ChatGPT lifted student scores nearly a point

A randomized experiment run at Bocconi University with OpenAI Economic Research found that ChatGPT access and critical-thinking training improved undergraduate work along different axes, and that students given both showed gains on the widest range of measures.

More than 1,000 first-year undergraduates worked on a real business case — marketing recommendations for the university's merchandise store — and were randomly assigned by class period to one of four conditions: ChatGPT access (GPT‑4o), training in causal reasoning, both, or neither. The training was unrelated to AI, teaching cause-and-effect reasoning through a game, examples, questions and feedback. Submissions were scored by trained human graders on a five-point rubric and separately run through automated text analysis measuring idea count and variety, signs of causal reasoning, and similarity to three expert-written submissions.

Students with ChatGPT access scored almost a full point higher on the five-point scale, with more ideas, clearer logic and answers closer to the expert baseline. The training group did not score higher on the rubric — which measured only how well recommendations addressed awareness and use of the store — but produced ideas that were more varied and more distinct from peers', and explained more clearly when their ideas might fail. The combined group matched the training group on idea variety and the ChatGPT group on rubric scores.

OpenAI frames the result as evidence that assessments may need to reward originality, not just polish. The company's post does not link a working paper, report effect sizes beyond the rubric gain, or say whether effects persist beyond the single assignment.