research reveals: just 10 minutes of interacting with an ai assistant can significantly undermine human independent problem-solving abilities
recent experiments show that even brief reliance on artificial intelligence assistants (such as gpt-5) rapidly erodes individuals’ cognitive autonomy. after only 10 minutes of interacting with ai, participants exhibited a noticeable decline in performance on cognitive tasks like arithmetic and reading comprehension.
in the experiment, some participants completed tasks with ai assistance before the system abruptly cut off service; at that point, their performance plummeted compared to the control group, which had never used ai throughout: accuracy dropped sharply from 73% to 57%, the frequency of skipping questions increased markedly, and their willingness to actively attempt solutions diminished significantly.
notably, this degradation effect was most pronounced among users who were accustomed to directly requesting answers rather than merely seeking hints. subsequent rounds of experiments further confirmed the universality of this phenomenon—over-reliance on ai led to a rapid decline in independent thinking across various types of reading-comprehension tasks.
the researchers emphasize that the risk does not stem from ai technology itself, but from the way humans collaborate with machines. when individuals continuously cede judgment and decision-making authority to ai models, the brain’s cognitive engagement diminishes, gradually fostering mental inertia and ultimately undermining the reasoning, analysis, and problem-solving skills that should remain under human control.
