Designing Effective AI Explanations for Misinformation Detection: A Comparative Study of Content, Social, and Combined Explanations

AI explanations work better when they align: The study found that when content explanations (about linguistic features) and social explanations (about broader context) agree with each other, users are significantly more effective at detecting misinformation. This suggests that responsible AI systems should ensure different types of evidence point in the same direction.

Context matters more than previously thought: Traditional AI explanations focus mainly on analyzing the text itself, but this research shows that social context—such as information about who is sharing content and how it's being spread—is equally important for helping people identify false information.

The order of information affects how people judge misinformation: Whether users see content analysis or social context first influences their ability to detect misinformation differently depending on the topic (COVID-19 versus politics). This finding suggests that AI system designers need to carefully consider how they sequence information for different domains.

One-size-fits-all explanations are insufficient: The research demonstrates that effective misinformation detection requires multiple types of explanations working together, not just linguistic analysis alone. This has implications for how AI systems should be designed to support human decision-making in real-world applications.

User experience depends on explanation design choices: The study shows that thoughtful design of AI explanations—including what information is presented and in what order—significantly impacts both users' ability to detect misinformation and their overall experience with the system.