Description

Professor Gillian Hadfield is seeking a highly-qualified postdoctoral scholar to join her team at the Normativity Lab in Baltimore, MD, or Washington, DC, to investigate the foundations of human normativity and how these insights can inform the development of AI systems aligned with human values. The ideal candidate will have a track record in computational modelling that explores the dynamics of AI systems and the development of autonomous AI agents, experience with machine learning, reinforcement learning, and generative AI, and a background in interdisciplinary research. This is a full-time one-year position, with the possibility of extension.About the Normativity Lab

How can we ensure AI systems and agents align with human values and norms? Maintain and enhance the complex cooperative economic, political and social systems humans have built? What will it take to ensure that the AI transformation puts us on the path to improved human well-being and flourishing, and not catastrophe? Existing approaches to alignment, such as RLHF, constitutional AI and social choice methods, focus on eliciting human preferences, aggregating them across multiple, pluralistic values if necessary, and fine-tuning models to satisfy those preferences. In the Normativity Lab we believe these approaches are likely to prove too limited to address the alignment challenge and that the alignment questions will require studying the foundations of human normativity and human normative systems.  We bridge computational modeling, specifically multi-agent reinforcement learning and generative agent simulations, and economic, political, and cultural evolutionary theory to explore the dynamics of normative systems and explore how to build AI systems and agents that have the normative infrastructure and normative competence to do as humans have learned to do:  create stable rule-based groups that can adapt to change while ensuring group well-being.Specific Duties and Responsibilities:

• Project Ownership: Take ownership of research projects, working independently and collaboratively with a diverse team of experts
• Model Development: Develop and refine computational models to simulate and analyze normative behaviors in various contexts.
• Data C...