ChatGPT and Critical-Thinking Training Lift Student Outcomes

As generative AI spreads through higher education, the debate is shifting from whether students should use it to how schools can preserve independent reasoning and measure genuine learning. Researchers at Bocconi University, working with OpenAI Economic Research, tested ChatGPT alongside causal-reasoning instruction on a real business assignment. The randomized design offers stronger evidence that AI and critical-thinking skills can improve different dimensions of student work rather than acting as substitutes.

OpenAI released the findings on Aug. 27, 2026. More than 1,000 first-year Bocconi students were assigned by class period to receive GPT-4o access, causal-reasoning training, both interventions or neither while developing marketing proposals for the university merchandise store. ChatGPT users scored almost one point higher on a five-point rubric. The reasoning exercise did not raise conventional scores but produced more distinctive ideas, while students receiving both interventions showed the broadest gains across quality, logical coherence and originality.