Revenue-Maximizing Stable Pricing in Online Labor Markets

Abstract

In online labor markets, millions of paid tasks are performed by workers every day. We solve the stable pricing problem, that is, given the information about tasks and workers, to find a revenue-maximizing mechanism — pricing and allocation — that is stable (where no worker or task is treated unfairly), and truthful (tasks reveal their true needs). We propose two truthful, stable mechanisms named SMUP and SMNP. In SMUP, we use randomized uniform pricing, and prove that it has (1 + log h)-guarantee on revenue where h is the maximum price of a task. In SMNP, we use randomized non-uniform pricing, and prove that it has (3+3 log h)-guarantee on revenue, slightly worse than SMUP analytically. However our experiments show, SMNP has much less variance than SMUP. For the online setting when tasks arrive over time, we present a truthful online stable mechanism with (2 + 2 log h)-guarantee on revenue.