ChatGPT and Critical Thinking Training Improve Students' Work Quality and Originality
An experiment conducted with over a thousand university students shows that combining ChatGPT support with critical thinking training provides the greatest benefit and better prepares students for the future.
A joint study conducted by Bocconi University and OpenAI's Economic Research unit has revealed that AI use and critical thinking training have complementary effects on students. According to the research, while access to ChatGPT improves the quality and logical coherence of students' work, causal reasoning training enables them to generate more original ideas.
Experiment Methodology and Participants
More than 1,000 first-year undergraduate students studying at Bocconi University worked on a real business case where they developed marketing proposals for the university's store. Students were randomly divided into four groups: those with access to ChatGPT (GPT-4o), those who received causal reasoning training, those with both, and those with neither.
Causal reasoning training was conducted independently of AI through an exercise involving games, examples, and feedback. The work submitted by the students was evaluated both by trained graders using a five-point assessment scale and through text analysis tools.
Different Effects of ChatGPT and Critical Thinking
Students with access to ChatGPT scored nearly one point higher on the assessment scale. These students' work contained more ideas, followed a clearer logical sequence, and showed greater similarity to expert recommendations.
While students who received critical thinking training did not score higher on the traditional assessment scale, they explained more clearly why their ideas would or would not work. Text analyses revealed that these students generated a much broader and more original range of ideas compared to their peers.
The Complementary Power of Both Methods and a New Era in Education
The group of students who were provided with both AI access and critical thinking training demonstrated the advantages of both methods combined. The diversity of ideas generated by these students matched those who only received training, while their scores and logical consistency were as high as those who only used ChatGPT.
Researchers point out that in this era where AI makes it easier for students to appear like experts, schools may need to change their assessment methods. The importance of evaluating originality, reasoning, and diverse approaches, rather than focusing solely on the final outcome, is emphasized.