Legible Machines, Contestable Truths

Abstract

The rise of AI-based fact-checking systems has intensified epistemic debates over transparency, legitimacy, and public accountability in algorithmic truth production. While explainable AI has aimed to make automated verification interpretable and trustworthy, the next frontier lies in contestability: designing systems that invite challenge, dialogue, and revision rather than passive understanding. Contestable AI fact-checking reframes explainability as a civic and procedural practice in which users become active interlocutors who can interrogate, dispute, and reshape machine-generated claims. Contestability functions not as a technical supplement but as an ethical and institutional condition for democratic epistemology, enabling deliberative feedback, evidence rebuttal, and iterative correction. By extending explainability into the domain of public reasoning, contestable fact-checking redefines the role of AI from an authoritative verifier to a participant in epistemic negotiation. The analysis advances the idea that legitimacy in AI-mediated verification depends not only on interpretability but on the capacity for contestation, where accountability is realized through interaction and argument rather than disclosure alone. This chapter proposes a dual framework of explainability and contestability as interdependent conditions for epistemic legitimacy in AI-driven verification systems. It calls for value-sensitive and reflexive designs that integrate transparency with procedural fairness, ensuring that truth remains open to scrutiny, disagreement, and revision within algorithmic environments.