Secret Agents: Synthetic Approaches to Language Evolution

Extensive work in the field of synthetic ethology demonstrates that multi-agent simulation is a useful tool for exploring a multitude of population-based phenomena, including economics, epidemiology, and language evolution. From a structured canon of local interactions and low-level rules, complex behaviors can emerge in a wide range of situations. This notion of ‘emergence’ is central to evolutionary studies, biological or otherwise. This paper addresses the current models of language evolution as put forward by contemporary researchers and attempts to codify the rules of language modeling. It seeks to modify the work done by Bart de Boer [7, 6, 8, 10] in self-organizing vowel systems by modifying certain principles of his simulation. With a few algorithmic changes, I hope to isolate the clustering mechanism at the heart of the simulation, so that it may be used in modeling self-organizing systems other than vowel inventories. Chapter 1 provides an introduction to the study of language evolution, then outlines the current research in the field and synthesizes the central principles of synthetic ethology as demonstrated by these systems. Chapter 2 describes the implementational details of reproducing de Boer’s simulation and discusses the results. Chapter 3 explains the simplified model in terms of algorithmic changes and organizational consequences. Chapter 4 discusses what can be learned from these experiments.