Search

US-20260127652-A1 - GENERATIVE ARTIFICIAL INTELLIGENCE-ENABLED AND AUGMENTED USER RESEARCH USING SYNTHETIC PERSONAS AS SUBJECTS

US20260127652A1US 20260127652 A1US20260127652 A1US 20260127652A1US-20260127652-A1

Abstract

An approach is provided for generative artificial intelligence (GenAI)-enabled user research using synthetic personas. Data for user research for a product or service is collected. The data is included in categories of individual profile data for users, product feature data, company metadata, and historical data. The collected data is evaluated using a neural network. Scores for types of data in each category are determined based on the evaluated data. An overall fit score is determined by aggregating the scores. Based on the overall fit score, a recommendation of areas of focus is generated. The areas of focus are associated with the product or service and require an evaluation. One or more synthetic personas are generated based on the individual profile data. The evaluation of the areas of focus is performed by using the one or more synthetic personas.

Inventors

  • Jennifer M. Hatfield
  • Lucia Larise Stavarache
  • Michael Jack Martine
  • Sarah Diane Green
  • Mark J. LUDLOW
  • Ira L. Allen

Assignees

  • INTERNATIONAL BUSINESS MACHINES CORPORATION

Dates

Publication Date
20260507
Application Date
20241105

Claims (20)

