Examining the Effect of Explanations of AI Privacy Redaction in AI-mediated Interactions

Explanations increase trust in privacy protection: When an AI system explained why it redacted sensitive information from messages, users were significantly more likely to believe the system was effectively protecting their privacy (p<0.05), suggesting that transparency about redaction decisions matters for user confidence.

More explanation is needed when more content is hidden: Users found explanations particularly helpful and relied on them more heavily when the AI system performed extensive redactions rather than minimal ones (p<0.05), indicating that the amount of explanation should scale with the scope of intervention.

One-size-fits-all explanations don't work: Individual differences like age and users' baseline familiarity with AI affected how much they trusted the system, suggesting that effective AI privacy interventions may need to be tailored to different user groups rather than applying uniform explanation strategies.

Context shapes how users perceive AI mediation: The study demonstrates that user trust in AI-mediated communication isn't determined by the system alone—factors like how much content is being removed and the specific context of the interaction significantly influence whether users accept the system's privacy decisions.

Transparency about AI decision-making is critical for responsible deployment: This research provides empirical evidence that explaining why an AI system takes privacy-protective actions—not just that it does—is essential for building appropriate user trust in AI systems that make consequential decisions on users' behalf.