ROBERT SUGDEN

Professor of Economics
University of East Anglia, Norwich, UK
 


Excerpt from Robert Sugden's interview: 3. W
hat is the proper role of game theory in relation to other disciplines? 

As will be obvious from the examples I have given, I am convinced that game theory can play an important role in the analysis of problems across the social sciences, and in moral philosophy and philosophy of language (which was the starting-point for Lewis’s work).  But I have tried to avoid using the word ‘application’, with its suggestion that game theory is a self-contained body of ‘pure’ theory, available for other disciplines to apply, but not dependent on those other disciplines for corroboration.  I believe that one of the biggest obstacles to progress in game theory is the perception or illusion that the theory is concerned with a world of a priori propositions, independent of empirical observation and investigation, but somehow accessible to theorists by intuition.

Of course, there need be no illusion in constructing a self-contained formal analysis of propositions that one takes as intuitive or axiomatic; but I believe it is an illusion to imagine that this is a good starting-point for an explanation of strategic interactions between people in the real world.  Theorising about actual human behaviour must surely be grounded in, and responsive to, evidence of how human beings think and act.  Thus, disciplines which study actual behaviour – for example, economics, psychology, anthropology, linguistics, biology – should not be seen merely as areas in which game theory can be applied.  They also provide evidence against which the success of game theory can be judged, and theories of behaviour which game theory might (dare I say it?) apply.     

One of the clearest signs of game theorists’ reluctance to engage with other disciplines on equal terms is the familiar assertion that game theory cannot be brought to bear on a real-world interaction until that interaction has been represented as a formal game, with well-specified payoffs.  As outsiders to game theory are often told: Get the payoffs right first.  This thought leads easily to another: that game theory cannot be tested by using common-sense interpretations of ‘payoff’ to represent real-world interactions as formal games, and by then investigating actual behaviour in those interactions.  If observed  behaviour in the resulting games is found to be contrary to some principle of game-theoretic rationality, game theory reserves the right to deny that the common-sense interpretation of payoff is the correct one.  For example, Ken Binmore famously denies that observations of cooperative choices in experimental Prisoner’s Dilemma games are evidence of the choice of dominated strategies.  He accepts that the material rewards in these experimental games have the structure of the Prisoner’s Dilemma payoffs, but argues that the observed behaviour is evidence that the game, when properly described, is not a Prisoner’s Dilemma after all.

I accept that game theory does not have to assume that payoffs are the same thing as material rewards.  But, I maintain, if it is to be capable of useful application, it must include some theory about what payoffs are.  That theory should allow us to assign payoffs to the games that we use to model real-world interactions before we observe how those games are played.  If game theory took this composite form, it would be capable of being tested against evidence.  I suspect that many game theorists will agree that ‘applications’ of the theory require this, but think that the definition of payoffs is a problem for the disciplines that use the theory, not a problem for the theory itself.  But it is just this attitude that is the obstacle to progress.  Conventional game theory presupposes that each player’s motivations can be represented by numerical payoffs, assigned to the outcomes of strategy profiles, and that the combined behaviour of the players of a game can be explained by using solution concepts that use these payoffs as data.  But what entitles game theory to claim that this strategy of explanation will work?  It is not a self-evident truth that players are motivated by individual payoffs, or that standard solution principles, such as dominance, hold when defined relative to such payoffs.16  To know if this strategy works, we need to be shown that there is a method of assigning payoffs to real-world games such that, when it is used, the solution concepts of game theory lead to successful predictions.  If this strategy doesn’t work, game theory is at fault and needs to be changed.  If progress is to be possible, the first essential is that game theorists recognise that it is their job to make their theories fit the world.  That is what science is all about. 

Read the remaining part of Robert Sugden's interview in Game Theory: 5 Questions, edited by Vincent F. Hendricks and Pelle Guldborg Hansen. The book is released in April 2007 by Automatic Press / VIP.

ISBN 87-991013-4-3
(paperback)
248 pages / $26 / £16

Available on Amazon:

AMAZON   US  CA  UK  DE

Check also for availability:


© 2006, Automatic Press / VIP