the following sentence: "To lower your blood pressure, it is important to first receive a diagnosis from a doctor and receive appropriate guidance. In general, improving your lifestyle, such as a balanced diet, moderate exercise, stress management, and sufficient sleep, is recommended. However, this is merely general advice, and we recommend that you consult a doctor or specialist about methods that suit your health condition. Please manage your health safely." If the risk level is "medium," the answer 432 includes an expression that encourages careful action.

[0221] FIG. 24 is a diagram showing an example of a third prompt 441 generated by the generation unit 124D and an answer 442 generated by the language generation AI in the fifth embodiment.

[0222] For example, the third prompt 441 includes a request sentence saying, "Below are the user's question and user characteristics. Please generate an answer to the user's question taking the user's characteristics into consideration," a question sentence saying, "What should I do to lower my blood pressure?", and a user characteristic sentence saying, "Since the user tends to accept advice without questioning it and act immediately, a wording that calls attention is needed." The question sentence is a question acquired by the question acquisition unit 123. The user characteristic sentence is a user characteristic acquired according to the risk level.

[0223] When the third prompt 441 is input to the language generation AI, the language generation AI generates an answer 442. For example, the answer 442 includes the following sentence: "You are asking about how to lower your blood pressure. First, it is generally recommended to improve your diet, exercise moderately, and manage stress, but these are merely general advice. Suddenly making major changes to your lifestyle based on your own judgment may pose risks to your health in some cases. In particular, if you are taking medication or have a chronic illness, it is important to first consult a doctor and receive appropriate guidance. Do not accept information lightly, but carefully check the opinions of experts and then proceed with lifestyle improvements within a reasonable range." If the risk level is "high," the answer 442 includes a wording that calls for caution.

[0224] The display control unit 126D displays the answer (message) acquired by the generation unit 124D on a display. The display control unit 126D transmits text information of the answer (message) to the information terminal 2 via the communication unit 11. The communication unit of the information terminal 2 receives the text information transmitted by the server 1D. The display of the information terminal 2 displays the answer (message) received by the communication unit.

[0225] The display control unit 126D causes the display to display a display screen including the text acquired by the generation unit 124D. The content of the display screen differs depending on the risk level. The display control unit 126D causes the display to display a display screen in which the expression of the answer included in the text differs depending on the risk level.

[0226] Next, the display process of the server 1D in the fifth embodiment will be described.

[0227] FIG. 25 is a flowchart showing an example of the display process of the server 1D in the fifth embodiment.

[0228] The processing in steps S71 to S73 is the same as the processing in steps S1 to S3 shown in FIG. 3, and therefore a description thereof will be omitted.

[0229] Next, in step S74, the generation unit 124D generates an answer to the question acquired by the question acquisition unit 123, the answer being written in an expression corresponding to the level of risk estimated by the risk level estimation unit 122. The generation unit 124D generates a prompt that includes the question acquired by the question acquisition unit 123 and user characteristics corresponding to the level of risk estimated by the risk level estimation unit 122, and requests that an answer to the question be generated that takes the user characteristics into consideration. The generation unit 124D inputs the prompt to the language generation AI, and acquires from the language generation AI an answer to the question written in an expression corresponding to the level of risk.

[0230] Next, in step S75, the display control unit 126D causes the answer generated by the generation unit 124D to be displayed on the display of the information terminal 2.

[0231] The process of step S76 is the same as the process of step S8 shown in FIG. 3, and therefore a description thereof will be omitted.

[0232] In the third to fifth embodiments, the risk level, reliability level, degree of uncertainty, and impact level are expressed as three values, "high," "medium," and "low," but the present disclosure is not particularly limited to this, and they may be expressed as two values, "high" and "low." "High" is, for example, 0.9, and "low" is, for example, 0.1.

[0233] Furthermore, in the flowchart of the fourth embodiment shown in FIG. 21 , the processing of step S54 may not be performed, and after the processing of step S57, the generation unit 124D of the fifth embodiment may generate an answer to the question acquired by the question acquisition unit 123, the answer being described in an expression corresponding to the risk level estimated by the risk level estimation unit 122A. In this case, the uncertainty estimation unit 142 may not need to estimate the degree of uncertainty CR. The risk level calculation unit 144 may calculate the risk level R based on the reliability UR estimated by the reliability estimation unit 141 and the influence level IR estimated by the influence estimation unit 143. Furthermore, the influence estimation unit 143 may estimate an influence level indicating the magnitude of the influence of the message on the user from the question acquired by the question acquisition unit 123. The influence estimation unit 143 may determine a topic addressed in the question. The influence estimation unit 143 may estimate the influence level according to the determined topic. The topic may be, for example, any one of finance, health, medicine, hobbies, preferences, and comfort. The specific method for estimating the influence is the same as the method for estimating the influence in the second embodiment.

