CHI99-- Paul Resnick

CHI 99 Online Communities Workshop
Position Paper

Paul Resnick

University of Michigan School of Information 314 West Hall Ann Arbor, MI 48109-1092 presnick@umich.edu last updated 2/26/99

Research Summary

I have two interests relevant to this conference.

First, I am studying reviews, recommendations, reputations, and other mechanisms that make it safe, fun, and profitable for people to interact with strangers, especially through electronic media. In particular, I am studying reputation systems, which gather, aggregate, and distribute people's comments about each other. The most prominent example of such a system is the feedback forum at eBay.

Second, I am interested in information systems that support non-virtual communities, especially on very small scales. In particular, I am experimenting with the use of neighborhood directories and email lists. This work is motivated in part by my participation in the Saguaro Seminar on Civic Engagement.

Research Methodology

For the study of reputation systems, my primary research method to date has been game theory. For example, in work with Eric Friedman, I have modeled the impacts of cheap pseudonyms on the emergence of trust in public spaces. We proved that there is no equilibrium with substantially less defection overall than one based on requiring newcomers to pay dues. Direct payment does it, as does a period of initiation waiting for the buildup of trust. This is an important result about the theoretical limitations of reputation systems. When a service is growing rapidly, there will be lots of newcomers. If players can freely rejoin a service under new pseudonyms, a natural setting on the Internet, then newcomers will have to be treated worse in some way than veterans. We also showed one way to alleviate this problem, using a cryptographic technique called blind signatures and a third party to enable newcomers commit not to return with different pseudonyms.

For the study of computational support for physical communities, I am doing iterative design and trials with informal evaluation.

Future Directions for My Research

Reputation Systems. In future work on reputation systems, I will extend the game theory analysis, characerize the design space of systems, and do empirical analysis of how reputation systems impact behavior in markets where they are used, including eBay.
Supporting physical communities. Once I have evolved what I believe are useful tools and social processes surrounding neighborhood directories and email lists, I would like to evaluate the impact of these tools, using social network measurement techniques (e.g., ask people on a block to mark which houses where they know someone, then ask again a year after introducing new information systems). If the tools are helpful, I will work on packaging them for easy adoption and adaptation by others.

Other Important Issues for the Field

Federated communities. What kind of support is useful for collections of small communities (either overlapping or non-overlapping)?

Selected Publications