MIT's Eric So warns heavy AI dependence can weaken thinking skills in new book

In artificial intelligence news that is drawing wide attention in the United States, MIT Sloan School of Management professor Eric So has warned that relying heavily on AI can weaken the very thinking skills people need to use it well. So, who teaches global economics and behavioural science, sets out the argument in his new book, "The Collision: What AI Does to Us", published this month by W.W. Norton. In a question-and-answer interview released by MIT on October 15, 2026, and in coverage the same day, he described the risk as a form of cognitive atrophy: people stop building, and stop keeping, core mental abilities because a machine does the hard part for them.

So's case does not rest on a single new laboratory experiment. It is a synthesis of existing research and his own observations, written by a behavioural scientist who studies how people make decisions. His central point is about effort. Learning, he argues, depends on productive struggle: wrestling with an idea, getting it wrong, and revising. AI tools remove much of that friction by producing quick, polished answers on demand. That suits a brain that naturally prefers instant reward and conserves energy, but it can cut out the work that builds lasting mastery.

He extends the concern to memory. Because many people now treat AI as their first stop for information, he says, it is changing how much they hold in their own heads. He also points to students as a group at particular risk, saying many are outsourcing their thinking to AI rather than working through problems themselves.

The idea of "AI gravity" explains why the pull is hard to resist. According to So, several forces push in the same direction: a biological instinct to save mental energy; social pressure, because when colleagues or classmates use AI everyone else feels they must keep up; and the difficulty of telling when someone has used AI at all. Together these create steady pressure to hand over more thinking in the name of efficiency. MIT Sloan has described the concept in exactly those terms, as the constant pressure to outsource more thinking to AI to become more efficient.

For India, the argument lands at a time when AI tools are spreading quickly through classrooms, coaching centres, offices and start-ups. Its relevance is practical rather than alarmist. So is not saying AI damages the brain instantly or that it should be avoided. He is saying that how it is used decides whether it strengthens or weakens a person. That is a question for students preparing for competitive exams, for teachers setting assignments, and for employers deciding how to measure the value of AI in their teams.

Short-term: The book and the interview add a well-known academic voice to a growing debate on whether AI assistants make people sharper or lazier. Schools, colleges and companies reviewing their AI rules are likely to find his framing useful when deciding where to allow or limit the tools.

Long-term: If his warning holds, the value of human skills such as sustained attention, judgement and the patience to solve hard problems could rise, because they become rarer. Education systems may need to design assignments that reward the process of thinking, not just the final answer, so that students still practise the struggle that builds expertise.

Who is affected: Students and young professionals who use AI chatbots daily, teachers and examiners, managers who track AI adoption in their teams, and anyone who uses AI as their main source of information.

Questions and Answers