[0234] Furthermore, some or all of the functions of the device according to the embodiment of the present disclosure may be realized by a processor such as a CPU executing a program.

[0235] Furthermore, all the numbers used above are merely examples to specifically explain the present disclosure, and the present disclosure is not limited to the numbers used as examples.

[0236] The order in which the steps are executed in the above flowchart is merely an example for specifically explaining the present disclosure, and other orders may be used as long as similar effects are obtained. Also, some of the steps may be executed simultaneously (in parallel) with other steps.

[0237] The technology disclosed herein is useful as a technology for presenting messages to users, as it can prevent users from being deceived by false information while maintaining users' willingness to use the system.

Claims

1. An information processing method executed by a computer, comprising: estimating the degree of risk that a message presented to a user poses to the user; determining a display format for a warning regarding the accuracy of the message based on the degree of risk; and displaying the message on a display, and displaying the warning on the display in accordance with the determined display format.

2. The information processing method according to claim 1, wherein determining the display mode includes increasing the degree of emphasis of the warning as the risk level increases.

3. An information processing method as described in claim 1 or 2, further comprising acquiring characteristic information of the user, and estimating the risk level comprises estimating the user's trustworthiness regarding the message as the risk level based on the characteristic information.

4. The information processing method according to claim 3, wherein the characteristic information includes at least one of a diagnostic result regarding the user's personality and a diagnostic result regarding the user's cognitive function.

5. The information processing method according to claim 4, wherein the characteristic information includes the results of any one of an Enneagram diagnosis, an MBTI (Myers-Briggs Type Indicator) diagnosis, a coaching typing diagnosis, and a Big Five personality diagnosis.

6. The information processing method according to claim 4, wherein the characteristic information includes test results of either an emotional intelligence (EQ) test or a cognitive ability test.

7. The information processing method according to claim 3, wherein the characteristic information includes an index representing one of cooperativeness, analytical thinking ability, and self-evaluation.

8. The information processing method according to claim 1 or 2, wherein the risk level estimation includes estimating a degree of uncertainty of the message as the risk level depending on the source of the message.

9. The information processing method of claim 8, wherein the generator is either a generating AI or a database in which all or part of the message is stored in advance, and the degree of uncertainty when the generator is the generating AI is higher than the degree of uncertainty when the generator is the database.

10. An information processing method according to claim 1 or 2, wherein the risk level estimation includes estimating, as the risk level, an impact level indicating the magnitude of the impact that the message will have on the user.

11. The information processing method according to claim 10, wherein estimating the risk level includes determining a topic addressed in the message and estimating the impact level according to the determined topic.

12. The information processing method described in claim 11, wherein the topic is one of finance, health, medicine, hobbies, preferences, and comfort, and the influence level when the topic is one of finance, health, and medicine is higher than the influence level when the topic is one of hobbies, preferences, and comfort.

13. An information processing method as described in claim 1 or 2, further comprising acquiring characteristic information of the user, wherein estimating the risk level comprises: estimating the user's reliability of the message based on the characteristic information; estimating the degree of uncertainty of the message depending on the source of the message; estimating an impact level indicating the magnitude of the impact the message has on the user; and calculating the risk level based on the reliability, the degree of uncertainty, and the impact level.

14. The information processing method according to claim 1 or 2, wherein the message is a response to a question received from the user.

15. The information processing method according to claim 1 or 2, further comprising inputting a question received from the user to a generation AI and obtaining an answer to the question from the generation AI as the message.

16. An information processing device having a processor, wherein the processor estimates the level of risk that a message presented to a user poses to the user, determines a display format for a warning regarding the accuracy of the message based on the level of risk, displays the message on a display, and also displays the warning on the display in accordance with the determined display format.

17. An information processing method executed by a computer, comprising: estimating the level of risk that an answer to a question received from a user poses to the user; inputting a prompt including the question into a generation AI and obtaining text including the answer output from the generation AI; and displaying a display screen including the text on a display, wherein the content of the display screen varies depending on the level of risk.

18. The information processing method of claim 17, wherein obtaining the text includes inputting the prompt edited according to the risk level into the generation AI and obtaining the text output from the generation AI, and displaying the display screen includes displaying on the display a display screen in which the expression of the answer contained in the text differs according to the risk level.

19. An information processing method as described in claim 17, wherein the display of the display screen includes displaying on the display a display screen in which the display mode of a warning regarding the accuracy of the answer contained in the text changes depending on the risk level.

20. An information processing method executed by a computer, comprising: estimating the degree of risk that an answer to a question received from a user poses to the user; generating a prompt for rephrasing the answer according to the degree of risk; inputting the prompt into a generation AI, obtaining a message corresponding to the answer output from the generation AI, and displaying the message on a display.

Images