Cognitive Epistemic Modeling in AI Fact-Checking

Abstract

As algorithmic verification systems proliferate, a central challenge is ensuring that they align with how people actually interpret, trust, and revise information. Existing fact-checking models often privilege technical accuracy while neglecting the cognitive and epistemic processes that shape user engagement. This chapter introduces Cognitive-Epistemic Modeling (CEM) as a framework for designing AI fact-checking systems that align with human reasoning and societal norms. CEM proposes a four-layer architecture (epistemic annotation, dual-objective learning, trust profiling, and epistemically sensitive interfaces) each aimed at closing the gap between algorithmic outputs and human uptake. Central to the model is epistemic calibration, the alignment of user trust with a system’s reliability and intelligibility, which avoids both blind acceptance and systemic skepticism. A case study on vaccine misinformation illustrates how CEM strengthens persuasive capacity by combining factual accuracy with epistemic resonance.