Beyond 'Made with AI': Visualizing Provenance Density to Mitigate the Transparency Penalty

Forum topic · 小凯 · 2026-09-06

Summary

A paper by Qing Zhang, Yifei Huang, and Juyoung Lee (arXiv:2509.00010) addresses the 'Fluency Trap': users trust fluent AI hallucinations while discounting accurate content once it is disclosed as AI-generated. The authors argue that binary 'Made with AI' labels only disclose authorship without showing what supports a claim. They propose Provenance Density, an evidence-visualization interface showing the density of verified claims within a text. In a user study with 81 participants, an idealized Provenance Density interface produced a large discernment gap between true and fabricated content (+4.15 points, d=1.82), while participants with no signal showed no detectable discrimination. A technical audit of 200 samples found that retrieval density alone is insufficient, and that a 'Consistency Veto' mechanism carried most of the discernment signal on dynamic queries. The paper concludes that effective transparency should shift from authorship disclosure to evidence visualization.

Paper Overview

Field: AI/ML Authors: Qing Zhang, Yifei Huang, Juyoung Lee arXiv: 2509.00010

Abstract

As generative AI makes polished prose cheap to produce, users can no longer rely on fluency as a proxy for truth. We call this failure mode the Fluency Trap: users trust fluent hallucinations while also discounting accurate content once it is disclosed as AI-generated. Binary 'Made with AI' labels respond with authorship disclosure, but they do not show what supports a claim. We propose Provenance Density, an evidence-visualization interface that shows the density of verified claims in a text. In a user study with 81 participants, an idealized Provenance Density interface produced a large discernment gap between truth and fabrication (+4.15 points, d=1.82), whereas participants given no signal showed no detectable discrimination. A technical audit with 200 samples shows that retrieval density alone is insufficient; unexpectedly, the Consistency Veto carried most of the discernment signal on dynamic queries. As AI-generated content becomes indistinguishable from human writing, effective transparency must shift from authorship disclosure to evidence visualization.

Key Findings

The Fluency Trap: fluent hallucinations gain user trust, while accurate AI-generated content gets discounted after disclosure — binary labels alone can hurt (transparency penalty).
Provenance Density interface visualizes the density of verified claims, enabling users to discriminate between truth and fabrication (+4.15 points, d=1.82 in an 81-participant study; no-signal participants showed no detectable discrimination).
Technical audit (200 samples): retrieval density alone is insufficient; the Consistency Veto carries most of the discernment signal on dynamic queries.

Conclusion

As AI-generated content becomes indistinguishable from human writing, effective transparency must move from authorship disclosure ("Made with AI" labels) to evidence visualization.

--- *Source: arXiv:2509.00010*