MIT professor: Excessive AI use is eroding cognitive muscles

Summary

MIT Sloan professor Eric So argues in an interview promoting his new book, 'The Collision: What AI Does to Us,' that over-reliance on AI tools leads to 'cognitive atrophy' by eliminating the productive struggle necessary to develop thinking skills. So describes the systemic pressures that reinforce this dependency as 'AI gravity.'

Why it matters

As AI becomes integrated into workplaces, classrooms, and daily life, the question of whether it erodes fundamental reasoning skills is no longer abstract. So's framework, which links individual habits, corporate incentives, and the flattering design of chatbots, offers a useful perspective for educators and policymakers who must decide how much scaffolding to leave in place without damaging the mastery process.

Highlights

• So uses the term 'AI gravity' to describe the compound social and institutional pressures (bonuses, promotions, peer competition) that drive people to use AI, even when they are aware of the long-term cognitive costs.
• So cites an Anthropic study as evidence of outsourced thinking, showing that approximately half of university students' queries are low-effort requests for simple answers.
• So draws a parallel to the post-industrial era: just as physical labor was replaced and exercise became intentional, 'cognitive challenge' may need to be practiced consciously rather than being the default.

Critical take

The text is fundamentally a book promotion interview with a weak evidence base: it relies on an Anthropic query analysis, an anecdote about completing a sentence with ChatGPT, and a broad analogy to industrialization. However, phrases like 'cognitive atrophy' and 'we are changing our identity as thinkers' read like settled science. Counter-evidence, controlled studies, or groups that might benefit from AI-assisted scaffolding (e.g., disabled learners) are not discussed. This weakens an argument that is timely and important.