Dan Cosley, Visiting Assistant Professor, Cornell University

"Computational social science with an HCI accent"

Monday, February 18, 4:00 PM

Packard Lab Room 416

Reception prior to talk in Packard Lab lobby at 3:30 PM

Abstract:  The rise of social web applications and online community provides a platform for doing "computational social science", understanding and improving the technologies that support these communities.  In this talk, we will focus on member-maintained communities, systems where many people participate in activities such as moderating contributions and building public goods like Wikipedia.  These systems often capture rich traces of activity, allowing us to study the impact of algorithms, social structures, and interfaces on people's behavior.  For example:

* A common interface idiom for collecting people's opinions about movies actually changes the opinions they express, with potentially awful consequences for recommendation systems based on these opinions.

* Simple algorithms for matching people with tasks they might want to do dramatically increases their willingness to perform work for the good of the community.

* Recommendations are usually based on what you do, but it turns out that who you know is sometimes a better predictor of your future behavior.

Along the way we will talk about the challenges of doing this kind of research in a theoretically grounded and generalizable way, and end with a grand challenge for doing computational social science.

Bio: Dan Cosley is a visiting assistant professor in communication at Cornell University, teaching courses in human-computer interaction. His primary interest is helping groups make sense and use of information. His recent work focuses on designing systems that encourage members to contribute more to shared community resources, combining social science theory, HCI design principles, and computational tools to motivate participation. He is also interested in the more general problem of how to apply theories of behavior to the design of systems in a way that helps future designers and theorists use the results.

His prior work focused on recommender systems, especially new ways to use recommendation algorithms, better interfaces for recommender systems, and better methods for evaluating their utility. As a systems builder, he developed the SuggestBot tool for Wikipedia, played an important role in building the successful research recommendation system MovieLens(http://movielens.umn.edu/), and developed the SmartShopper, a successful mobile shopping list application. He also helped teach Google how to play "Who Wants to Be a Millionaire?"