Selective information disclosure in contests

Contests are important mechanisms to elicit work (ideas) from crowds. While contests have been used throughout history (e.g. the British government's 1714 Longitude Prize), they have gained popularity in the current Internet era, and, in particular, in the context of crowdsourcing [2, 7, 14, 41, 59, 60].Well known examples include the Netflix prize (netflixprize.com), Darpa challenges [3, 57] and the Hult prize (hultprize.org), as well as various public platforms that allow requesters to solicit contributions through contests with monetary prizes, such as taskCN (www.taskcn.com), TopCoder (www.topcoder.com) and Kaggle (www.kaggle.com). As such, the study and analysis of contests have become prominent in mechanism design and multi-agent systems literature [6, 14, 23, 36-39, 43, 54]. These include both the analysis and determination of optimal strategies - for the contestants, and methods for the design of effective contests - for the contest's organizer. In this work we concentrate on the latter issue - that of contest design.