In plain words: Fake news detectors were tested on human-written and AI-written stories, and many judged writing style instead of truth: they called AI text fake while passing human-written lies as real. Retraining them on AI-rewritten real news improved accuracy on both kinds.
Abstract · Fake News Detectors are Biased against Texts Generated by Large Language Models
The spread of fake news has emerged as a critical challenge, undermining trust and posing threats to society. In the era of Large Language Models (LLMs), the capability to generate believable fake content has intensified these concerns. In this study, we present a novel paradigm to evaluate fake news detectors in scenarios involving both human-written and LLM-generated misinformation. Intriguingly, our findings reveal a significant bias in many existing detectors: they are more prone to flagging LLM-generated content as fake news while often misclassifying human-written fake news as genuine. This unexpected bias appears to arise from distinct linguistic patterns inherent to LLM outputs. To address this, we introduce a mitigation strategy that leverages adversarial training with LLM-paraphrased genuine news. The resulting model yielded marked improvements in detection accuracy for both human and LLM-generated news. To further catalyze research in this domain, we release two comprehensive datasets, \texttt{GossipCop++} and \texttt{PolitiFact++}, thus amalgamating human-validated articles with LLM-generated fake and real news.
Jinyan Su, Terry Yue Zhuo, Jonibek Mansurov, Di Wang, Preslav Nakov
arXiv:2309.08674 · cs.CL, cs.AI · submitted Sep 15, 2023
abstract · pdf · html · The first two authors contributed equally
I think this is more interesting:
> [LLMs] are more prone to flagging LLM-generated content as fake news while often misclassifying human-written fake news as genuine.
> To address this, we introduce a mitigation strategy that leverages adversarial training with LLM-paraphrased genuine news. The resulting model yielded marked improvements in detection accuracy for both human and LLM-generated news.