Study: Generative AI predicts personality traits based on open-ended narratives
If you say a few words, generative AI will understand who you are-maybe even better than your close family and friends.
A new University of Michigan study found that widely available generative AI models (e.g., ChatGPT, Claude, LLaMa) can predict personality, key behaviors and daily emotions as or even more accurately than those closest to you.
"What this study shows is AI can also help us understand ourselves better, providing insights into what makes us most human, our personalities," said the study's first author Aidan Wright, U-M professor of psychology and psychiatry. "Lots of people may find this of interest and useful. People have long been interested in understanding themselves better. Online personality questionnaires, some valid and many of dubious quality, are enormously popular."
Researchers looked into whether AI programs like ChatGPT and Claude can act like general "judges" of personality. To test this, they had the AI read people's own words-either short daily video diaries or longer recordings of what happened to be on their mind-and asked it to answer personality questions the way each person would. The study included stories and thoughts from more than 160 people collected in real-life and lab settings.
The results showed that the AI's personality scores were very similar to how people rated themselves, and often matched them better than ratings from friends or family. Older text-analysis methods did not perform nearly as well as these newer AI systems.
"We were taken aback by just how strong these associations were, given how different these two data sources are," Wright said.
AI's personality ratings could also predict real parts of people's lives, like their emotions, stress levels, social behavior and even whether they had been diagnosed with mental health conditions or sought treatment, according to the findings.
This research indicates that personality naturally shows up in our everyday thoughts, words and stories-even when we're not trying to describe ourselves.
Chandra Sripada, U-M professor of philosophy and psychiatry, says the findings support the long-held idea that language carries deep clues about how people differ in psychological traits such as personality and mood. He adds that open-ended writing and speech can be a powerful tool for understanding personality. Thanks to generative AI, researchers can now analyze this kind of data quickly and accurately in ways that weren't possible before.
At the same time, important questions remain. The study relied on people rating their own personalities and did not test how well AI compares with judgments from friends or family, or how results might differ across age, gender or race.
Researchers also don't yet know whether AI and humans rely on the same signals-or whether AI could one day outperform self-reports when predicting major life outcomes like relationships, education, health, or career success.
"The study shows that AI can reliably uncover personality traits from everyday language, pointing to a new frontier in understanding human psychology," said Colin Vize, assistant professor of psychology at the University of Pittsburgh.
Whitney Ringwald, assistant professor of psychology at the University of Minnesota, says the results "really highlight how our personality is infused in everything we do, even down to our mundane, everyday experiences and passing thoughts."
The study's other authors were Johannes Eichstaedt of Stanford University and Mike Angstadt and Aman Taxali, both from U-M. The findings appear in the journal Nature Human Behavior.
