AI Use May Reduce Thinking, but Lasting Decline Not Proven
Recent studies suggest AI tools may reduce critical thinking and brain engagement, but experts say this does not prove permanent cognitive decline. The risk lies in over-reliance, not the technology itself, depending on how users apply it.
Will artificial intelligence make people less intelligent? Two recent studies suggest a possible link, but the evidence does not prove lasting decline, according to a new commentary.
An MIT Media Lab experiment, not yet peer-reviewed, found lower brain connectivity among participants who wrote essays with ChatGPT compared with those who wrote unaided. Separately, a Microsoft and Carnegie Mellon survey of 319 knowledge workers found that greater confidence in AI for a given task was associated with less self-reported critical thinking.
Both findings warrant attention, but neither establishes that AI causes a permanent reduction in intelligence. Thinking less while the machine works is not the same as becoming less able to think, the authors argue.
Drawing an analogy, the commentary compares AI to a cart carrying heavy bags: muscles do not weaken that afternoon because the load has moved. However, if one stops lifting for years, strength will fade. The MIT study hints at this risk, as participants who switched from ChatGPT to working unaided stayed less engaged, but the sample was small and the period short.
The purpose of the task matters, the authors write. When learning is the goal, as with a student's essay, delegating the thinking may defeat it. When delivering a result is the aim, reducing effort can be sensible. The challenge is knowing which kind of work is being done.
Two professionals using the same AI on the same analysis might diverge: one accepts the result and moves on, while the other uses it to test more alternatives, challenge assumptions, and tackle analysis they could not have managed alone. The first saves effort; the second reinvests it in becoming capable of more.
The commentary concludes that AI will not make people dumber by itself. The bigger danger is mistaking what the machine produced for what users have become. The cart carries the load, but it does not decide where people walk.
The piece was co-authored by Frank Ng, a retired NASDAQ CEO, and his son Ryan, following their book Hey AI, Let’s Talk!