Explainable AI and Adoption of Financial Algorithmic Advisors: an Experimental Study

Explanations matter most when things go wrong: The study found that detailed explanations (whether feature-based or accuracy-based) significantly reduced users' abandonment of an AI advisor after it made mistakes, suggesting that transparency is particularly valuable for maintaining trust during system failures.

Different explanations work better at different times: Accuracy-based explanations led to higher adoption rates in early interactions, but when the AI system performed perfectly, the type of explanation provided became less important—users were willing to adopt regardless.

People will pay more for explained AI advice: Participants were willing to pay a higher price for financial advice when it came with explanations compared to identical AI advice without explanations, indicating that explainability adds perceived value to algorithmic services.

"Autopilot" features boost adoption significantly: Enabling automatic acceptance of recommendations substantially increased adoption rates, suggesting that convenience and reduced friction can be as influential as transparency in driving user acceptance of AI systems.

Explainability is critical for responsible AI deployment: The research demonstrates that how AI systems communicate their reasoning directly influences whether people trust and adopt them, highlighting that responsible AI requires not just accurate algorithms but also thoughtful explanation design.