AI Business LensTHE BUSINESS OF AI, FOR PEOPLE WHO TEACH IT OR LEARN FROM IT
Phys.org — Technology · September 7, 2026

AI digital twins struggle to predict human behavior, creating 'funhouse mirror' distortions

PsychologySociologyMarketing
THE AI ANGLE
Simulating human behavior and decision-making for research and polling

A large-scale study published in Science Advances found that AI 'digital twins' fail to reliably predict human behavior, acting instead like 'funhouse mirrors' that systematically distort real human choices across tasks like hiring, privacy, and political attitudes. Despite being trained on over 500 individual responses, the models performed only slightly better than generic AI and introduced flaws such as socioeconomic bias, demographic stereotyping, and excessive rationality. For instructors in behavioral and market research, these findings highlight critical limitations in using synthetic AI agents to replace human research subjects or survey respondents.

THE TEACHING ANGLE
Students can evaluate the scientific and ethical risks of substituting human participants with AI, examining why models fail by relying on broad demographic stereotypes and unrealistic rationality instead of capturing true individual decision-making.

Read the original at techxplore.com   Generate teaching or study materials

Instructors get discussion guides, assignments, and mini-cases. Students and readers get a plain summary, class prep, and an exercise. All built from the full article. Three are free with an account.

More in Psychology