When a principal needs to recruit agents to carry out a task incrementally — each agent knowing more about the task after partial completion — standard procurement auctions fail to elicit truthful bids. This project develops theoretically grounded auction mechanisms for the gradual procurement of strategic service providers.
Work includes characterising optimal mechanisms under various information structures, establishing dominant-strategy and Bayesian incentive compatibility conditions, and extending results to multi-agent settings with complementarities. Recent work addresses robust information design when principals face model uncertainty about agents’ beliefs.
Publications span the Journal of Artificial Intelligence Research (JAIR) and workshop venues at IJCAI and AAMAS.