  1. 1 . A computer-implemented method comprising: collecting data for user research for a product or a service, the data being included in a first category of individual profile data for a plurality of users, a second category of product feature data, a third category of company metadata, and a fourth category of historical data; evaluating the collected data using a neural network; determining scores for types of data in each category based on the evaluated data; determining an overall fit score by aggregating the scores; based on the overall fit score, generating a recommendation of areas of focus associated with the product or service that require an evaluation; generating, by a processor set, one or more synthetic personas based on the individual profile data; and performing the evaluation of the areas of focus by using the one or more synthetic personas.
  2. 2 . The method of claim 1 , further comprising: defining an architecture of a structured model; training the structured model using an iterative optimization algorithm that minimizes loss; and evaluating an output of the trained structured model, wherein the evaluating the output includes identifying a gap in an expertise of the plurality of users, the expertise being included in the evaluated collected data, and the gap indicating that the expertise of the plurality of users is inadequate for an evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the expertise of the plurality of users, and includes generating the one or more synthetic personas to include additional expertise that corrects the gap and is adequate for the evaluation of the areas of focus by the one or more synthetic personas.
  3. 3 . The method of claim 2 , wherein the evaluating the output of the trained structured model further includes identifying a gap in a demographic representation of the plurality of users, the demographic representation being specified in the evaluated collected data, and the gap in the demographic representation indicating that the demographic representation of the plurality of users is inadequate for an unbiased evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the demographic representation of the plurality of users, and includes generating the one or more synthetic personas to include additional demographic representation that corrects the gap in the demographic representation and is adequate for an unbiased evaluation of the areas of focus by the one or more synthetic personas.
  4. 4 . The method of claim 3 , further comprising: performing a final evaluation and harmonization of the structured model by verifying an improvement in the scores; and performing a feedback loop that updates the structured model with new data based on the evaluation of the areas of focus by the one or more synthetic personas.
  5. 5 . The method of claim 1 , further comprising: generating multiple synthetic personas based on the individual profile data, the multiple synthetic personas including at least a first synthetic persona and a second synthetic persona; and collecting information about a conversation among the multiple synthetic personas, wherein the conversation includes a first feedback about the product or the service from the first synthetic persona and a second feedback about the product or the service from the second synthetic persona, the second feedback being based on a processing of the first feedback by the second synthetic persona.
  6. 6 . The method of claim 1 , further comprising: generating multiple synthetic personas based on the individual profile data, the multiple synthetic personas having different respective modalities of thinking and learning; and performing the evaluation of the areas of focus by using the multiple synthetic personas based on the multiple synthetic personas having the different respective modalities of thinking and learning.
  7. 7 . The method of claim 1 , wherein the performing the evaluation includes collecting feedback from the one or more synthetic personas, wherein the feedback includes a description of one or more characteristics and a recommendation that the one ore more characteristics be added to the product or the service, and wherein the recommended one or more characteristics do not currently exist in the product or the service.
  8. 8 . The method of claim 1 , wherein the performing the evaluation includes reverting a state of a synthetic persona included in the one or more synthetic personas to a state that the synthetic persona had at an earlier phase in the evaluation, the earlier phase being earlier than a current phase of the evaluation, wherein the reverting causes the synthetic persona to selectively forget aspects of an experiment included in the evaluation, the experiment being conducted subsequent to the earlier phase.
  9. 9 . The method of claim 1 , further comprising: generating synthetic personas that simulate users located in a geo-location, wherein the performing the evaluation includes: obtaining feedback about the product or the service from the synthetic personas, the obtained feedback being based at least in part on the geo-location in which the users being simulated by the synthetic personas are located; and generating a recommendation to modify the product or the service based on the obtained feedback and the geo-location, the recommendation being generated exclusively for users located in the geo-location.
  10. 10 . A computer system comprising: a processor set; one or more computer-readable storage media; and program instructions stored on the one or more computer-readable storage media to cause the processor set to perform operations comprising: collecting data for user research for a product or a service, the data being included in a first category of individual profile data for a plurality of users, a second category of product feature data, a third category of company metadata, and a fourth category of historical data; evaluating the collected data using a neural network; determining scores for types of data in each category based on the evaluated data; determining an overall fit score by aggregating the scores; based on the overall fit score, generating a recommendation of areas of focus associated with the product or service that require an evaluation; generating one or more synthetic personas based on the individual profile data; and performing the evaluation of the areas of focus by using the one or more synthetic personas.
  11. 11 . The computer system of claim 10 , wherein the operations further comprise: defining an architecture of a structured model; training the structured model using an iterative optimization algorithm that minimizes loss; and evaluating an output of the trained structured model, wherein the evaluating the output includes identifying a gap in an expertise of the plurality of users, the expertise being included in the evaluated collected data, and the gap indicating that the expertise of the plurality of users is inadequate for an evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the expertise of the plurality of users, and includes generating the one or more synthetic personas to include additional expertise that corrects the gap and is adequate for the evaluation of the areas of focus by the one or more synthetic personas.
  12. 12 . The computer system of claim 11 , wherein the evaluating the output of the trained structured model further includes identifying a gap in a demographic representation of the plurality of users, the demographic representation being specified in the evaluated collected data, and the gap in the demographic representation indicating that the demographic representation of the plurality of users is inadequate for an unbiased evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the demographic representation of the plurality of users, and includes generating the one or more synthetic personas to include additional demographic representation that corrects the gap in the demographic representation and is adequate for an unbiased evaluation of the areas of focus by the one or more synthetic personas.
  13. 13 . The computer system of claim 12 , wherein the operations further comprise: performing a final evaluation and harmonization of the structured model by verifying an improvement in the scores; and performing a feedback loop that updates the structured model with new data based on the evaluation of the areas of focus by the one or more synthetic personas.
  14. 14 . The computer system of claim 10 , wherein the operations further comprise: generating multiple synthetic personas based on the individual profile data, the multiple synthetic personas including at least a first synthetic persona and a second synthetic persona; and collecting information about a conversation among the multiple synthetic personas, wherein the conversation includes a first feedback about the product or the service from the first synthetic persona and a second feedback about the product or the service from the second synthetic persona, the second feedback being based on a processing of the first feedback by the second synthetic persona.
  15. 15 . The computer system of claim 10 , wherein the operations further comprise: generating multiple synthetic personas based on the individual profile data, the multiple synthetic personas having different respective modalities of thinking and learning; and performing the evaluation of the areas of focus by using the multiple synthetic personas based on the multiple synthetic personas having the different respective modalities of thinking and learning.
  16. 16 . The computer system of claim 10 , wherein the performing the evaluation includes collecting feedback from the one or more synthetic personas, wherein the feedback includes a description of one or more characteristics and a recommendation that the one ore more characteristics be added to the product or the service, and wherein the recommended one or more characteristics do not currently exist in the product or the service.
  17. 17 . A computer program product comprising: one or more computer-readable storage media; and program instructions stored on the one or more computer-readable storage media to perform operations comprising: collecting data for user research for a product or a service, the data being included in a first category of individual profile data for a plurality of users, a second category of product feature data, a third category of company metadata, and a fourth category of historical data; evaluating the collected data using a neural network; determining scores for types of data in each category based on the evaluated data; determining an overall fit score by aggregating the scores; based on the overall fit score, generating a recommendation of areas of focus associated with the product or service that require an evaluation; generating one or more synthetic personas based on the individual profile data; and performing the evaluation of the areas of focus by using the one or more synthetic personas.
  18. 18 . The computer program product of claim 17 , wherein the operations further comprise: defining an architecture of a structured model; training the structured model using an iterative optimization algorithm that minimizes loss; and evaluating an output of the trained structured model, wherein the evaluating the output includes identifying a gap in an expertise of the plurality of users, the expertise being included in the evaluated collected data, and the gap indicating that the expertise of the plurality of users is inadequate for an evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the expertise of the plurality of users, and includes generating the one or more synthetic personas to include additional expertise that corrects the gap and is adequate for the evaluation of the areas of focus by the one or more synthetic personas.
  19. 19 . The computer program product of claim 18 , wherein the evaluating the output of the trained structured model further includes identifying a gap in a demographic representation of the plurality of users, the demographic representation being specified in the evaluated collected data, and the gap in the demographic representation indicating that the demographic representation of the plurality of users is inadequate for an unbiased evaluation of the areas of focus by the plurality of users, and wherein the generating the one or more synthetic personas is performed in response to the identifying the gap in the demographic representation of the plurality of users, and includes generating the one or more synthetic personas to include additional demographic representation that corrects the gap in the demographic representation and is adequate for an unbiased evaluation of the areas of focus by the one or more synthetic personas.
  20. 20 . The computer program product of claim 19 , wherein the operations further comprise: performing a final evaluation and harmonization of the structured model by verifying an improvement in the scores; and performing a feedback loop that updates the structured model with new data based on the evaluation of the areas of focus by the one or more synthetic personas.

