Computing and Information Systems - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Promoting cooperation in multi-player evolutionary games
    Chiong, Raymond ( 2012)
    Understanding how cooperation can be promoted and maintained in a population of individuals when selfish actions are clearly favoured is a long-standing scientific endeavour. A simple yet powerful framework that has become a standard tool for investigating this fundamental problem is evolutionary game theory (EGT). Within the general EGT framework, social dilemma models in the form of two players and two strategies (2 x 2) games are typically used to study cooperative phenomena. A great deal of research has been carried out on such 2 x 2 games with pair-wise interactions. However, the fact that numerous cooperative enterprises involve the actions of more than two players implies the limitations of these 2 x 2 game models to some extent. Multi-player games have important real-world implications, but compared to two-player games significantly less attention has been devoted to games with multi-player interactions. Consequently, many important issues in the multi-player context have been overlooked. The aim of this research is therefore to perform a systematic and comprehensive study of mechanisms that promote cooperation in multi-player games. To this end, we investigate three related issues based on EGT and computational simulations. First, we study the role of spatial, iterated interactions in two multi-player evolutionary games, namely the N-player Iterated Prisoner's Dilemma (N-IPD) and N-player Iterated Snowdrift (N-ISD) games. Different model parameters, which include the cost-to-benefit ratio, the size of groups, the number of repeated encounters, and the interaction topology, have been explored through extensive numerical simulations by using a genetic algorithm to evolve game strategies. Our results reveal that, while the introduction of iterated interactions does promote high levels of cooperative behaviour across a wide range of parameter settings, the cost-to-benefit ratio and group size are important factors in determining the appropriate length of beneficial repeated interactions. It is generally accepted that more repeated encounters would promote higher levels of cooperation. Our findings, however, suggest that increasing the number of iterated interactions may have a detrimental effect in multi-player spatial games when the cost-to-benefit ratio and group size are small. Next, we consider the effects of migration on the evolution of cooperation, again using the spatial N-IPD and N-ISD games as the basis of our investigation. Here, players are allowed to move to new locations based on the migration schemes in place. Key parameters examined include the cost-to-benefit ratio, group size, and movement range (i.e., how far a player can move). Our results show that cooperation can still be maintained in a population of mobile individuals. Significantly, mobility can enhance system-wide cooperation levels when compared to the "never-move" case, if the movement of players is kept to within a limited range. To the best of our knowledge, this is the first time mobility issues have been studied in the context of iterated N-player games. Finally, the inherent cognitive complexity of spatial, iterated interactions has led us to propose an alternative stigmergy-based approach, a form of indirect reinforcement learning, that works via the use of social information. Inspired by recent research suggesting that social information use is a widespread phenomenon not only in human society but also among the non-human animals, and that it may trigger cultural evolution, the proposed stigmergy-based model relies on external cues in the environment rather than individual-specific information for the decision-making process. Detailed simulations across different cost-to-benefit ratios and group sizes based on the "one-off" N-player PD and SD games - without iterated interactions or spatial extensions - provide strong supporting evidence that stigmergic interactions enable the promotion of cooperation across a wide range of conditions. This is despite the fact that the make-up of the interacting groups is continually changing. The new results and outcomes of this research describing the promotion and maintenance of cooperation in multi-player games may lead to an improved understanding of various phenomena found in natural and social systems, especially when efficient collective actions matter.