Research & Science

AI helped students start research, then tempted them to hand it off

In a graduate software course, AI helped students find starting points for research reviews. It also pulled some attention away from judging evidence and toward using the tool itself. The authors watched one day of work by ten doctoral students in three groups. Their report supports careful supervision, not a claim that AI improves research training.

Beyond Prompting
Read original source

What the source reports

Danilo Monteiro Ribeiro and Gilberto Sussumu Hida watched ten doctoral students in three groups. They worked in a one-day graduate software class. The students tried planned reviews of past studies, both with and without AI help. The authors read class notes, student work, and chats with AI helpers. AI helped students get started and find options. It also brought risks. Some students handed over too much work, checked results too lightly, and spent time on the tool instead of the review. The authors call for human checks, records of choices, and care about the tool’s limits. This is one class session. It is not a controlled test of learning gains. The authors plan such a study later.

Original source

Title
Observing the Conduct of Systematic Reviews with Generative AI Support: An Experience Report from a Graduate Software Engineering Course
Author
Danilo Monteiro Ribeiro, Gilberto Sussumu Hida
Publication
arXiv
Date
Wednesday, September 23, 2026