Are AI Tutors Making Students Learn Less?
Shownotes
What if giving students access to AI tutors actually makes them learn less?
This week, Ray and Dan examine two new randomised trials that challenge some of the assumptions about AI in education. One found that AI tutoring was as effective as expert human tutoring, but students covered less material and saw no improvement in final test scores. Another found that students offered AI tutors performed worse, participated less, and rarely used the tools.
We explore what these findings mean for student motivation, engagement and the future of teaching.
News this episode:
- Universities embrace ChatGPT: Princeton and Carnegie Mellon announce wider access, while Dartmouth's provost resigns amid controversy over AI-written content.
- Google changes course: Gemini will no longer be enabled by default in schools, raising questions about AI and Google Search. Google's blog, and follow up clarification tweet Reporting on it: NY Times story and The Atlantic Story
- Rethinking assessment: Microsoft's medical AI highlights why traditional exams may not reflect real-world skills.
- AI policy in Australia: Major changes to national education bodies and renewed debate about screen time and testing.
Research
- Measuring the impact of AI tutoring, from near transfer to standardised tests: An expanded RCT in UK mathematics classrooms [Blog post] [Research Paper]
- AI humanizers and the political economy of academic integrity in higher education
- The Effects of Course-Integrated AI Tutoring on Student Performance and Engagement: A Randomized University Trial