Machine Learning, Incentives, and Markets

Program Description

Decision-making in modern economic and financial environments is increasingly carried out by machine learning and AI agents. Prominent examples include critical applications such as stock trading, market making, and auction bidding. At the same time, the widespread deployment of machine learning systems introduces new challenges: strategic, self-interested agents may manipulate their data to influence system outcomes in their favor. Notable examples include adversarial perturbations of input data. As a result, understanding and designing modern economic and machine learning systems requires an interdisciplinary approach that integrates game theory, machine learning, and economics. 

The P1 program, Machine Learning, Incentives, and Markets, aims to establish an interdisciplinary research initiative in Denmark that brings together expertise in algorithmic game theory, theoretical machine learning, and economics, and build a diverse and cohesive research community. By addressing fundamental challenges at the intersection of the three fields, the program will explore the following research areas of (1) machine-learning-based decision-making in strategic environments, (2) incentive-aware machine learning, (3) market design with adaptive and data-driven agents, and (4) data-driven mechanism design.

The program will emphasize both the identification and rigorous mathematical analysis of fundamental challenges within these research areas, as well as practical applications, with a particular focus on markets and energy systems. Through this dual emphasis, the program seeks to create new research directions and to strengthen Denmark’s position in the study of multi-agent AI systems. 

Structured activities and workshops will be organized focusing on:

1. Incentive-Aware Machine Learning: Integrating learning algorithms with mechanism design and game theory to ensure robustness and efficiency under strategic behavior
2. Market Design under Adaptive Agents: Understanding convergence, stability, and welfare in markets where agents learn and adapt over time
3. ML-Guided Decision Making: Applying statistical, online and reinforcement learning to sequential economic decision problems such as auctions, trading, and market making

Together, these activities will support interdisciplinary collaboration, connect theory with practice, and strengthen the Danish research ecosystem in incentive-aware AI and market design.