Developing Computational Models of Players' Identities and Values from Videogame Avatars

Videogame avatars, while technically virtual identities, also often reflect the aspects of players’ identities and values in the physical, social, and cultural world of the player outside of the game (i.e., the “real world”). These aspects of real world identity include phenomena such as social group categorization and stereotyping. In this paper, we present AIRvatar, an Artificial Intelligence (AI) system that models player values and social identity related phenomena from the computational data structures of virtual avatars. We conducted a user-study with 191 participants who created characters using an avatar creation system of our own design together with AIRvatar. We then constructed identity models of players by analyzing the graphical images, text descriptions, and numerical attribute distributions of these constructed avatars. These allowed us to identify and characterize aspects of players’ real-world values related to identity (e.g., gender-based stereotyping and stereotypical associations of traditional videogame roles,) from the technical data structures and implementations of players’ avatars.