According to game theory, dominant strategy of Prisoner’s Dilemma game is defecting. Since online trading between two strangers falls in the realm of a Prisoner’s Dilemma, failure should be the only logical conclusion of such electronic commerce situation. Traders are less likely to cooperate since they are not required to deal with the same buyer again given the enormous population of online traders. Thus one could argue that markets like eBay should never exist. Then what is the reason behind resounding success of such electronic markets? The answer lies in the reputation system that they established. Google page rankings and reader reviews on various sites like Amazon.com and Epinions.com are some other examples of the reputation system pointing towards the enormous influence that they wield in our everyday life. This research thus attempts to analyze importance of reputation system in an online trading using Agent Based Simulation. The effect of reputation system on the sustainability of mutual cooperation between online traders is analyzed by taking into account key factors like level of gullibility of online traders and the weight of influence given to their past behavior. Results indicate that reputation system is certainly capable of sustaining cooperation among online traders. However, weightage given to the past behavior of agents should be decided only after taking into account the level of gullibility of agents. Use of agent based modeling for studying dynamics of online trading is a relatively new concept. Sustaining cooperation among online traders is often perceived to be difficult task. The reputation system proposed in this research is expected to make contribution towards this goal.

Games have been investigated as computationally complex problems since the inception of artificial intelligence in the 1950’s. Originally, search-based techniques were applied to create a competent (and sometimes even expert) game player. The search-based techniques, such as game trees, made use of human-defined knowledge to evaluate the current game state and recommend the best move to make next. Recent research has shown that neural networks can be evolved as game state evaluators, thereby removing the human intelligence factor completely. This study builds on the initial research that made use of evolutionary programming to evolve neural networks in the game learning domain. Particle Swarm Optimisation (PSO) is applied inside a coevolutionary training environment to evolve the weights of the neural network. The training technique is applied to both the zero sum and non-zero sum game domains, with specific application to Tic-Tac-Toe, Checkers and the Iterated Prisoners Dilemma (IPD). The influence of the various PSO parameters on playing performance are experimentally examined, and the overall performance of three different neighbourhood information sharing structures compared. A new coevolutionary scoring scheme and particle dispersement operator are defined, inspired by Formula One Grand Prix racing. Finally, the PSO is applied in three novel ways to evolve strategies for the IPD – the first application of its kind in the PSO field. The PSO-based coevolutionary learning technique described and examined in this study shows promise in evolving intelligent evaluators for the aforementioned games, and further study will be conducted to analyse its scalability to larger search spaces and games of varying complexity. / Dissertation (MSc)--University of Pretoria, 2005. / Computer Science / unrestricted

L’altruisme réciproque, le mécanisme le plus vraisemblable expliquant l’existence de la coopération entre individus non-apparentés, peut être modélisé par le Dilemme du Prisonnier. Ce jeu prédit que la coopération devrait évoluer lorsque les joueurs prévoient d’interagir ensemble à maintes reprises et adoptent des stratégies conditionnelles telles que «Tit-For-Tat» ou Pavlov. Bien que la coopération soit à la source de toutes sociétés humaines, celle-ci est rarement observée chez les animaux. Une explication plausible serait que ces derniers sont plus impulsifs que les humains. Plusieurs études ayant évalué les effets de l’impulsivité sur la coopération ont en effet trouvé un impact négatif du phénomène de « discounting » sur la réciprocité. Néanmoins, l’impulsivité n’est pas un concept unitaire et le rôle de l’impulsivité motrice, une autre facette de l’impulsivité, reste inexploré, alors qu’elle pourrait également restreindre la coopération en altérant la capacité des individus à ajuster de manière flexible leur comportement face aux décisions prises par leur partenaire. En effet, l’impulsivité motrice se définit comme étant l’incapacité à inhiber un comportement qui n’est plus approprié suite à un changement de situation et est donc contreproductif (Broos et al., 2012; MacLean et al., 2014). Pour résoudre cette hypothèse, nous avons mené une expérience avec des diamants mandarins (Taenyopigia guttata) que nous avons appariés en fonction de leur niveau d’impulsivité motrice, puis nous les avons fait jouer dans un Dilemme du Prisonnier Alterné. Tel qu’attendu, nous avons trouvé que la coopération mutuelle survenait plus fréquemment entre les partenaires autocontrôlés que les paires d’individus impulsifs, ce qui serait dû à une différence entre les stratégies employées par les deux types d’individus. Plus précisément, les individus autocontrôlés utilisaient une stratégie « Generous TFT », tel que prédit par la théorie, alors que les oiseaux impulsifs choisissaient de coopérer avec une probabilité fixe, laquelle était indépendante de la décision précédemment prise par le partenaire. Si l’incapacité des individus impulsifs à utiliser des stratégies réactionnelles est due à une capacité de la mémoire de travail réduite, nos résultats pourraient alors contribuer à expliquer les différences interspécifiques qui existent au niveau des comportements coopératifs. / Reciprocal altruism, the most probable mechanism for cooperation among unrelated individuals, can be modelled as a Prisoner’s Dilemma. This game predicts that cooperation should evolve whenever the players, who expect to interact repeatedly, adopt conditional strategies. Yet, experimental data suggest that reciprocity would be rare in animal societies, maybe because animals, compared to humans, are very impulsive. Several studies examining the effect of impulsiveness on cooperation have indeed found a negative impact of temporal discounting. On the other hand, the role of impulsive action, another facet of impulsiveness, remains unexplored, though it could also impede cooperation by affecting the capacity of individuals to flexibly adjust their behaviour to their partner’s decision. To address this hypothesis, we conducted an experiment with zebra finches (Taenyopigia guttata) that were paired assortatively with respect to their level of impulsive action and then played an Alternating Prisoner’s Dilemma. As anticipated, we found that mutual cooperation occurred more frequently between self-controlled partners than between impulsive ones, a difference that was caused by differences in the strategy used by both types of individuals. Specifically, self-controlled individuals used a Generous TFT strategy, as predicted by theory, whereas impulsive birds chose to cooperate with a fixed probability, which was independent of their partner’s previous decision. If the inability of impulsive individuals to use reactive strategies are due to their reduced working memory capacity, our findings might contribute to explaining interspecific differences in cooperative behaviour.