PhD Position in Learning Algorithms for Undeceivable Policies

This doctoral position at Institut Polytechnique de Paris (Télécom SudParis) investigates Personalized Demand-Side Mitigation Strategies (PDSMS) for sustainable behavior, such as reducing pollution. While current state-of-the-art models rely on Random Utility Theory and assume agent honesty, this project addresses scenarios where agents might manipulate regulators to receive favorable signals. The candidate will design reinforcement learning formulations rooted in utility theory and apply mechanism design principles to ensure regulatory robustness. Responsibilities include theoretical modeling, algorithm development, and experimental validation in a collaborative academic lab setting.

• Formulate novel reinforcement learning frameworks based on Random Utility Theory
• Apply mechanism design principles to create policies robust against strategic agent manipulation
• Conduct mathematical modeling and analytical evaluations of regulatory learning processes
• Develop and execute programming tasks to implement and test the proposed learning algorithms
• Design and perform laboratory experiments to validate theoretical findings