Social Identity Enactment Through Linguistic Style: Using Naturally Occurring Online Data to Study Behavioural Prototypicality

Abstract

Social identity prototypes refer to the quintessential representation of a particular social identity; prototypes define and prescribe the characteristics, behaviours and attitudes of a particular group, as distinguished from other groups (Hogg, 2001). For the most part, identity prototypicality is studied using self-reported methods used to assess perceptions of the prototypicality of self and others. However, in this thesis we provide behavioural evidence to demonstrate how linguistic style data can be used to measure identity-prototypical behaviour in real world contexts. Combining naturally-occurring online data with experimental data, the first chapter demonstrates that individuals behave in an identity-prototypical way regardless of the context in which they are communicating. Further, we show that this identity-prototypical style of communication is robust to topic, demographics, personality and platform, and moreover that the same identity-prototypical communication style can be detected in experimentally controlled conditions. In the second chapter, we demonstrate the small but statistically significant link between identity-prototypical communication and influence in real-world forum data. This finding provides insight into how group members respond to other ingroup members based on their prototypical communication style in real-world situations. Finally, in the third chapter, we use the group prototypical behaviour observed in naturally occurring online forum data to construct a typology of social identities, demonstrating the existence of five different types of social identity in line with the research of Deaux et al. (1995). We also demonstrate that it is possible to use this measurement of behavioural prototypicality to observe identity change over time. Using eight years’ worth of forum data, we illustrate the slow movement of the transgender identity from being a stigmatised identity in 2012, to shifting towards a collective action identity in 2019. In sum, the findings outlined in this thesis provide evidence to support the idea that it is possible to use machine learning algorithms and naturally occurring online data to study behavioural prototypicality in real world environments. Moreover, this methodology enables us to study identities ‘in the wild’ thus transcending the limitations associated with using self-reported methodologies or experimental approaches to study how individuals express and enact their group memberships. Further, we also demonstrate the value in using naturally-occurring online behavioural data to test and extend the key components of social identity theory.