We use a controlled experiment to show that ability and measurement beliefs jointly determine the benefits of working with Artificial Intelligence (AI). AI improves performance significantly for people with low basic skills. However, to hold the ability constant, AI help is more important for people who are averaged, meaning they have accurate beliefs about their ability. People who know they have low skills benefit the most from working with AI. In a counterfactual analysis, we show that removing anomalies can cause AI to reduce performance inequality by almost twice as much as it already does.
That’s according to a new NBER working paper on
Source link