# Zoom Out and Observe: News Environment Perception for Fake News Detection

**Authors:** Qiang Sheng, Juan Cao, Xueyao Zhang, Rundong Li, Danding Wang, Yongchun Zhu  
**arXiv:** [2203.10885](https://arxiv.org/abs/2203.10885)  
**Date:** 2022-03-21  
**Summary page:** https://www.summarizepaper.com/en/paper/2203.10885v2/

## Key points

- Fake news detection is important in preventing the spread of misinformation on social media
- Existing methods focus on analyzing language patterns and verifying content, but overlook the influence of the external news environment
- The news environment can inspire fake news by capitalizing on popular events and capturing public interest
- The proposed NEP framework "zooms out" to observe the broader news environment surrounding each post
- NEP constructs macro and micro signals from recent mainstream news articles for each post
- Two modules focused on popularity and novelty perceive useful signals from the news environment
- Incorporating NEP significantly improves the performance of basic fake news detectors
- Fake news detection benefits more from incorporating signals from the news environment compared to real news detection
- NEP captures information from the broader news environment, complementing verified knowledge in detecting fake news.

## Summary

, , , , 


Fake News Detection: Enhancing Performance by Perceiving Signals from the News Environment

 is a critical task in preventing the spread of misinformation on social media. Existing methods primarily focus on analyzing the language patterns of news posts and verifying their content through knowledge sources or user replies. However, these approaches often overlook the influence of the external news environment in which fake news is created and disseminated. The  represents the prevailing opinions in mainstream media and public attention, which can inspire the fabrication of fake news by capitalizing on popular events and capturing public interest with unexpected content. To address this limitation, we propose a  that "zooms out" to observe the broader news environment surrounding each post. We construct both macro and micro  from recent mainstream news articles for each post. Additionally, we design two modules - one focused on popularity and another on novelty - to perceive useful signals from the  and enhance the final prediction. Our experiments using newly built datasets demonstrate that incorporating NEP significantly improves the performance of basic fake news detectors. Specifically, all six base models show enhanced accuracy and macro F1 scores when combined with NEP. Furthermore, our results indicate that fake news detection benefits more from incorporating signals from the  compared to real news detection. We also compare our approach with existing methods that rely solely on post content or related post-level signals like social context and knowledge. While knowledge-based methods show some improvement by leveraging relevant articles, our NEP provides additional gains by capturing information from the broader . This suggests that perceived signals from the  play a complementary role to verified knowledge in detecting fake news. In conclusion, our NEP framework offers a novel perspective by considering the influence of the external  in detecting fake news on social media platforms. By zooming out and observing this environment, we can capture valuable signals that enhance the performance of basic fake news detectors.