Study Finds AI Users Perform Worse Than Non-Users When Assistance Is Removed

Three randomized experiments involving 1,222 participants investigated the impact of removing AI assistance from work tasks. While AI initially improved performance, participants performed worse than those who had worked without AI once the assistant was removed.

The study highlights a potential dependency effect, where reliance on AI tools may degrade baseline problem-solving abilities when the tools are unavailable. However, the experiments only measured short-term effects, leaving the long-term implications for professional workflows open to question.