Overview
- Researchers used a think-aloud method with 20 undergraduates, recording screens and spoken thoughts to see AI choices as they happened.
- Students most often turned to chatbots for ideas, revision, and help when stuck while keeping control of their argument and voice.
- Participants frequently edited, redirected, or rejected AI output rather than accepting it word for word.
- The research team plans to expand the pilot to 100 students to test whether these patterns hold across a larger group.
- Classroom guidance is diverging, with a Harvard professor requiring bounded AI use as a research tool and editor while critics warn such tools feed easy answers and weaken learning.