AI Homework Hack Backfires: Students' Exam Scores Drop 20%, Study Finds
Bastillepost · 1 SOURCES1 day ago2 MIN

Summary
A groundbreaking two-and-a-half-year study tracking 26,811 middle school students has revealed a counterintuitive phenomenon: those who use AI to complete homework assignments finish faster and score higher on homework, yet their exam performance declines by approximately 20%. Researchers from Stockholm University and the University of Hong Kong found that 80% of the grade decline was concentrated among students who treat AI as a substitute rather than a learning tool. The findings, reported by The Economist, challenge assumptions about AI's educational benefits and suggest that the way students use AI matters more than whether they use it at all.
Key Points
- A longitudinal study monitored 26,811 middle school students over two and a half years, involving scholars from Stockholm University and the University of Hong Kong
- Students who used AI to complete homework finished assignments in about 20 minutes with perfect scores, compared to the previous average of two hours per assignment
- The research discovered that 80% of significant grade declines occurred among students who "outsourced" their work to AI rather than using it as a learning aid
- A group of "outlier" students used AI but maintained their exam scores—they asked AI to explain concepts and identify mistakes, then practiced the material themselves
- The study's core finding emphasizes that the distinction lies not in whether students use AI, but in how they use it: as a coach or as a substitute
Why It Matters
This study carries significant implications for Hong Kong's education system, where tutorial centers and private tutoring are already deeply embedded in student culture. As AI tools become more accessible, parents and teachers face the challenge of distinguishing between genuine learning and homework completion that masks knowledge gaps. The research suggests that simply restricting AI use may be insufficient; instead, educators need to teach students how to leverage AI as a coaching tool that enhances understanding rather than replaces effort.
This study carries significant implications for Hong Kong's education system, where tutorial centers and private tutoring are already deeply embedded in student culture. As AI tools become more accessible, parents and teachers face the challenge of distinguishing between genuine learning and homework completion that masks knowledge gaps. The research suggests that simply restricting AI use may be insufficient; instead, educators need to teach students how to leverage AI as a coaching tool that enhances understanding rather than replaces effort.