2026 Theses Doctoral

Platform Design under Uncertainty and Strategic Behavior

Yu, Chenkai

Digital platforms such as ridesharing and delivery systems make operational decisions under network spillovers, strategic behavior, and incomplete information. This dissertation studies three frictions: spatial demand imbalance, strategic supply withholding, and missing ground truth in customer support.

Chapter 2 studies welfare-improving pricing in ridesharing. Origin-based surge pricing clears local markets but ignores destination conditions, inducing inefficient driver flows and distorted incentives. I develop an origin-destination pricing mechanism with weekly iterative updates that use only observable quantities from prior weeks. In simulations based on Chicago morning-rush data, the mechanism achieves over 95% of weekly welfare optimum for most weeks, compared with below 90% under naive origin-based pricing, and remains effective during large demand shocks.

Chapter 3 studies strategic supply withholding in ridesharing. I analyze a continuous-time model in which drivers coordinate online and offline behavior, generating price and supply cycles. I characterize conditions under which these cycles form a Nash equilibrium, show that in sufficiently dense markets they can reduce total driver payoffs, and derive price-floor policies that prevent stable cycles under those conditions.

Chapter 4 studies fraud detection in two-sided marketplaces when ground truth is unavailable. I develop a belief propagation method that jointly infers strategic customers and drivers from sparse interaction histories. On marketplace data with 164 million orders, the method outperforms naive score-based detection over a wide range of thresholds. A composite score that combines structural inference with historical averages can further improve performance.

Together, these chapters show how incentive-aware optimization and structure-aware inference can improve efficiency, reliability, and trust in strategic marketplaces.