The Wharton School Research Paper: Generative AI Can Harm Learning

“Generative artificial intelligence (AI) is poised to revolutionize how humans work…”

In this 2024 study, the authors examine how access to generative AI, specifically GPT-4, influences human learning in real educational environments. Conducted across three high school grade levels and involving nearly one thousand students, the experiment compared two AI tutors: a standard GPT-based model and a modified version with built-in learning safeguards. Both versions improved short-term performance, but when AI access was removed, students who had used the unrestricted version performed worse than those without AI support. The findings suggest that overreliance on AI during practice can impede independent problem-solving and long-term skill development.

The research underscores the need for thoughtful design and regulation of AI tools in education. While AI can enhance learning outcomes, it must be implemented with intentional safeguards that promote reflection, reasoning, and gradual autonomy. The authors conclude that effective educational use of generative AI requires human-centered strategies to ensure technology strengthens, rather than replaces, core cognitive skills.