AI Should Serve as Trainer, Not Crutch, Researchers Advise

Recent research indicates that artificial intelligence is most beneficial when it acts as a trainer, offering feedback and examples, rather than providing direct answers. The study, led by behavioral scientist Trent Cash of the University of Waterloo and published in *Trends in Cognitive Sciences* on 9 July, found that skills such as solving math problems or detecting polyps during colonoscopy deteriorated when participants relied entirely on AI tools. After a period of AI‑assisted work, performance fell once the assistance was removed.

The authors recommend using AI to guide learning while keeping the learner mentally active, a approach that can preserve or even improve abilities. The findings echo earlier commentary in Croatian media warning that unchecked AI use may weaken cognitive capacities, but they also stress that balanced, supportive AI applications can enhance education and professional training.