Multi-modal Analysis of Misleading Political News

The internet is a valuable resource to openly share information or opinions. Unfortunately, such internet openness has also made it increasingly easy to abuse these platforms through the dissemination of misinformation. As people are generally awash in information, they can sometimes have difficulty discerning misinformation propagated on these web platforms from truthful information. They may also lean too heavily on information providers or social media platforms to curate information even though such providers do not commonly validate sources. In this paper, we focus on political news and present an analysis of misleading news according to different modalities, including news content (headline, body, and associated image) and source bias. Our findings show that hyperpartisan news sources are more likely to spread misleading stories than other sources and that it is not necessary to read news body content to assess its validity, but considering other modalities such as headlines, visual content, and publisher bias can achieve better performances.