Incentivizing Truthful Reporting from Strategic Sensors in Dynamical Systems

Abstract: Human agents are increasingly serving as data sources in the context of dynamical systems. Unlike traditional sensors, humans may manipulate or omit data for selfish reasons. Therefore, this paper studies the influence of effort-averse strategic sensors on discrete-time LTI systems. In our setting, sensors exert costly effort to collect data, and report their effort to the system operator. However, sensors do not directly benefit from the output of the system, so they will not exert much effort to ensure accuracy and may even falsify their reported effort to maximize their utility. We explore payment mechanisms that incentivize truthful reporting from strategic sensors. We demonstrate the influence of the true and reported effort on the expected operational cost. Then, we use the realizations of the system cost to construct a payment function. We show that payment functions typically used in static settings will not be able to elicit truthful reports in general, and present a modified payment function that elicits truthful reporting, which requires terms that compensate for the dynamic impact of reported efforts on the closed-loop performance of the system.