Skip to main content


Using #AI makes people less likely to be correct

Accuracy collapsed when AI was available

- Baseline: 27% gave correct answer
- With AI advice, 9% gave #correct answer

Fact: Some would-be correct people got AI advice & became #wrong

AI advice made people #confident there're correct

- Baseline: 30%
- With AI help, confidence rose to 76%

People became worse, accuracy was only 1/3, but they were twice as confident

theregister.com/ai-and-ml/2026…

#digital #DumbedDown #DDD = Digitally Dumbed Down

in reply to Mark

this is a bad study. The biggest single flaw is that the authors designed the study so the AI was guaranteed to be wrong, then drew conclusions about what happens when AI is present. Those are different things. I'm all for studying how AI affects people but this one is not well done.
in reply to jcrabapple

@jcrabapple I do think that still says something valuable though? I would say it illustrates that if they were made reliably wrong in a way that is useful to someone, people as they currently are, would seem to believe what it says. Sounds like all it would take is injecting pre-baked responses into specific outputs to make that happen.