Influence of prior and task generated emotions on XAI explanation retention and understanding

Emotions shape how people interpret AI explanations, not just whether they remember them. The study found that while prior emotions (like being induced to feel happy or fearful before seeing an explanation) didn't significantly affect what people remembered about AI features, they did influence how people understood the relevance of those features—suggesting emotions can create confirmation bias when people evaluate AI reasoning.

AI explanations need to account for users' emotional states. The research measured heart rate variability, facial expressions, and self-reported emotions while people received explanations, revealing that emotional arousal during the explanation process is a measurable factor that responsible AI systems should consider when designing how to communicate their decisions.

Personal attitudes trigger emotional reactions to specific AI features. The study identified that certain features related to users' personal attitudes generated measurable emotional arousal in individual participants, indicating that the same AI explanation may provoke different emotional responses depending on what values or beliefs the explanation touches on.

Current AI explanation research overlooks social and emotional factors. The authors highlight a "surprising lack of knowledge" about how emotions affect explanation comprehension in AI systems, suggesting that the field of explainable AI (XAI) has focused too narrowly on technical clarity without considering the human psychological factors that determine whether explanations actually help users understand AI decisions.