This experiment conducted at Bocconi University examined how ChatGPT access and causal reasoning training independently affect student work quality and originality. Researchers found that while AI access improves professional coherence and expert-like quality, critical-thinking training encourages a broader, more unique set of ideas. The results suggest that educational assessments must shift away from evaluating polished final outputs to better measure student originality and underlying reasoning.
Key points
Access to LLMs significantly improves the professional quality and coherence of novice work by closing expertise gaps.
Critical-thinking training increases idea diversity and uniqueness, acting as a functional complement to AI assistance.
Traditional grading rubrics often fail to capture originality and causal reasoning, rewarding only polished and conventional outputs.
Effective student evaluation in the AI era requires shifting from measuring the quality of a final artifact to evaluating the process and depth of reasoning.
