Detecting Conspiracy Theories on Social Media: Improving Machine Learning to Detect and Understand Online Conspiracy Theories

Abstract

Conspiracy theories circulated online over social media contribute to a shift in public discourse away from facts and analysis and can contribute to direct public harm. Social media platforms are encountering a difficult technical and policy challenge as they try to mitigate harm from online conspiracy theory language. As part of an effort to confront emerging threats and incubate new technology to help create a safer world, Google's Jigsaw unit asked RAND Corporation researchers to conduct a modeling effort to improve machine-learning technology for detecting conspiracy theory language by using linguistic and rhetorical theory to boost performance. We also aimed to synthesize existing research on conspiracy theories with new insight from this improved modeling effort. In this report, we share insights from the effort and offer recommendations to mitigate harm and reduce the effects of conspiracy theories online.The research reported here was completed in January 2021 and underwent security review with the sponsor and the Defense Office of Prepublication and Security Review before public release.