Avatar Interfaces for Biobehavioral Feedback Tylar Murray 1 , Delquawn Hardy 1 Donna Spruijt-Metz 2 , Eric Hekler 3 , and Andrew Raij 1 1 USF Electrical Engineering, 2 USC Keck School of Medicine, 3 ASU School of Nutrition and Health Promotion tylarmurray@mail.usf.edu; dhardy@mail.usf.edu; dmetz@usc.edu; ehekler@asu.edu; raij@usf.edu, Abstract. The combination of inexpensive wearable sensors, powerful mobile phones, and always-connected cloud computing are enabling new, real-time feedback and coaching via mobile technologies. This paper ex- plores the use of avatars - digital representations of the self - as an ideal user interface for mobile health application. Specifically, a justification for using avatars is provided, both based on empirical studies and the psychology of human body interpretation. We then provide an organized, theoretical description of how an avatar’s traits (appearance, behavior, and virtual environment) can be manipulated to convey specific health- related behavior change messages. Keywords: avatars, development, Proteus Effect 1 Introduction Through real-time processing of sensor and self-report data, mobile health (mHealth) systems can provide appropriate and timely biobehavioral feedback anytime and anywhere. Indeed, several mHealth systems have changed health behaviors [24] and outcomes [20] with varying levels of success. One under-explored source of feedback are avatars, digital representations of the self. While some research and commercial health systems use avatars [11, 12, 17, 18, 28–30], the work to date does not provide a clear explanation for why avatars could be a good tool for behavior change, nor is it clear how to best use avatars for effective behavior change. When the avatar is a more prominent part of the interface, its evalu- ation is minimal, or there is no discussion of what aspects of the avatar were influential. Direct comparison of systems is difficult due to differences in domain, implementation, and design of previous works. Methods of comparison between implementations are not well explored in this area, and the components which make up an avatar interface are not well defined. In this paper, we begin to ad- dress this gap in the literature by providing 1) a clear motivation for the use of avatars in behavior change applications, and 2) a design language and methodol- ogy to facilitate more structured design, analysis, and evaluation of avatar-based behavioral change interventions.