2014 • Computer and Information Sciences
A web-based laboratory for computational social science
Lead Presenter: Ceyhun Karbeyaz
Additional Presenters: David Lazer, Waleed Meleis, Alan Mislove, Brian Keegan
Faculty Advisor / PI: David Lazer, Waleed Meleis
While laboratory experiments are a methodological cornerstone across the social and behavioral sciences, physical laboratories have drawbacks that lead to unrepresentative populations, small participant pools, and substantial organizational complexity. Web-based platforms have the potential to transform this experimental research paradigm by running larger experiments more quickly and more efficiently. We developed a web-based experimental behavioral laboratory called Volunteer Science to enable researchers to quickly design and deploy behavioral experiments while recruiting participants from a substantially larger subject pool of any active Facebook user. This platform enables us to conduct behavioral experiments involving hundreds of participants while supporting synchronous and client-side interactions while using commodity web service platforms. We describe the design of Volunteer Science and review features that allow researchers to design highly customizable experiments such as synchronous interaction and text chat while also logging real-time and detailed user behavior. We review preliminary results of our replications of well-known experiments such as the traveling salesperson problem, iterated prisoner’s dilemma, public goods games, and the ultimatum game and conclude by discussing implications this platform has for experimental behavioral research in the social and behavioral sciences.