The smart assistant solves the problem in seconds and everything seems fine. But what happens when it suddenly disappears?
Researchers from Carnegie Mellon, Berkeley, Oxford and MIT conducted three experiments on 1,222 people and found that about ten minutes of relying on a chatbot is enough to weaken a person''s desire to try alone.
In the first experiment, 354 participants solved fraction problems, half of them with the help of artificial intelligence. They initially performed better, but then the researchers removed the AI without warning. The result was then reversed: their performance declined, and they skipped problems more often than their peers who hadn''t used it at all.
The result was replicated with 667 other participants, and then with 201 people in a reading comprehension test. The effect was strongest among those who requested the answer directly rather than a hint.
Lead researcher Grace Liu, from Carnegie Mellon University, says that "useful difficulty" is what builds skill over time. Those who always get the answer may win the moment but lose their independent ability.
The team suggests that artificial intelligence should act as a guide pointing the way, not as a machine delivering the solution.