Bot or Not? Social Identity Biases in the Perceived Humanity of Online Profiles

• In Event: 20220 - Section on Communication, Information Technology, and Media Sociology Roundtables
  In Refereed Roundtable (60 minute): Section on Communication, Information Technology, and Media Sociology Roundtable I: Artificial Intelligence I

Abstract

Social categorization and identity are central to how people navigate their social world. The rise of generative artificial intelligence systems, particularly in online social settings, complicates these fundamental processes by blurring traditional boundaries between human and machine-generated social performances. While human-AI interaction now emerges in a variety of settings, social media platforms provide a venue where expectations about authenticity are increasingly challenged by ambiguity. We seek to develop this understudied space by investigating the impact of salient identity characteristics– race, gender, and partisanship– on perceptions as human or bot. Drawing on theories of social identity, one might expect individuals to question the authenticity of perceived out-group members while readily accepting those who share their identity characteristics. However, our experiment reveals a more complex reality: while white Republican men were likely to label accounts displaying identity cues of their outgroup as bots, participants more frequently questioned the humanity of apparent in-group members– with this effect particularly pronounced among members of marginalized groups. We interpret these findings through the theoretical lenses of Goffman’s stigma consciousness and Du Bois' double-consciousness, suggesting that the heightened self-awareness experienced by marginalized groups may extend to increased skepticism about the authenticity of apparent in-group members in digital spaces. As artificial intelligence systems increasingly incorporate into public life, our findings begin to show how these emerging technologies may disrupt the formation and perception of meaningful social categories. Our results suggest that pre-existing patterns of bias may create conditions that cause the holders of stigmatized identities to disproportionately suffer dehumanization as a product of identity-ambiguity created by new technologies.

Authors

• Benjamin Rochford, Duke University

• Cathy Lee, Duke University

• Michelle Qiu, Duke University

• Alexander Volfovsky, Duke University

• D. Sunshine Hillygus, Duke University

• Christopher A. Bail, Duke University