Description

BACKGROUND The present invention relates to user research, and more particularly to user research enhanced by generative artificial intelligence (GenAI). SUMMARY In one embodiment, the present invention provides a computer-implemented method. The method includes collecting data for user research for a product or a service. The data is included in a first category of individual profile data for a plurality of users, a second category of product feature data, a third category of company metadata, and a fourth category of historical data. The method further includes evaluating the collected data using a neural network. The method further includes determining scores for types of data in each category based on the evaluated data. The method further includes determining an overall fit score by aggregating the scores. The method further includes, based on the overall fit score, generating a recommendation of areas of focus associated with the product or service that require an evaluation. The method further includes generating, by a processor set, one or more synthetic personas based on the individual profile data. The method further includes performing the evaluation of the areas of focus by using the one or more synthetic personas. In one embodiment, the present invention provides another computer-implemented method. The method includes defining an initial subject for user research. The initial subject is required to be evaluated in the user research. The method further includes identifying user characteristics that one or more users are required to have to evaluate the initial subject. The user characteristics include (i) levels of experience and expertise associated with using the product or the service, (ii) cultural backgrounds of the one or more users, and (iii) learning and thinking modalities of the one or more users. The method further includes identifying one or more synthetic personas whose characteristics match the identified user characteristics. The method further includes collecting feedback about the initial subject from the one or more synthetic personas. The method further includes receiving a validation of the collected feedback from a user group consisting of humans. The method further includes, based on the collected feedback and in response to receiving the validation, updating the initial subject to generate an updated subject. The method further includes repeating the defining, the identifying the user characteristics, the identifying the one or more synthetic personas, the collecting the feedback, and the receiving the validation, with the initial subject being replaced by the updated subject. Respective computer systems and computer program products corresponding to the above-summarized computer-implemented methods are also described herein. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of a system for GenAI-enabled and augmented user research using synthetic personas as subjects, in accordance with embodiments of the present invention. FIG. 2 is a block diagram of a first set of modules included in code included in the system of FIG. 1, in accordance with embodiments of the present invention. FIG. 3 is a flowchart of a process of GenAI-enabled and augmented user research using synthetic personas as subjects, where operations of the flowchart are performed by modules in FIG. 2, in accordance with embodiments of the present invention. FIG. 4 is a block diagram of a system that implements the process of FIG. 3, in accordance with embodiments of the present invention. FIG. 5 is a flowchart of another process of GenAI-enabled and augmented user research using synthetic personas as subjects, in accordance with embodiments of the present invention. FIGS. 6A-6C depict an example of an implementation of a user research augmentation system that performs the operations in the flowchart of FIG. 3, in accordance with embodiments of the present invention. DETAILED DESCRIPTION Overview User research is time-consuming, expensive, and is complicated by the necessity to find representative sample users. Human focus groups are expensive, time-consuming, and resource-intensive. Furthermore, researchers do not always have easy access to human users. It is not feasible for user researchers to have their human research subjects interact with each other over extended periods of time and generate and test ideas. Experiments that user researchers can run are limited because their human subjects remember previous iterations, changes in variables, etc. Companies struggle to appeal to a new geographic location and/or culture, because of difficulties in not only empaneling a truly representative human focus group, but also knowing the right questions to ask members of the focus group. User researchers have limited visibility into the thinking and reasoning modalities of their human focus groups and test subjects. Many conventional approaches for an organization's user research use humans as focus group