Grant: Discovery Grants Program - Individual

In recent years, information campaigns have often been organized through communicative activities, providing users with malicious intentions with an opportunity to reach and deceive a large user base to advance their agenda. As part of their strategic information operations, social media have been strategically leveraged by numerous actors for political and commercial gains from the manipulation of democratic processes in Britain and the U.S. in 2016, to the coordinated spreading of the China bioweapon conspiracy theory about coronavirus in 2020. The problem is of paramount importance not only because manipulative campaigns are designed to circulate disinformation to shape public opinions but also for their insidious nature in exploiting the concerns, fears and prejudices of ordinary users and enrolling them to circulate misinformation. Existing research has primarily focused on characterizing influence campaigns; assessing their impacts; and inferring their strategies and tactics based on observed behaviors. Recent advancements propose novel techniques to detect information campaigns by exploring their propagation patterns built from the network structure, coordinated behaviors, temporal user activities, time-series features, and content features. However, existing approaches either detect information campaigns at a later stage of its development or rely on historical digital traces to perform postmortem analysis of campaigns after takedowns. The objective of this research program is to develop computational techniques for early prediction of information campaigns, in the context of which we have limited to no prior knowledge about the information campaign. We will develop methods that identify different characteristics of manipulative information campaigns from different information sources: a) content level to predict if the social content has a potential to turn into information campaigns; b) user perspective such as metrics driven from the social network graph structure or behavioral characteristics of the user to predict if the socially-shared contents by the user have the potential to turn into information campaigns; and c) propagation data to predict the occurrence of the information campaigns from their observed propagation patterns. The outcome of this research will be highly impactful as timely detection of manipulative campaigns minimizes their negative impacts on individuals and society by soliciting early interventions from authorities before they become widespread. The research program will train in-demand HQP skilled in researching and tackling this growing area of research for diverse roles in academia, government and industry.