This project studies the design of information systems like wikis and information markets. Research in social science has established that often there is a "wisdom of the crowd" -- i.e., collectives can display more intelligence than the individuals they are composed of. When such collective information systems work, they serve as superb aggregators and disseminators of information. However, fundamental computational challenges remain in understanding how to design them optimally.
This research is advancing along several lines, including
(1) general theories of how information is aggregated in different social media, developed and validated using real data gathered from existing databases and generated from user experiments;
(2) algorithms for facilitation of user interactions so that the medium in question can deliver the promised results (for example, market-making algorithms for liquidity provision in information markets);
(3) theoretical and practical characterization of the possibilities for rogue users to manipulate collective wisdom systems;
(4) algorithms for detecting malicious users, and mechanisms that thwart miscreants.
The research is naturally interdisciplinary in nature, drawing from machine learning and probabilistic reasoning, data mining and social networks, as well as finance and economics. It contributes to our understanding of complex social phenomena like the growth of information in wikis and blogs, as well as to the development of intelligent reasoning algorithms for agents in complex, uncertain multi-agent environments like markets.
The design of agents that participate in markets and social systems improves the quality of online markets and improves information flow in virtual spaces. Further, insights gained from modeling market structures and social spaces can tell us how to design them better. For example, understanding the impact of different levels of central control on wiki articles or open source software projects yields guidelines for how much central control is optimal in different settings.
In a world where computation and social systems are increasingly intertwined, the PI's research and education program exposes students to multidisciplinary ideas through the introduction of a new class on collective intelligence, social networks and e-commerce, and the development and extensive use of the very objects of study -- information markets and wikis -- in classroom and lab settings. The PI is also developing an experimental project for putting freely accessible course wikis online, similar to online course materials at other universities, but open to editing by the community.
