Hiding Radical Speech in Plain Sight: Covert Identity Signaling on Social Media

Abstract

There has recently been a surge in online political polarization and radicalization, occurring alongside an increase in the popularity of offline radical movements. In response, researchers have developed techniques to detect and measure extreme forms of radical speech online. However, these techniques are aimed at detecting overt radical speech, which refers to messages that are broadcast honestly and without disguise. In many cases, radical ideas are instead spread using coded language, so-called covert communication, which is designed to be accurately received by its intended audience but obscured from those who do not share the same views. Examples of covert communication include conversational humor, flirting, and political dog whistles. Today there is little understanding about how covert radical speech is used online and when and where it is preferred to overt forms of speech. This interdisciplinary project uses innovative, theoretically-driven methodology to disambiguate overt and covert radical political speech, and to explain the mechanisms underlying the changes in their relative use. Covert and overt speech can be viewed within the framework of identity signaling, whereby individuals signal their identity to find others with similar norms, views and affiliations, or to reaffirm existing alliances. Recently, one of the PIs proposed a formal mathematical theory to explain the functional advantages of overt versus covert speech (Smaldino, Flamson, & McElreath, 2018). Simply put, overt signaling maximizes information transfer, and so should be favored when social costs of alienating dissimilar others are low. Covert signaling minimizes the chances of detection by dissimilar others, and should be favored when such detection is potentially costly. The project will use this theoretical framework to investigate the dynamics of covert and overt radical political speech in real-world settings. There are three primary research objectives: (1) Identify and quantify overt and covert radical speech online. (2) Investigate how the level and ease of assortment with similar others offline affect the frequency of overt as opposed to covert signaling about radical views online. (3) Analyze how external events that change social costs of expressing radical views affect the frequency of covert compared to overt signaling. A combination of research methods will be used to achieve these objectives. An initial dataset of tweets along with behavioral studies will be used to build a training sample of radical covert and overt speech. This annotated sample will then enable us to uncover such speech in larger metropolitan and national samples of tweets by applying natural language processing techniques and machine learning. Finally, this project will analyze how trends in overt and covert speech over time and geographical locations relate to the level and ease of assortment in offline social networks, and to external events that change the costs of expressing radical views. The results of this project will be valuable for the study of human behavior and for the public good. This project will advance social and behavioral science related to social networks and communication. It will be the first to empirically test the formal theory of covert signaling and to develop methods for detecting the use of covert radical speech online. Results from this project will enable a deeper understanding of how the use of covert and overt speech relates to social network structure and culturally relevant events. In addition, this project will provide training opportunities to several early-career scholars.

Document Details

Document Type
DoD Grant Award
Publication Date
Jul 09, 2020
Source ID
W911NF2010220

Entities

People

  • Paul E. Smaldino

Organizations

  • Army Contracting Command
  • United States Army
  • University of California

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Irregular Warfare and Special Operations Cyberspace Operations against Adversarial Threats.
  • Political Violence and Terrorism Studies.

Technology Areas

  • AI & ML