Verification limits reduce confidence in ai-generated knowledge artifacts
More Claims, Less Evidence: Bounded Verification of AI-Generated Digital Knowledge Artifacts
Artificial IntelligenceDigital Libraries
Summary
As AI systems create more digital summaries and documents with many claims, it's hard for humans to check every claim for accuracy. The authors developed a model that predicts how reliable a document is if only some claims are verified. Their study shows that when documents get bigger but checking remains limited, confidence in the whole document drops, especially if errors are rare and scattered. They also provide a way to estimate how many claims need checking to keep trust high.
What this means in practice
- •For digital library managers: Determine how many claims to verify in AI-generated summaries to maintain trust in library collections.
- •For content moderation teams: Set verification budgets that balance effort and reliability for fact-checking large volumes of AI-generated content.
Authors
Feliks Bańka, Jarosław A. Chudziak
Abstract
Digital libraries, repositories, and AI-mediated knowledge services increasingly rely on generative systems to produce summaries, descriptions, and other multi-claim knowledge objects. Yet generation can scale far more easily than verification capacity: a reviewer may need to decide whether an object is suitable for publication or downstream use after checking only a small fraction of its claims. This creates a fundamental gap between claim-level verification and confidence in the object as a whole. The central question is therefore what a successful partial check implies about the reliability of the complete artifact when its size grows but the verification budget does not. This paper contributes a Bayesian model of bounded verification centered on the Predictive Value of Pass (PVP). The model predicts that evidentiary value decreases as artifacts grow under fixed verification capacity, improves with larger verification budgets, and is especially fragile when errors are sparse. Controlled experiments on FEVEROUS and FEVER support these predictions and show that adding supported claims around a fixed number of false or unsupported claims can make passing more likely while making a pass less informative. The model further yields the minimum verification budget required to maintain a target PVP, providing a practical component for AI-assisted quality-assurance workflows in which generated knowledge objects must be checked before publication or downstream use.