The Creative Link Between Words and Ideas is Weakening in the AI Era

Le résumé fourni par la source

Concerns that AI homogenizes human thinking appear at odds with findings that LLM writing is perceived as more creative than human writing. We propose that LLMs enhance superficial semantic diversity while simultaneously homogenizing underlying ideas. Four large-scale natural experiments tested this preregistered “disjunctive homogenization” hypothesis in 372,793 personal statements written in high-stakes college admissions contexts. Comparing before-versus-after ChatGPT’s release revealed increasing word-level diversity simultaneous with decreasing conceptual diversity at sentence and whole-document levels. Controlled experiments provided causal evidence of AI’s influence and identified a plausible mechanism: a positive association of word-to-idea diversity in humans was reversed in GPT. Despite conceptual homogenization, raters perceived post-ChatGPT essays as more creative due to increased word-level diversity, even when considering multiple essays. Disjunctive trends were strongest among minoritized applicants.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.