Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of GraphSearch.
We consider a participatory sensing scenario where a group of private sensors observes the same phenomenon, such as air pollution. Since sensors need to be installed and maintained, owners of sensors are inclined to provide inaccurate or random data. We design a novel payment mechanism that incentivizes honest behavior by scoring sensors based on the quality of their reports. The basic principle follows the standard Bayesian Truth Serum (BTS) paradigm, where highest rewards are obtained for reports that are surprisingly common. The mechanism, however, eliminates the main drawback of the BTS in a sensing scenario since it does not require sensors to report predictions regarding the overall distribution of sensors' measurements. As it is the case with other peer prediction methods, the mechanism admits uninformed equilibria. However, in the novel mechanism these equilibria result in worse payoff than truthful reporting.
Fabio Nobile, Sundar Subramaniam Ganesh
Edoardo Charbon, Pouyan Keshavarzian, Francesco Gramuglia, Mario Stipcevic