GoKawiilNature correspondence urges dual testing of student skills with and without AI
A letter in Nature responds to Marc Schmitt's argument that universities should train students to actively argue with AI models and grade that interaction. The correspondence contends that assessing only AI-assisted performance is incomplete, citing a randomized trial of secondary-school math students in which chatbot access boosted practice scores but reduced scores on a later unassisted exam.
GoKawiil's interpretation of the reporting above, not reported fact.
The cited trial suggests that skill demonstrated alongside an AI tool may not reflect retained independent ability, which could mean current assessment methods overstate genuine learning gains. This implies universities may need to design evaluations that separately test unaided competence, not just collaborative performance with AI present.
- A letter responds to Marc Schmitt's Nature piece on teaching students to interact critically with AI
- A cited PNAS trial found chatbot access raised practice scores but lowered unassisted exam performance
- The author argues assessments must measure student ability both with and without AI support
Source: nature.com — Trattner, 2026-11-03
Published there as: “Universities must measure learning, both with and without AI assistance”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.