AI can boost your test scores, but it might mislead you
A new study from Aalto University says using AI tools like ChatGPT can help you score better on tough problems—but it also makes people think they did even better than they actually did.
In studies involving nearly 500 participants, those who used AI overestimated their results, especially if they already knew a bit about how AI works.
The Dunning-Kruger effect
Normally, people who aren't great at something tend to overestimate themselves (that's the Dunning-Kruger effect).
But here, it was the folks with higher AI know-how who got extra confident.
Researchers think this happens because people start relying on AI for thinking and just trust whatever answer it gives.
Users rarely double-checked ChatGPT's answers
Most users only asked ChatGPT once per problem and rarely double-checked its answers.
This lack of back-and-forth means less feedback and more false confidence—even though their scores were higher overall.
Researchers suggest future AI tools encourage users to reflect more
The researchers suggest that future AI tools should encourage users to reflect more on their answers, not just trust the tech blindly.
Otherwise, there's a risk of mistakes piling up as confidence outpaces actual skill.