Community-Based Fact-Checking on Social Media

• Applicant: Professor Dr. Nicolas Pröllochs
• Subject Area: Accounting and Finance
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 492310022

Misinformation undermines the credibility of social media and poses significant threats to modern societies. As a countermeasure, Twitter has recently introduced “Birdwatch,” a community-driven approach to address misinformation by harnessing the wisdom of crowds. On Birdwatch, users can identify tweets they believe are misleading, write fact-checks that provide context to the tweet, and rate the quality of other users’ fact-checks. Given the scale of misinformation on social media platforms, there is an urge to understand the mechanisms and challenges in the context of community-based fact-checking. Corresponding findings would have direct implications for Twitter’s Birdwatch platform and future attempts to implement community-based approaches to combat misinformation on social media. Therefore, this project aims to analyze community-based fact-checking on Twitter holistically. Specifically, we will address three main research goals: (1) we will analyze factors that contribute to a helpful fact-check. Here we aim to understand the rating mechanism on Birdwatch, which is a key feature that should help to identify the context that people will find most helpful. (2) We will analyze how community-based fact-checking is associated with the spreading dynamics of the fact-checked tweet on Twitter. For this purpose, we will use the Twitter Historical API to reconstruct the retweet cascades for fact-checked tweets. We will then empirically analyze the relationship between the veracity of fact-checked tweets (i.e., whether a fact-checked tweet presents misinformation) and user sharing behavior on Twitter. The overall goal is to understand whether community-based fact-checking actually mitigates the spread of misinformation on social media (or rather yields adverse effects). (3) We will shed light on the role of emotions in community-based fact-checking. Here we will build upon emotion theory and employ quantitative text analysis to analyze emotions embedded both in the fact-checked tweets and the fact-checks. We are particularly interested in understanding whether tweets embedding certain emotions are more likely to be fact-checked and to yield certain emotions in the fact-check (e.g., anger). As an extension, we will evaluate whether state-of-the-art machine learning models can help to predict the spread of fact-checked tweets. Based on the generated findings, we will derive important implications for social media platforms.

• DFG Programme: Research Grants