Beyond Surface Judgments: Human-Grounded Risk Evaluation of LLM-Generated Disinformation

TL;DR

This study evaluates how well LLM-based judges reflect human reader responses in assessing disinformation, revealing significant gaps and differences in evaluation criteria and alignment.

Contribution

The paper systematically audits LLM judges against human responses, highlighting their internal consistency but poor alignment with actual reader perceptions.

Findings

LLM judges are generally harsher than humans.

Judges weakly recover human item rankings.

Judges rely more on logical rigor and less on emotional cues.

Abstract

Large language models (LLMs) can generate persuasive narratives at scale, raising concerns about their potential use in disinformation campaigns. Assessing this risk ultimately requires understanding how readers receive such content. In practice, however, LLM judges are increasingly used as a low-cost substitute for direct human evaluation, even though whether they faithfully track reader responses remains unclear. We recast evaluation in this setting as a proxy-validity problem and audit LLM judges against human reader responses. Using 290 aligned articles, 2,043 paired human ratings, and outputs from eight frontier judges, we examine judge--human alignment in terms of overall scoring, item-level ordering, and signal dependence. We find persistent judge--human gaps throughout. Relative to humans, judges are typically harsher, recover item-level human rankings only weakly, and rely on different textual signals, placing more weight on logical rigour while penalizing emotional intensity more strongly. At the same time, judges agree far more with one another than with human readers. These results suggest that LLM judges form a coherent evaluative group that is much more aligned internally than it is with human readers, indicating that internal agreement is not evidence of validity as a proxy for reader response.