University of Cambridge > Talks.cam > Machine Learning @ CUED > Modeling Behaviour in Economic Games using Game-Theoretic POMDPs

Modeling Behaviour in Economic Games using Game-Theoretic POMDPs

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Zoubin Ghahramani.

Economic games such as Dictator, Ultimatum and Investor-Trustee games have provided an empirical basis for investigating social cooperation and reputation formation. Even in completely anonymous settings, human subjects show rich patterns of behaviour that can be seen in terms of personality concepts. They also behave as if they model these aspects of other players in games, and express different abilities to model their opponents.

We have built a generative model of behaviour in multi-round trust games. The critical features are covered using Type parameters for social utility functions, and finite cognitive hierarchy levels. This is tied together in a game-theoretic Partially Observable Markov Decision Process framework. Particle filtering and Point-based Value Iteration is used to get approximations to the system.

The model is able to capture classes of behaviour in these games. Further it provides more realistic and tractable explanations of behaviour in these games than other equilibrium notions such as Bayes Nash Equilibria.

We also consider the recognition model that is used to fit the parameters for individual interactions. And we’ll look at Neuropsychological correlates of our model parameters with actual behaviour in the Investor-Trustee game.

Finally, l`ll discuss how this model can be extended to capture behaviour in interactions within larger groups such as in Public Goods games.

This talk is part of the Machine Learning @ CUED series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2020 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity