Does Explainable Artificial Intelligence Improve Human Decision-Making?

AI accuracy matters more than explanations: The study found that "the strongest predictor for human decision accuracy was AI accuracy," suggesting that whether an AI system works well is more important than whether users understand why it makes decisions. This challenges the assumption that explanations automatically improve outcomes.

Explanations don't consistently help users catch AI errors: While users were "somewhat able to detect when the AI was correct versus incorrect," this ability "was not significantly affected by including an explanation." This means that providing the "why" behind AI predictions didn't meaningfully help people identify when the AI was wrong.

Any AI assistance improves decisions, but explanations add little value: The research found that "providing any kind of AI prediction tends to improve user decision accuracy, but [there is] no conclusive evidence that explainable AI has a meaningful impact." Users benefited from AI help itself, not specifically from understanding how it worked.

Current explanation methods may need redesign: The authors conclude that "the 'why' information provided in explainable AI may not enhance user decision-making," suggesting that how explanations are currently designed and presented to users may not be effective and requires rethinking.

More research needed on real-world implementation: The study calls for "further research...to understand how to integrate explainable AI into real systems," indicating that laboratory findings about explainability may not translate directly to practical applications.