Can AI Make Us Smarter? Research Points to the Value of Intellectual Friction

Summary

This article examines whether generative AI improves human thinking or encourages people to outsource it. It describes cognitive offloading as the tendency to accept fast, confident AI responses without sufficient questioning, potentially weakening critical-thinking skills even as productivity and time savings increase. A study led by Monideepa Tarafdar and colleagues, involving 45 knowledge workers at a U.S. public university, points to a different approach: use friction-generating queries that ask AI to contradict a user’s view, explore unlikely perspectives, or act as a devil’s advocate. The article says workers using AI in this way improved their performance, while earlier research on human-AI loan evaluations similarly found better decisions when people challenged initial AI recommendations. Examples include a marketing professional using negative AI-generated customer personas to uncover overlooked preferences, an attorney asking AI to explore legal loopholes and unethical scenarios, and a supply-chain researcher adapting ideas from signal processing and wireless communication. Other cited work suggests creative writers benefit more from using AI as a sounding board than as a ghostwriter. The article emphasizes that AI supplies possibilities, while domain expertise, verification, ethics, and final judgment remain human responsibilities. It recommends that organizations teach workers to scrutinize outputs, share practices through AI groups, provide bootcamps and workshops, and offer private sandboxes for experimentation. The central argument is that AI is most beneficial when treated as an intellectual sparring partner rather than an oracle or substitute for thinking, although the article also warns that factual correctness does not necessarily produce insight.