It is interesting to see that the drop in performance was a function of the users' perception of the AI (i.e. how good they thought the AI was) rather than the performance of the AI itself. I wonder - What would the results would look like if the users were asked to work with the AI without informing them about the efficiency of the AI? How long it would take for the same pattern to show up?
To be honest, I don't agree with the conclutions. We don't have to change approach of designing AI - we have to educate people how to use it proreply. But overall study is interesting, thank you for covering. So, recruiters were real, right?
Yes, real recruiters, hired at market rates. Agree with idea of being trained in how to use AI. A few videos ago (ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-ogQbgdZQiaI.html), the paper I looked at got into the idea of "centaurs" vs "cyborgs" as an analogy for how to use AI.