A peer-reviewed study co-authored by Brian Christian of UC Berkeley’s Center for Human-Compatible AI found that brief reliance on AI tools significantly impairs human focus and persistence on challenging tasks. The research, presented at the Conference on Language Modeling, involved 1,222 online participants across three randomized controlled trials. In the first experiment with 354 participants, those granted access to ChatGPT while solving 15 fraction problems initially outperformed the control group but performed markedly worse once the AI was removed after the 12th problem. A replication with 667 participants produced similar results, showing that AI-assisted users either gave up or answered incorrectly when support was withdrawn, while the non-AI group maintained higher accuracy and engagement. A third trial using 201 participants and an SAT reading comprehension task confirmed these findings. The study highlights the erosion of “productive struggle,” a critical component of cognitive development and skill mastery, when AI provides instant answers. Researchers from UC Berkeley, Carnegie Mellon University, MIT, the University of Oxford, and UCLA contributed to the work, which underscores concerns about the long-term impact of AI on scholarly rigor and collaborative problem-solving in both education and scientific research.