Abstract
The spread of misinformation or fake-news is a global concern that undermines progress on issues such as protecting democracy and public health. Past research aiming to combat its spread has largely focused on identifying its semantic content and media outlets publishing such news. In contrast, we aim to identify individuals who are more likely to share fake-news by studying the language of actors in the fake-news ecosystem (such as fake-news sharers, fact-check sharers and random twitter users), and creating a linguistic profile of them. Fake-news sharers and fact-check sharers use significantly more high-arousal negative emotions in their language, but fake-news sharers express more existentially-based needs than other actors. Incorporating psycholinguistic cues as inferred from their tweets into a model of socio-demographic predictors considerably improves classification accuracy of fake-news sharers. The finding that fake-news sharers differ in important ways from other actors in the fake-news ecosystem (such as in their existential needs), but are also similar to them in other ways (such as in their anger levels), highlights the importance of studying the entire fake-news ecosystem to increase accuracy in identification and prediction. Our approach can help mitigate fake-news sharing by enabling platforms to pre-emptively screen potential fake-news sharers' posts.
Abstract (translated)
URL
https://arxiv.org/abs/2203.10560
