Dispatch
Bocconi and OpenAI randomized 1,000+ students and found ChatGPT and critical-thinking training fix different things
Students were randomly assigned by class period to one of four arms, GPT-4o access, causal-reasoning training, both, or neither, then asked to build a marketing strategy for Bocconi's merchandise store. ChatGPT access improved the quality and coherence of submissions; the causal-reasoning exercise increased the number of unique ideas; students who got both showed both effects. Grading combined a five-point human rubric with automated text analysis measuring idea count and variety, evidence of causal reasoning, and similarity to three expert submissions, which is a cleaner design than the survey studies that dominate this literature.
Source
↳ Follow the thread