Source
Times Higher Education
Summary
Thibault Schrepel argues against speculation and for empirical classroom experiments to measure how generative AI truly affects student learning. He outlines simple, scalable experimental designs—e.g. groups forbidden from AI, groups using it without guidance, groups trained in prompting and critique—to compare outcomes in recall, writing quality, and reasoning. Schrepel also suggests activities like having students build AI research assistants, comparing human and AI summaries, and using AI as a Socratic tutor. He emphasises that AI won’t uniformly help or hurt; its impact depends on how it’s used, taught, and assessed.
Key Points
- Use controlled classroom experiments with different levels of AI access/training to reveal real effects.
- Recall or rote learning may not change much; AI’s effects show more in reasoning, argumentation and writing quality.
- Activities like comparing AI vs human summaries or having AI play the role of interlocutor can highlight strengths and limitations.
- Prompting, critique, and metacognitive reflection are central to converting AI from crutch to tool.
- Banning AI outright is less useful than enabling pedagogical experimentation and shared insight across faculty.
Keywords
URL
https://www.timeshighereducation.com/campus/how-test-genais-impact-learning
Summary generated by ChatGPT 5
