More Than Words: Computational Models of Emergence and Evolution of Symbolic Communication.

The study of symbolic communication is a key research area in both the social and natural sciences. However, little has been done in order to bridge these scientific domains, so an unfortunate gulf between them still persists. Even less has been done in the field of computational sociology, in which most research using agent-based models has disregarded the importance of symbolic communication. It is this lacuna that the thesis addresses. In the thesis, it is claimed that the type of emergent properties that are inherent to social phenomena are likely to result from the unique fact that the participating entities are symbolic agents. It is proposed that symbolic communication is a threshold phenomenon that emerges in the intersections among human cognition, social interactions and human biology. A theoretical framework with which to clarify this connection is also presented. In order to test in silico some hypotheses derived from this theoretical framework, the analysis relies upon two agent-based models. Different simulation methods and techniques were used, such as reinforcement learning algorithms, genetic algorithms, graph theory, and evolutionary game theory. To investigate the simulation results, multivariate analysis techniques, social network analysis and differential equations were used. The first agent-based model was developed to study the properties of an emergent communication system, in which groups of 'speechless' agents create local lexicons and compete with each other to spread them throughout the whole population. The model results indicate that a common lexicon can emerge on the condition that a group of agents develops a communicative strategy that favours their mutual understanding and allows them to reach more recipients for their utterances. An analysis of the agents' social networks reveals that strong mutual relations among agents from the same group, high 'immunity' to external influence and high capability of speaking to agents from different groups play a fundamental role in the process of spreading lexicons. The second agent-based model was built to study the pre-linguistic stage of cooperation among individuals required for the emergence of symbolic communication. In this model, agents reproduce sexually, males and females differ in their reproductive costs and they play the iterated prisoner's dilemma. The model results show that, when male reproductive costs are less than female reproductive costs, males cooperate with females even when females do not reciprocate. This non-reciprocal cooperation, in turn, produces a sustained population growth, because females can reproduce faster despite their high reproductive costs. Finally, a mathematical model of cumulative cultural evolution is used to investigate different patterns of population dynamics, and it is demonstrated that the artificial societies in which non-reciprocal cooperation emerges are able to sustain more complex cultural artefacts, such as communicative symbols. Linking computational sociology to appropriate theories of language evolution, communication, evolutionary biology and cognitive research, the thesis provides concept- ually grounded mechanisms to explain the emergence and evolution of symbolic communication. In so doing, the thesis contributes both substantively and methodologically to academic work on computational sociology, as well as agent-based models of symbolic communication