Exploring the Interaction of Explanation Styles, Context, and Trust of AI Privacy Redaction in AI-mediated Interactions

Explanations improve trust in AI privacy decisions: When an AI system that removes sensitive information from messages provided explanations of why it redacted content, users had significantly greater confidence that the system was effectively protecting privacy (p<0.05), suggesting that transparency about AI decision-making can build user confidence in sensitive contexts.

Context matters: More redaction requires better explanations: Users found explanations particularly valuable and helpful when the AI system removed large amounts of content rather than small amounts (p<0.05), indicating that the more aggressive an AI intervention, the more important it becomes to explain that intervention to maintain user trust.

One-size-fits-all explanations don't work: The study found that individual differences—including age and prior familiarity with AI—significantly affected how much users trusted the system, suggesting that effective AI transparency requires customizing explanations to different user groups rather than providing identical explanations to everyone.

Privacy-preserving AI requires careful communication design: The research highlights a fundamental challenge in AI-mediated communication: systems must simultaneously protect privacy through redaction while maintaining user trust through clear explanation, and getting this balance right depends on understanding both the context and the individual user.