Empirical study: generative AI harms half of students, no benefit for the rest
birchlse · x · 2026-08-30
Benjamin Riley's Substack essay "The 50 Percent Problem" summarizes rigorous empirical evidence on students using generative AI. The research indicates AI substantially harms learning for about half of students — what researchers call the "Generative AI penalty" — while providing no discernible benefit to the other half.
By the study's end, 50% of regular AI users had fully outsourced their homework to generative AI and simply stopped thinking about schoolwork at home. Riley notes that nearly four years after ChatGPT launched, high-quality empirical research remains scarce: a Stanford landscape analysis found only 20 of 800+ studies of AI in education used true causal measures.
More from AGI Musings
- 1996 Sugarscape Model: Early Origins of Agent Tech — generativist · 2026-09-01
- New paper: a structured ladder for scaling large reasoning models beyond human supervision — Zhiqin Yang · 2026-09-01
- Trust: The Biggest Barrier and Driver for Personal Agent Adoption — petergyang · 2026-09-01
- The Next AI Revolution Won't Be One Assistant. It Will Be An Entire Team of AI Agents — CurieuxExplorer · 2026-09-01
- Agentic commerce is the future, replacing websites — thisiskp_ · 2026-09-01
- Is AI Leading Us Toward Singularity? The Shift in Control and Power — CurieuxExplorer · 2026-09-01