Using Prediction Markets to Incentivize and Measure Collective Knowledge Production

One of the toughest challenges to ensure the emergence of collective intelligence and knowledge production is to rightly set individual incentives towards the achievement of a common goal [Pickard et al. 2011]. Unfortunately, most attempts to establish and maintain long-term collective action are confronted with diverging individual incentives [Hardin 1968], and organizational issues [Ostrom 1990]. On the Internet, collective action is best embodied by open source software (OSS) and other non-software knowledge production projects, like Wikipedia. Our current knowledge of OSS communities reveals how they achieve knowledge production in a seamless self-organized way called peerproduction, which relies on two fundamental organizational rules : (i) task self-selection and (ii) peerreview [Benkler 2002]. Beyond these rules, the success or the failure of OSS projects is most often bound to people’s capabilities to build true communities, through institutions that help solve the tensions existing between individual incentives and the collective interest. Building and adapting institutions for open collaboration is often a hard and painful enterprise, which can put at stake the very survival of an OSS project [O’Mahony and Ferraro 2007]. Here we present a mechanism design, based on a Wiki-like collaboration platform coupled to a prediction market, which aligns individual incentives with the goals of collective knowledge peer-production, and does not require further governance mechanism. We have implemented and tested four instances of this system in the context of higher education for the courses Entrepreneurial Risks (ETH Zurich; Course No 351-0564-00L; instructor Didier Sornette) in 2011, 2012, 2013, and Environmental Entrepreneurship (EPFL; instructor Marc Vogt) in 2013. Complementing business management experiences [Hoyt and Rao 2006; Benbya and Van Alstyne 2011], this system explicitly formalizes the rules of peer-production. Most prediction markets are designed to organize information rather than to engage individual knowledge production as a collective good. Our implementation helps value not only information, but also individual and collective behaviours, which are at the very roots of collective intelligence.