Toward Telemetry-driven Analytics for Understanding Players and their Avatars in Videogames

Abstract

A great deal of research has gone into understanding the relationships between social behaviors, player preferences, and identities in both real-world and virtual environments. In this works-in-progress report, we describe AIRvatar, a tool that telemetrically collects data such as session-based time durations and click events during the process of avatar customization within a videogame of our own creation. We present results from a user-study of 181 players, highlighting how social phenomena such as gender-related stereotypes of users can be revealed, particularly when players' self-identified real world genders contrast with the gender identities of their avatars.

The Marginal: A Game for Modeling Players' Perceptions of Gradient Membership in Avatar Categories

“…Currently, our database contains 191 created avatars that were obtained as a result from previous user study. More details on both AIRvatar and the user-study involving Heroes of Elibca are available in (Lim and Harrell 2015a;2015c). This database is used an AI-system AIRlib that constructs various different computational models using the data.…”

“…This emergent, "bottom-up" approach better reflects the implicit values and preferences of players compared to system creators determining them on their own. Studying players' avatar customization behaviors in this way has previously shown interesting findings, such as players conforming to common RPG class or gender-related stereotypes (Lim and Harrell 2015b;2015c). The broader implications could involve alternative data sets based on other aspects of society in the real world (e.g., demographic, education, employment), using it to model and reveal society's values of other socially-charged topics.…”

A Data-Driven Approach for Computationally Modeling Players' Avatar Customization Behaviors

“…Previous Work In our previous work, we studied "infrastructural values" that were built into systems by analyzing the end-product of constructed avatars, including how statistical attributes and visual characteristics reflected social phenomena potentially symptomatic of developer bias or implicitly shared worldviews (Lim and Harrell 2015a;2015b;2015c). Building on those findings, here we focus on modeling "user values" through behaviors enacted out within systems, which particularly include the temporal properties of performing sequences of actions within a given system.…”