Modelling human fairness in cooperative games
: a goal programming approach

  • Nerda Zura Zaibidi

Student thesis: Doctoral Thesis


The issues of rationality in human behavior and fairness in cooperation have gained interest in various economic studies. In many prescriptive models of games, rationality of human decision makers implicitly assumes exchange-ability. This means that real people are assumed to adopt the beliefs of a player as expressed in the game when placed in the shoes of that particular player. However, it is a well debated topic in the literature that this modeling assumption is not in accordance to what behavioral economists have observed in some games played with real human subjects. Even when assuming the role of the same player in the game, different people think differently about the fairness of a particular outcome. People also view fairness as an essential ingredient of their decision making processes in games on cooperation.
The aim of this research is to develop a new modeling approach to decision making in games on cooperation in which fairness is an important consideration. The satisficing and egilatarian philosophies on which weighted and Chebyshev Goal Programming (GP) rely, seem to offer an adequate and natural way for modeling human decision processes in at least the single-shot games of coordination that are investigated in this work. The solutions returned by the proposed GP approach aim to strike the right balance on several dimensions of conflicting goals that are set by players themselves and that arise in the mental models these players have of other relevant players.
Fairness concerns are important in the well-known Ultimatum and Dictator games. These games are modeled using a Chebyshev GP approach. Parallels are drawn between the approach and concepts of human decision making from the field of cognitive neuroscience and psychology. The Chebyshev GP is the universal mechanism in the model for players to decide how fair outcomes are to be identified, but allows individuals to differ in their belief which outcomes are fair. Computer simulations of these GP models,testing a large number of Ultimatum, Dictator and Double Blind games, lead to distributions of proposals made and accepted that correspond reasonably well with experimental findings.
In our study of some simple but classic cooperative games, a fairness model is developed using weighted GP by taking into account players' aspiration goals and preferences in terms of profit and fairness concerns explicitly.The model offers a framework by which players can make decisions considering the different viewpoints of the potential partners of a coalition. The application of this framework to the Drug and Land games shows that the inclusion of fairness in the game produces solutions that may sometimes deviate significantly from solutions obtained from standard methods of cooperative game theory.
Date of AwardMar 2012
Original languageEnglish
SupervisorPatrick Beullens (Supervisor) & Dylan Jones (Supervisor)

Cite this