AI could make scientists do more work less well, not less work better, study argues
decoder ·
A theoretical study argues that even perfect language models could make scientific research worse by increasing opportunity costs. Researchers modeled LLMs as time-saving tools using optimal foraging theory. They found that when AI speeds up idea evaluation or publishing, researchers become less thorough because time is better spent on new projects. Only when AI accelerates the voluntary deep-dive phase does research quality improve. The authors conclude that LLMs impel scientists to do more, less well, not the same amount better. They cite real-world signs: faster software rewrites but new bottlenecks, and developers feeling faster while actually taking longer. They also note rising submissions straining peer review.
decoder