Evaluating the Influences of Explanation Style on Human-AI Reliance

Different explanation styles affect trust differently: The study found that feature-based explanations (showing which factors influenced a decision) and example-based explanations (showing similar past cases) produced different patterns of human reliance on AI systems, even when both were equally interpretable—suggesting that how explanations are presented matters beyond just clarity.

One-size-fits-all explanations don't work: The research demonstrates that the effectiveness of XAI methods depends on individual user characteristics and task complexity, indicating that organizations should tailor explanations to specific users and situations rather than assuming a single explanation style will work for everyone.

Interpretability alone isn't enough: While explainable AI is designed to increase interpretability, this study reveals that interpretability is only part of what influences whether people rely on AI appropriately—other qualities of explanation styles play important roles that current XAI approaches may overlook.

Context matters for appropriate reliance: The findings suggest that the same explanation style can lead to either appropriate or inappropriate reliance depending on factors like task difficulty and individual performance, highlighting the need to consider the full context when designing AI explanations.