How reliable is Crowdsourcing?
- When:
- Friday, August 3, 2012, 11:00 am - 12:00 pm
- Where:
- 11th fl Large CR (rm 1135)
- Speaker:
- Manuel Cebrian
- Description:
- AI SEMINAR Host - Kristina Lerman
Webcast link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=f017f9c8b05543e8b200b89838b6cfe51d
Coffee and snacks will be ready at 10:30am so you may come early to get settled in.Title: How reliable is Crowdsourcing?
Abstract:
If the Internet was designed with the goal to have a robust, fault-tolerant, and distributed computer network, Crowdsourcing, Internet's latest spawn, aims to be serves as a robust, fault-tolerant, and distributed collective problem solving platform. To put this idea to test, government agencies held the 2009 DARPA
Network Challenge, the 2011 DARPA Shredder Challenge, and 2012 Department of State's Tag Challenge. In the DARPA Network Challenge competing teams were asked to locate 10 red weather balloons in the continental US under 8 hours. In the Shredder Challenge, DARPA asked participants to reconstruct 5 puzzles of the hardest computational nature under 5 weeks. In the Tag Challenge, team had to locate 5 jewel thieves in cities across Europe and the US under 12 hours. Using a recursive incentive mechanism that both spread information about the task and incentivized individuals to act, our team was able to win the Network and Tag Challenges, and make significant progress in the Shedder Challenge while suffering constant attacks. This talk outlines the theoretical and empirical properties of this mechanism, but places emphasis on the observed adversary capability and intent. The multiplicity of aggression episodes serves to illustrate the different factors that make Crowdsourcing vulnerable.Bio:
Manuel Cebrian is currently an assistant research scientist with the Department of Computer Science and Engineering at the University of California, San Diego, and a senior research fellow with the Media Laboratory at the Massachusetts Institute of Technology. Manuel received a Ph.D. in computer science from Universidad Autonoma de Madrid (Spain) in 2007. He recently finished a stint as a visiting scientist with the Data Science team at Facebook. Previously, Manuel worked as a junior researcher with the User Modeling Laboratory at Telefonica Research, after a first postdoctoral experience with the
Department of Computer Science at Brown University. Manuel's work lies at the intersection of the computer and social sciences. His primary interests include social and financial networks, crowdsourcing, consumer research, urban economics, epidemiology, behavioral game theory, and evolutionary dynamics. Manuel is particularly involved in hands-on research in time-critical social mobilization, which is best illustrated by his participation in the 2009 DARPA Network Challenge, the 2011 DARPA Shredder Challenge, and the recent 2012 Department of State Tag Challenge.-Alma Navaanava@isi.edu
