Learning in Markets

In the last decade, we have seen a meteoric rise in massive online marketplaces such as Uber, Airbnb, and Amazon Mechanical Turk. In these settings, the platform’s role is to serve as an allocator, whether it be matching riders to drivers, guests to hosts, or workers to jobs. These allocation should satisfy multiple objectives, such as maintaining stability, maximizing revenue, or abiding by resource constraints.

We are interested in the role of pricing in such marketplaces, as intelligent selection of prices can be used to simultaneously achieve market objectives while learning agent preferences and behavior. In certain settings, data-driven methods such as multi-armed bandits can be employed to jointly achieve optimal allocation and pricing. We are exploring the application of such methodologies to better control the dynamics of complex online marketplaces.

Publications

1. Y.E. Erginbas, S. Phade, K. Ramchandran, “Interactive Learning with Pricing for Optimal and Stable Allocations in Markets”, International Conference on Artificial Intelligence and Statistics 2023.

2. Y.E. Erginbas, S. Phade, K. Ramchandran, “Interactive Recommendations for Optimal Allocations in Markets with Constraints”, arXiv 2022.

3. A. Ghosh, A. Sankararaman, K. Ramchandran, T. Javidi, A. Mazumdar, “Decentralized Competing Bandits in Non-Stationary Matching Markets”, arXiv 2022.