[Paper Review] Using co-sharing to identify use of mainstream news for promoting potentially misleading narratives

Pranav Goel, Jon Green+2 more|arXiv (Cornell University)|Aug 12, 2023

Misinformation and Its Impacts5 citations

TL;DR

This paper identifies that mainstream news articles from reputable outlets are strategically repurposed to amplify misleading narratives by being co-shared with fake news on Twitter. Using NLP to detect misinformation narratives, the authors show that co-shared articles contain these narratives at significantly higher rates than non-co-shared articles, even after accounting for partisanship, demonstrating how factual reporting can unintentionally bolster disinformation campaigns.

ABSTRACT

Much of the research quantifying volume and spread of online misinformation measures the construct at the source level, identifying a set of specific unreliable domains that account for a relatively small share of news consumption. This source-level dichotomy obscures the potential for users to repurpose factually true information from reliable sources to advance misleading narratives. We demonstrate this potentially far more prevalent form of misinformation by identifying articles from reliable sources that are frequently co-shared with (shared by users who also shared) "fake" news on social media, and concurrently extracting narratives present in fake news content and claims fact-checked as false. Specifically in this study, we use Twitter/X data from May 2018 to November 2021 matched to a U.S. voter file. We find that narratives present in misinformation content are significantly more likely to occur in co-shared articles than in articles from the same reliable sources that are not co-shared, consistent with users using information from mainstream sources to enhance the credibility and reach of potentially misleading claims.

Motivation & Objective

• To investigate how factual news from reliable outlets is co-shared with fake news on Twitter, despite not being false themselves.
• To examine whether co-shared mainstream articles contain misinformation narratives at higher rates than non-co-shared articles from the same outlets.
• To determine whether the co-sharing phenomenon is driven by narrative content rather than partisan curation or article popularity alone.
• To demonstrate how legitimate news can be repurposed to enhance the reach and persuasiveness of false narratives in online misinformation ecosystems.
• To develop and apply a graph-based method to identify articles disproportionately co-shared with fake news, enabling narrative-level analysis.

Proposed method

• The authors use a graph-based approach where nodes represent fake news and mainstream news articles, and edges are weighted by the number of shared users on Twitter.
• They apply a statistical model to score the likelihood of co-sharing, controlling for individual article virality, to isolate disproportionate sharing.
• Narrative structures are extracted using an unsupervised NLP technique from false claims and used to identify entity-action relationships indicative of misinformation.
• Articles are classified as 'co-shared' if they are shared by a significantly higher proportion of users who also shared fake news, compared to expected co-sharing rates.
• The analysis controls for partisan curation by comparing co-shared and non-co-shared articles from the same outlets, using strict reliability criteria.
• A case study is conducted on a Washington Post article about vaccinated people comprising a majority of COVID-19 deaths, showing its disproportionate sharing by anti-vaccine networks.

Experimental results

Research questions

• RQ1 Do mainstream news articles that are co-shared with fake news contain more misinformation narratives than those not co-shared?
• RQ2 Is the co-sharing of mainstream articles with fake news driven by narrative content rather than partisan alignment or article popularity?
• RQ3 To what extent can factual news from reliable outlets be repurposed to support misleading narratives in online misinformation ecosystems?
• RQ4 How do narrative structures in fake news propagate through co-sharing with legitimate news articles on Twitter?
• RQ5 Can a graph-based method effectively identify mainstream articles that are disproportionately co-shared with fake news, enabling narrative-level analysis?

Key findings

• Mainstream news articles co-shared with fake news contain misinformation narratives at a significantly higher rate than non-co-shared articles from the same outlets, even after controlling for partisanship and popularity.
• The Washington Post article titled 'Vaccinated people now make up a majority of covid deaths' was disproportionately shared by users with a history of sharing anti-vaccine fake news, despite being factually accurate.
• The presence of specific misinformation narratives—such as claims about vaccine harm or election fraud—was a strong predictor of co-sharing, independent of outlet bias or article virality.
• Co-sharing is not merely a result of partisan curation, as the same outlets publish articles with and without such narratives, and only the narrative-rich ones are disproportionately co-shared.
• The study identifies 318 high-dimensional misinformation narratives from false claims, which are used to detect and validate narrative reuse in co-shared mainstream content.
• The findings suggest that mainstream news is not just a passive victim of misinformation but can be strategically repurposed to enhance the credibility and reach of false narratives.