Generative AI may erode human persistence and critical thinking, studies warn

The Dawn/AFP piece surfaces a tension that’s been simmering under the productivity narrative: if generative AI makes it too easy to offload effort, we may be trading short‑term gains for long‑term cognitive atrophy. The studies cited suggest that when people use AI to solve problems, they often learn less, persevere less and are less willing to try without the tool later. That’s not surprising for anyone who has watched calculators reshape maths education, but the scope is different when the tool can help with almost any cognitive task.([dawn.com](https://www.dawn.com/news/2015756))

From an AGI race perspective, the risk isn’t just that individual students get lazier. It’s that entire organisations re‑optimise around “ask the model” rather than cultivating deep domain expertise. If frontier models keep improving while human skill investment shrinks, you get a feed‑forward loop where ever more decisions are made by systems people barely understand. That makes society more brittle when models fail and increases the temptation to hand over genuinely strategic choices to AI.

That said, the same research also nudges labs toward building more “Socratic” modes that prompt users to think rather than just deliver answers. If AGI is going to be integrated into education and work at scale, making it a good teacher – one that preserves persistence and curiosity – may be as important as raw capability.