Evolutionary game theory provides tools to study complex, repeated interactions in large populations. Important applications are cooperative games such as common-pool resource problems, coalitional games, or matching markets. A new strand of behavioural models with ``Learning by Trial and Error" (Peyton Young, 2009) provides the basis for models that are particularly suited to study such games using simulations and software applications. The resulting dynamics of such behaviour - which is ``completely uncoupled" from others - are surprising: the long-run stable outcomes turn out to be appealing solutions to the games in many applications. This proves that very simple learning dynamics may be sufficient for agents to learn to cooperate and play socially desirable outcomes. While the first part of the talk discusses the background and outlines the theoretical results of existing work, the second part is about the design of learning dynamics, software agents, and computer matching mechanisms.
About the Speaker: Heinrich Harald Nax holds a D.Phil. in Economics from the University of Oxford. His research focusses on cooperation, the evolution of social outcomes and learning in games.
