Could make people dull, one scientist fears. Here's why.

Could make people dull, one scientist fears, as new research warns that large language models may reduce the diversity of human choices. The concern centers on how AI apps produce answers by predicting what comes next, often generating information that matches what is most common or “average” in a population. Sandra Matz, a Columbia Business School professor and co-author of the research, said this can homogenize decisions by steering users toward normative options. The study analyzed more than 110,000 real-world decisions made by 1,000 people and compared results from generic and personalized AI agents. Researchers also used data from a Facebook study called myPersonality, based on personality tests shared for research. Matz argues developers could introduce an “exploration mode” to support more unexpected recommendations and prevent culture from collapsing into a single set of preferences.