Blended Bots: Infiltration through Identity Deception on Social Media

TL;DR

This paper investigates how bots on social media mimic human identities, especially gender stereotypes, revealing that gender bias is more prevalent in human profiles and identifying identity features that differentiate bots from humans.

Contribution

It introduces an analysis of identity patterns in social media profiles, highlighting gender stereotypes and their role in distinguishing bots from humans, and discusses implications for bot detection training.

Findings

Some identity types differentiate between human and bot profiles.

Gender bias is more prevalent in human accounts than in bots.

Identifies identities strongly associated with gender in bot profiles.

Abstract

Bots are automated social media users that can be used to amplify (mis)information and sow harmful discourse. In order to effectively influence users, bots can be generated to reproduce human user behavior. Indeed, people tend to trust information coming from users with profiles that fit roles they expect to exist, such as users with gender role stereotypes. In this work, we examine differences in the types of identities in profiles of human and bot accounts with a focus on combinations of identities that represent gender role stereotypes. We find that some types of identities differentiate between human and bot profiles, confirming this approach can be a useful in distinguishing between human and bot accounts on social media. However, contrary to our expectations, we reveal that gender bias is expressed more in human accounts than bots overall. Despite having less gender bias overall, we provide examples of identities with strong associations with gender identities in bot profiles, such as those related to technology, finance, sports, and horoscopes. Finally, we discuss implications for designing constructive social media bot detection training materials.