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

Generative AI in education refers to AI systems that produce new text, images, audio, or code in response to prompts — applied to learning contexts. Examples include ChatGPT generating essay feedback, DALL-E creating visual aids for lessons, and AI tools like Curipod building interactive lesson slides from a topic prompt.

Generative AI benefits schools by dramatically reducing the time teachers spend creating differentiated materials, providing students with personalized explanations at scale, enabling instant practice question generation for any topic, and making content creation accessible to educators without specialized technical skills.

Risks include students submitting AI-generated work as their own, exposure to inaccurate AI outputs presented confidently, potential reduction of writing and research skills from over-reliance, and privacy concerns from data sharing with AI vendors. Schools that teach students to critically evaluate AI outputs mitigate many of these risks.

Teachers use generative AI to draft lesson plans in minutes, create rubrics aligned to standards, adapt reading passages for different grade levels, generate discussion questions, and produce multiple versions of assessments to reduce cheating. Magic School AI and Diffit are among the most widely adopted tools for these tasks.