Generative AI without guardrails can harm learning: Evidence from high school mathematics

Checked 3 Nov 2026. Non-commercial or no-derivatives license: full text kept internally; only metadata and abstract are indexed and used.

Generative AI is poised to revolutionize how humans work, and has already demonstrated promise in significantly improving human productivity. A key question is how generative AI affects learning-namely, how humans acquire new skills as they perform tasks. Learning is critical to long-term productivity, especially since generative AI is fallible and users must check its outputs. We study this question via a field experiment where we provide nearly a thousand high school math students with access to generative AI tutors. To understand the differential impact of tool design on learning, we deploy two generative AI tutors: one that mimics a standard ChatGPT interface ("GPT Base") and one with prompts designed to safeguard learning ("GPT Tutor"). Consistent with prior work, our results show that having GPT-4 access while solving problems significantly improves performance (48% improvement in grades for GPT Base and 127% for GPT Tutor). However, we additionally find that when access is subsequently taken away, students actually perform worse than those who never had access (17% reduction in grades for GPT Base)-i.e., unfettered access to GPT-4 can harm educational outcomes. These negative learning effects are largely mitigated by the safeguards in GPT Tutor. Without guardrails, students attempt to use GPT-4 as a "crutch" during practice problem sessions, and subsequently perform worse on their own. Thus, decision-makers must be cautious about design choices underlying generative AI deployments to preserve skill learning and long-term productivity.

• Across schooling, workplace, and modeling studies, AI's performance gains evaporate or reverse once assistance is withdrawn. Bastani et al. 2025 found students "actually perform worse than those who never had access" after GPT-4 removal; Wu et al. 2025 found the "performance augmentation effect did not persist in subsequent tasks performed independently by humans"; and Park et al. 2026's model predicts "20% mandatory practice preserves 92% more capability." Trace →
• Whether AI assistance produces durable learning appears to depend on design and context rather than on access alone. Brynjolfsson et al. 2024 found a workplace conversational assistant "facilitates worker learning and improves English fluency," whereas Bastani et al. 2025 found unguarded GPT-4 access harmed student learning—negative effects that were "largely mitigated by the safeguards in GPT Tutor." Trace →
• Wu et al. 2025 and Bastani et al. 2025 converge on a consistent pattern across workplace and classroom experiments: GenAI collaboration boosts immediate performance, but the gains do not persist once the AI is removed—independent performance and intrinsic motivation deteriorate instead. Trace →
• Contrasting Brynjolfsson et al. 2024 with Bastani et al. 2025 suggests that whether AI assistance builds or erodes human skill depends on deployment design: a workplace assistant facilitated worker learning, whereas unguarded GPT access harmed students' learning and safeguarded prompts largely mitigated that harm. Trace →

• Supervision and doctoral education