Artificial Intelligence in User-Centered Innovation Processes: The Impact of Synthetic Users on Empathic User Understanding and Ideation

Abstract

Recent advancements in artificial intelligence (AI) have led to the emergence of synthetic users, AI-generated entities designed to simulate human behavior. They hold substantial potential for innovation and design processes, however, despite growing interest, they remain largely unexplored in research. In particular, little is known about how the usage of synthetic users influences innovation processes and whether the lack of real human interaction compromises essential empathic user understanding. This thesis investigates how interactions with synthetic versus real users influence innovators’ empathy, user insight identification, and idea generation. A laboratory experiment using a post-test-only control group design was conducted with 58 students. Participants conducted semi-structured interviews with either a synthetic user in the experimental group or a real user in the control group. They completed an empathy map to summarize user insights, a self-report empathy questionnaire, and an ideation task based on the identified user’s problem statement. The results show that participants who interacted with synthetic users reported significantly lower levels in the dimensions perspective-taking, emotional contagion, and emotional interest, indicating a reduction in cognitive and affective empathy. No significant differences were found for empathic concern or for the quality and accuracy of identified user insights. Ideas generated in the synthetic user condition were rated significantly lower in novelty and feasibility, while relevance and specificity did not differ between conditions. These findings provide empirical evidence on a trade-off of using synthetic users in innovation processes: while they can offer efficiency gains and support structured user insight extraction, they appear to limit the intensity of empathic engagement and negatively impact ideation outcomes. For practitioners, this implies that synthetic users should be used with caution and primarily to complement, rather than replace, real user research.

Date of Award	2025
Original language	English
Supervisor	Kristiana Roth (Supervisor)