Out-of-distribution Evidence-aware Fake News Detection via Dual Adversarial Debiasing

TL;DR

This work tackles the OOD generalization problem in evidence-aware fake news detection caused by news-content and evidence-content biases. It introduces Dual Adversarial Learning (DAL), a plug-and-play framework that adds news-aspect and evidence-aspect debiasing discriminators to the standard evidence-aware detection backbone, while training the main predictor to maximize factual reasoning between news and evidences. By reversely optimizing the debiasers and positively optimizing the main model, DAL reduces bias-driven signals and enhances news–evidence reasoning, yielding consistent improvements across four backbones and two OOD settings. The approach demonstrates strong, stable gains over existing debiasing methods, suggesting practical utility for robust misinformation detection under distribution shifts.

Abstract

Evidence-aware fake news detection aims to conduct reasoning between news and evidence, which is retrieved based on news content, to find uniformity or inconsistency. However, we find evidence-aware detection models suffer from biases, i.e., spurious correlations between news/evidence contents and true/fake news labels, and are hard to be generalized to Out-Of-Distribution (OOD) situations. To deal with this, we propose a novel Dual Adversarial Learning (DAL) approach. We incorporate news-aspect and evidence-aspect debiasing discriminators, whose targets are both true/fake news labels, in DAL. Then, DAL reversely optimizes news-aspect and evidence-aspect debiasing discriminators to mitigate the impact of news and evidence content biases. At the same time, DAL also optimizes the main fake news predictor, so that the news-evidence interaction module can be learned. This process allows us to teach evidence-aware fake news detection models to better conduct news-evidence reasoning, and minimize the impact of content biases. To be noted, our proposed DAL approach is a plug-and-play module that works well with existing backbones. We conduct comprehensive experiments under two OOD settings, and plug DAL in four evidence-aware fake news detection backbones. Results demonstrate that, DAL significantly and stably outperforms the original backbones and some competitive debiasing methods.

Figures (4)

• Figure 1: Causal diagrams of evidence-aware fake news detection.
• Figure 2: The overview of DAL with a piece of news $n_i$ and its corresponding evidences $E_i = \left\{ {e_{i,1} , e_{i,2} ,..., e_{i,\left| {E_i } \right|} } \right\}$: (a) the structure of evidence-aware fake news detection models; (b) the structure of the news-aspect debiasing; (c) the structure of the evidence-aspect debiasing.
• Figure 3: Sensitivity of hyper-parameters, i.e., $\alpha$ and $\beta$, of DAL plugged in four different evidence-aware fake news detection backbones, tested on PoliticFact and Snopes under the cross-platform setting measured by F1-Macro.
• Figure 4: Sensitivity of hyper-parameters, i.e., $\alpha$ and $\beta$, of DAL plugged in four different evidence-aware fake news detection backbones, tested on PoliticFact and Snopes under the cross-topic setting measured by F1-Macro.