Behavioural Game Theory and Interacting Agents

We study boundedly rational and heuristic decision making, planning, and learning dynamics, individually and collectively.

We focus on situations with interacting and heterogeneous agents at multiple levels of socio-economic organization and governance, in simple and complex environments with risk, ambiguity and potentially large stakes, with collaboration, coordination, and competitive aspects.

To this end, we use normative theory, algorithm design, numerical simulations, and behavioural experiments, in collaboration with strong partners from the behavioural, formal, and normative sciences, including experimental economics, social psychology, machine learning, social choice theory, and moral philosophy.

If necessary, we challenge disciplinary paradigms such as expected utility maximization, and replace them by alternative principles such as precautionary, ambiguity-averse, multi-criteria, or aspiration-based decision-making.

We sometimes use additional concepts from complexity science such as complex networks, nonlinear dynamics, information theory, or agent-based models.

A major motivation for our research is that mitigating global risks such as climate change requires cooperation on an unprecedented scale, from local communities to the international sphere. Significant progress has been made in describing how individuals behave and interact: while game theory describes fundamental problems such as free-riding, networks can represent the complex structure of the global community, and behavioral economics and social psychology analyze how agents learn from experience and social interactions. Still, modeling decision making across different scales remains an interdisciplinary challenge. How cooperation may or may not emerge is an open question.

(This working group is part of PIK's Research Department 4, Complexity Science, and partially continues some research directions started in the former cross-institutional FutureLab on Game Theory and Networks of Interacting Agents (GaNe) between PIK and MCC)

Topics

Our current research focusses on

• Mechanisms for cooperation: coalition formation, conditional commitments, group decision methods, evolutionarily stable strategies
• Mathematical formalization and operationalization of concepts such as agency and responsibility (as part of the mathematics "Exzellenzcluster" MATH+)
• Relationships between game-theoretical and physical equilibrium concepts, human and machine learning, and other dynamical behaviours in games
• Design of boundedly-rational / non-maximization-based decision algorithms for stochastic sequential decision making under various forms of uncertainty

Examples of possible thesis topics

• Solving a certain set of partial differential equations representing a model of farsighted strategy updating in simple games (mathematics)
• Design and implementation of a decentralized and confidential communication structure for a social app for collaboration via conditional commitments (computer science)