aiqbench analysis
This study tests how well people can tell GPT-4 and Llama-2 text from human writing, using 918 evaluations and causal models. Its headline findings are that political orientation barely predicts detection accuracy, while prior exposure to fake news does, and that GPT-4 output can pass as human because of its fluency. For readers tracking AI risk, the results argue for targeting how people judge sources rather than demographic groups, though the causal claims rest on observational correlations and should be read as hypotheses.
Eroding the Truth-Default: A Causal Analysis of Human Susceptibility to Foundation Model Hallucinations and Disinformation in the Wild
arXiv:2601.22871v3 Announce Type: replace-cross Abstract: As foundation models (FMs) approach human-level fluency, distinguishing synthetic from organic content has become a key challenge for Trustworthy Web Intelligence. This paper presents JudgeGPT and RogueGPT, a dual-axis framework that decouples "authenticity" from "attribution" to investigate the mechanisms of human susceptibility. Analyzing 918 evaluations across five FMs (including GPT-4 and Llama-2), we employ Structural Causal Models (SCMs) as a principal framework for formulating testable causal hypotheses about detection accuracy. Contrary to partisan narratives, we find that political orientation shows a negligible association with detection performance ($r=-0.10$). Instead, "fake news familiarity" emerges as a candidate mediator ($r=0.35$), suggesting that exposure may function as adversarial training for human discriminators. We identify a "fluency trap" where GPT-4 outputs (HumanMachineScore: 0.20) bypass Source Monitoring mechanisms, rendering them indistinguishable from human text. These findings suggest that "pre-bunking" interventions should target cognitive source monitoring rather than demographic segmentation to ensure trustworthy information ecosystems.
Abstract:As foundation models (FMs) approach human-level fluency, distinguishing synthetic from organic content has become a key challenge for Trustworthy Web Intelligence. This paper presents JudgeGPT and RogueGPT, a dual-axis framework that decouples "authenticity" from "attribution" to investigate the mechanisms of human susceptibility. Analyzing 918 evaluations across five FMs (including GPT-4 and Llama-2), we employ Structural Causal Models (SCMs) as a principal framework for formulating testable causal hypotheses about detection accuracy. Contrary to partisan narratives, we find that political orientation shows a negligible association with detection performance ($r=-0.10$). Instead, "fake news familiarity" emerges as a candidate mediator ($r=0.35$), suggesting that exposure may function as adversarial training for human discriminators. We identify a "fluency trap" where GPT-4 outputs (HumanMachineScore: 0.20) bypass Source Monitoring mechanisms, rendering them indistinguishable from human text. These findings suggest that "pre-bunking" interventions should target cognitive source monitoring rather than demographic segmentation to ensure trustworthy information ecosystems.
8 pages, 4 figures, plus supplementary appendix. Published in Companion Proceedings of the ACM Web Conference 2026 (WWW Companion '26), pp. 1125-1132
Computers and Society (cs.CY); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Human-Computer Interaction (cs.HC)
I.2.7; H.1.2; H.5.2
arXiv:2601.22871 [cs.CY]
(or arXiv:2601.22871v3 [cs.CY] for this version)
https://doi.org/10.48550/arXiv.2601.22871 Focus to learn more arXiv-issued DOI via DataCite
Companion Proceedings of the ACM Web Conference 2026 (WWW Companion '26), pp. 1125-1132
https://doi.org/10.1145/3774905.3795832 Focus to learn more DOI(s) linking to related resources
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 299 words
Related coverage
An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency
arXiv · Sep 21, 2026
MotionHalluc: Diagnosing Kinematic Hallucinations in Fine-Grained Motion Reasoning
arXiv · Oct 23, 2026
Mitigating Cross-Modal Hallucinations by Correcting Modality Misrouting in Omnimodal LLMs
arXiv · Oct 15, 2026
Eliciting Intrinsic Hallucinations in LLMs via Semantically Equivalent Adversarial Attacks
arXiv · Aug 6, 2026