The effects of anthropomorphic avatars vs. non-anthropomorphic avatars in a jumping game

Abstract

Avatar identification is a topic of increasingly intense interest. This is largely because avatar identification can promote a wide variety of outcomes: game enjoyment, intrinsic motivation, quality of made artifacts, and more. Yet we still understand very little about how different avatar types affect users. Here, we contribute one of the few highly controlled studies of this nature (N=1074). Specifically, we compare three avatar types in a jumping game: 1) Human (high anthropomorphism), 2) Block-like (low anthropomorphism), and 3) Robot (high anthropomorphism). We find that players randomly assigned to the Robot condition have significantly higher player experience. We find that both Robot and Human conditions lead to higher avatar identification. Finally, using linear hierarchical regression, we find that avatar identification significantly promotes player experience (29.8% variance) and time played (3.5% variance). Our study demonstrates the importance of considering avatar type in designing virtual systems.