Effects of Social Belief on Spatial Learning in Virtual Reality

Dylan Arena · Daniel Schwartz · Jeremy Bailenson

Thu., June 11, 11:00–12:30, Old Madison (3rd floor, East/Southeast)

As videogames and other virtual environments become increasingly sophisticated, new possibilities are emerging for interesting “hybrid” avatars that combine features of traditional avatars (specifically, behaving as the user commands) with those of computer agents (specifically, behaving independently). It is possible that such hybrid avatars will trigger social schemas for user interaction (negotiation for control, theory of mind questions, etc). Recent research from the LIFE Media lab group at Stanford indicates that the mere belief that a social interaction is occurring increases conceptual learning (e.g., Okita, Bailenson, and Schwartz, 2007). The main question of the current study is: How will manipulation of social belief related to interaction with an avatar in an educational virtual environment affect users’ learning of the layout of that environment?

To explore this issue, we asked undergraduate students in an introductory Communication class (n = 38) to guide an avatar through an immersive virtual reality environment (a maze of rooms) in an item-finding task. In certain rooms of the environment, participants received negative feedback whenever they made incorrect directional choices while searching for the target items. Half of participants (the "social" condition) were told that the avatar they were guiding through the virtual environment represented the experimenter and that the corrective feedback was coming from the experimenter; the other half of participants (the "solo" condition) were told that the avatar represented themselves and that the feedback was coming from the computer. The primary dependent measure was time (in seconds) to find all three target objects in the virtual environment in each of three successive trials. A covariate survey measure of self-reported sense of direction (the Santa Barbara Sense of Direction Scale) was included to account for individual differences in wayfinding ability.

Preliminary data analyses show that participants in the “social” condition learned the layout of the virtual environment more quickly than participants in the "solo" condition, even when taking into account self-reported wayfinding ability. Also, after the first item-finding trial, wayfinding ability no longer predicted performance for participants in the "social" condition, although it continued to predict performance for participants in the "solo" condition all the way through the third and final item-finding trial. Future directions and implications of these results for design of educational virtual environments will be discussed.

References

Okita S. Y., Bailenson, J., and Schwartz, D. L. (2007). The mere belief of social interaction improves learning. Paper presented at the Cognitive Science Conference, Nashville, TN.