*Social Computing and User Generated Content: A Game-theoretic Approach*

Arpita Ghosh
Senior Research Scientist
Microeconomics and Social Systems Group
Yahoo! Research

Wednesday, May 30
4:30 - 5:30 PM
Y2E2 101

Abstract:

User-generated content (UGC), such as the reviews on Amazon and Yelp, comments, and contributions in online Q\&A and discussion forums, now constitute a large fraction of the useful content on the Web. But while there is a large amount of UGC online, not all of it is of the same {\em quality}. What can we understand, using an economic approach, about what {\em incentive schemes} elicit high quality contributions, as well as adequate participation, in such systems?

We provide a game-theoretic model with strategic, exposure-motivated contributors, and use this model to investigate the widely used {\em rank-order} allocation, where users' contributions are displayed on a webpage in decreasing order of their ratings. Such an allocation of attention constitutes a mechanism, which can influence the quality of content elicited from attention-motivated contributors. We show that the {\em lowest} quality that can arise in {\em any} mixed strategy equilibrium of the rank-order mechanism becomes optimal as the available attention diverges, while maintaining high participation. Next we compare rank-order mechanisms against the more equitable proportional mechanism, which distributes attention in proportion to the number of positive ratings, and show that the rank-order mechanism, while less fair?, almost always incentivizes higher quality equilibrium contributions in the limit of diverging attention. We then move to non- diverging reward regimes, and use a model with {\em endogenous entry} to analyze mechanisms for crowdsourcing environments such as conventional crowdsourcing contests (Innocentive, TopCoder) as well as crowdsourced content, as in online Q\&A forums. Unlike models which treat participation as an exogenous choice, the expected number of participants here can be increased by subsidizing entry, potentially improving the expected value of the best contribution. However, we show that free entry is, in fact, dominated by taxing entry--- making all entrants pay a small fee which is rebated to the winner can improve the quality of the best contribution over a winner-take-all contest with no taxes. Based on joint work with Patrick Hummel (EC 2011) and Preston McAfee (WWW 2012).

Bio:
Arpita Ghosh is a Senior Research Scientist at Yahoo! Research in the Microeconomics and Social Systems Group. Her research focuses on algorithms and mechanism design in the context of the Web, particularly social computing and user-generated content, online advertising, matching markets, and privacy. She holds a PhD from Stanford University.

Operations Research Colloquia: http://or.stanford.edu/oras_seminars.html