AI and critical thinking: enhancing student outcomes in complementary ways

A recent study conducted by researchers at Bocconi University in collaboration with OpenAI Economic Research has revealed that access to ChatGPT and training in causal reasoning can significantly enhance student performance in distinct yet complementary ways. The experiment involved over 1,000 first-year undergraduate students who were randomly assigned to one of four groups: those with access to ChatGPT (GPT-4o), those who received training in causal reasoning, those who received both, and a control group that received neither. The students were tasked with developing marketing recommendations for the university’s merchandise store as part of a real-world assignment. Students who had access to ChatGPT scored almost a full point higher on a five-point rubric, with their answers showing improved quality, coherence, and similarity to expert recommendations. Despite this, the students were not simply outsourcing their work to the AI; they still had to evaluate and select the best responses for their final submissions. Meanwhile, the causal reasoning training led to students generating more unique ideas, though their scores on the rubric did not improve. The rubric, which focused on how well recommendations addressed standard marketing goals, did not capture the broader range of ideas produced by students who completed the critical-thinking exercise. Text analysis revealed that students who underwent the causal reasoning training produced a wider variety of distinct ideas compared to their peers. This highlights a gap in traditional assessment methods, which may overlook originality and the depth of reasoning. The study also emphasized that the combination of ChatGPT access and critical-thinking training yielded the most comprehensive benefits, with students demonstrating improved logical coherence, a wider range of ideas, and stronger evidence of questioning assumptions. The randomized design of the experiment allowed researchers to isolate the effects of each intervention, contributing to the growing body of research on AI’s impact on education. The findings suggest that assessments should evolve to reward originality, reasoning, and consideration of multiple approaches, rather than focusing solely on polished answers. The study underscores the importance of a holistic approach to evaluating student progress, especially as AI continues to reshape educational practices and learning outcomes.