Iterated learning : the exemplar-based learning approach
Item statusRestricted Access
MetadataShow full item record
The Iterated Learning Framework (Kirby 2002a, Kirby 2002b, Kirby and Hurford 2002) has been used to explore how human-type languages can emerge and evolve through a process of cultural transmission, given that agents have the necessary biological hardware to learn, use and process it. Different models have different built-in assumptions about the mechanics of language learning, use and processing, and can range from connectionist models (i.e. Tonkes 2002) to rule-induction models and other symbolic models (i.e. Kirby 2002a, Kirby 2002b, Kirby and Hurford 2002) to models in which agents are driven mostly by an attempt to find a compressible grammar (i.e. Teal and Taylor 1999, Brighton and Kirby 2001, Brighton 2002). The research described in this dissertation will deal mostly with an exemplar-based learning model described by Batali (2002), and will go into an in-depth examination of certain aspects of the model and attempt to answer two specific questions: 1) How does the lack of population turnover affect the behaviour of the model, if at all? 2) Does exemplar discouragement and pruning in the model implement a linguistic "bottleneck" with effects on the model similar to those described by Kirby and Hurford (2002, Kirby 2002b). I will also go into some comparisons with other models of the emergence of compositional language, specifically with the symbolic rule-induction model described by Kirby (2002b) and with findings discussed by K. Smith (2002), as well as other interesting things that came up in the course of the research, especially having to do with the effect of the types, complexity, and distribution of meanings given to the agents to discuss on the behaviour of the model.