Do Students Rely on AI? Analysis of Student-ChatGPT Conversations from a Field Study

Students struggled to use AI effectively early on: The study found that college students exhibited "overall low reliance on AI and many of them could not effectively use AI for learning," suggesting that simply having access to powerful AI tools doesn't automatically translate to productive educational outcomes without proper training.

Bad habits are hard to break: Once students had unsuccessful initial experiences with ChatGPT, "negative reliance patterns often persisted across interactions," indicating that students had difficulty adjusting their approach after early failures—a concerning pattern for responsible AI adoption that requires intervention.

AI interfaces need built-in safeguards: The researchers recommend that "AI interfaces should be designed with reliance-calibration mechanisms to enhance appropriate reliance," meaning the tools themselves should be engineered to help users develop healthy, balanced relationships with AI rather than over- or under-relying on them.

Better onboarding is essential: The study emphasizes "the need for enhanced onboarding processes to improve student's familiarity and effective use of AI tools," suggesting that institutions cannot assume students will naturally learn to use AI responsibly without structured guidance and training.

Behavioral patterns can predict problematic AI use: The research identified "behavioral metrics [that] strongly predicted AI reliance," offering potential early warning signs that educators and designers could use to identify and support students who are developing unhealthy AI dependency patterns.