AI Chatbots Impair Problem-Solving Skills in Just 10 Minutes, Research Finds

Summary

A new study from researchers at Carnegie Mellon, MIT, Oxford, and UCLA reveals that using AI chatbots for even brief periods significantly degrades human problem-solving abilities and persistence. The researchers conducted three experiments with hundreds of participants tasked with solving problems like fractions and reading comprehension. When AI assistance was available, users performed better initially, but when the AI was removed, those same users were significantly more likely to give up or produce incorrect answers compared to control groups who solved problems without AI assistance from the start.

The findings suggest that while AI tools boost immediate productivity, they may come at the cost of developing fundamental problem-solving skills and cognitive resilience. Michiel Bakker, an MIT assistant professor involved in the study, argues that the issue isn't with AI itself, but with how it's deployed. The research indicates that systems providing direct answers may have different long-term effects than systems that coach, scaffold, or challenge users. This raises important questions about how AI assistants should be integrated into educational and workplace settings to maximize both productivity and human capability development.

• Widespread AI adoption requires rethinking how these tools are designed to enhance rather than replace human capability

Editorial Opinion

This study raises critical questions about how we integrate AI assistants into learning and work environments. While the research shows concerning short-term cognitive effects, it's important to view these findings as a blueprint for better AI design rather than a reason to reject the technology. The challenge ahead is developing AI systems that enhance human capability—coaching rather than solving—a nuanced goal that requires collaboration between AI companies, educators, and researchers.