Auditing the Truth: A Pluralistic Framework for Disinformation Analysis

Abstract

Disinformation costs the global economy an estimated $78 billion a year, fueling a frantic race to build AI fact-checkers. Yet, this arms race is creating a dangerous new problem: an unaccountable 'black box of truth' that delivers an authoritative answer without showing its work, further eroding public trust. The world is trying to build an AI referee to make the final call. This paper presents a radical alternative: instead of an AI referee, we need an AI auditor. This paper details the blueprint for a Pluralistic Framework that achieves this by integrating a community-driven Endorsement model with a Comprehensive Truth Verification engine powered by Dempster-Shafer theory. This approach synthesizes conflicting information from experts, officials, and the public to produce not an answer, but a transparent audit that makes the degree of consensus and conflict easy for anyone to understand. This is not just a better fact-checker; it is a framework for turning fact-checking from a private judgment into a public audit, a vital tool for rebuilding trust in our shared reality.