THE DEVELOPMENT AND APPLICATION OF COMPUTATIONAL MULTI-AGENT MODELS FOR INVESTIGATING THE CULTURAL TRANSMISSION AND CULTURAL EVOLUTION OF HUMPBACK WHALE SONG

Abstract

Three different multi-agent models are presented in this thesis, each with a different goal. The first model investigates the possible role migratory routes may have on song evolution and revolution. The second model investigates what social networks could theoretically facilitate song sharing in a population of whales. The third model implements a formal grammar algorithm in order to investigate how the hierarchal structure of the song may affect song evolution. Finally, the thesis attempts to reconnect the models with their origins and discusses how these models could potentially be adapted for composing music. Through the development of these different models, a number of findings are highlighted. The first model reveals that feeding ground sizes may be key locations where song learning from other population may be facilitated. The second model shows that small world social networks facilitate a high degree of agents converging on a single song, similar to what is observed in wild populations. The final model shows that the ability to recognise hierarchy in a sequence coupled with simple production errors, can lead to songs gradually changing over the course of time, while still retaining their hierarchal structure.