Aion

Fusing Quantitative and Qualitative Methods in Virtual Worlds Behavioral Research

A mixed-methods study conducted by SAIC and Georgia Tech examines whether real-world characteristics of players can be inferred from how they behave and appear as avatars in massively multiplayer online role-playing games.

• Gender was predicted from virtual-world observation with 84% overall accuracy, the strongest result in the study of 80 Guild Wars and Aion participants, followed by extraversion (73%), aggressive ideology (73%), submissive ideology (71%), and approximate age (66%).
• The study fused standardized avatar ratings with 'thick description' field notes converted into quantitative predictors via automated text analysis, then analyzed them with discriminant analysis and leave-one-out cross-validation.
• Key predictors included male avatar appearance, movement patterns, and avatar class; aggressive ideology was associated with non-strategic play, melee combat, and frequent quest log checking, though its 53% precision indicates many false positives.