Mediating Community-AI Interaction through Situated Explanation: The Case of AI-Led Moderation

Community context matters more than technical accuracy: The research found that how communities understand AI moderation decisions depends heavily on their shared values, norms, and practices—not just on how technically correct the AI's explanation is. This suggests that one-size-fits-all explanations from AI developers may fail to resonate with actual users.

Collective explanation is a form of community resilience: Rather than passively accepting AI decisions, community members actively work together to develop their own explanations for why the AI took certain actions (particularly automated punishments). This collaborative sense-making helps communities maintain trust and agency when interacting with AI systems.

Individual-focused XAI research misses the bigger picture: Most explainable AI research focuses on helping individual users understand AI decisions in controlled settings. This study reveals that real-world AI impact—like content moderation—happens at the community level, requiring different approaches to explanation and accountability.

Automated punishment decisions need special attention: The case study of AI-led moderation highlighted that communities struggle most with understanding and accepting automated punishments. This suggests that high-stakes AI decisions affecting community members warrant more transparent and community-informed explanation strategies.

Explanation is a social process, not just a technical one: The research demonstrates that explanation mediates the relationship between communities and AI systems. Rather than treating explanation as something AI systems simply "provide," organizations should recognize it as an ongoing dialogue shaped by community context and values.