Egress Online: Towards Leveraging Massively, Multiplayer Environments for Evacuation Studies
User studies and evaluation
Large datasets of real human behaviors are of huge benefit across numerous domains, including evacuation safety, urban planning, marketing, and ergonomics. However, because large-scale experiments involving real human subjects are expensive and prohibitively difficult to organize, such datasets are scarce. Thus in this paper, we propose the use of massively multiplayer online (MMO) communities as an inexpensive and innovative way to capture datasets of large numbers of people under different conditions. We describe our implementation of an online data collection system, based on games, inside the popular massively multiplayer, online environment of Second Life. We evaluate the use of this system for performing evacuation experiments using a mix of Second Life residents and players recruited on campus. Our system was able to draw online participants, support data collection needs, and provide potential insights into high-level evacuation behaviors such as the choices of exit, effects of building debris, and the use-patterns of a building. Through experiments performed using our system, we found that Second Life residents found the game controls and environment to be significantly more compelling than lab participants; that players unfamiliar with our office building tended to evacuate primarily via the front entrance; and that in-game debris significantly increased the numbers of participants who failed to exit a building safely.