From AI Fact-Checks to User Understanding: Explaining Misinformation Detection to Non-Expert Audiences

Misinformation presents a critical threat to society, impacting public health, democratic processes, and social cohesion.As AI-driven tools are increasingly deployed to detect false or misleading online content, effectively communicating these detection results to nonexpert users is crucial.This paper introduces guidelines for designing explanations of AI-detected misinformation, addressing both the technical challenges of AI detection and the cognitive biases that shape how people interpret flagged content.The discussion covers key aspects of explanation design, including different explanation types (feature-based, free-text, evidence-based, uncertainty-aware, and multimodal), timing strategies (pre-exposure inoculation, interventions during exposure, and post-exposure corrections), and user-specific approaches (belief-consonant framing, progressive disclosure, and personalized explanations).Drawing from AI, psychology, and communication research, the guidelines aim to tackle the "black box" problem-where opaque AI decision-making undermines user trust and comprehension.The paper also highlights factors that influence explanation effectiveness, such as prior attitudes, source credibility, and information coherence.These insights support the proposed guidelines, which are designed to improve user understanding and encourage informed engagement with online content.