ID No.: 2007HZ67003
NAME OF THE STUDENT: Ronan Sandford
DISSERTATION TITLE: Multi-Agent Models of Language Acquisition and Usage. Focus on Exophoric Use of Demonstratives
ABSTRACT
In this work, after reviewing ordinary language philosophy, we describe an approach in opposition to symbolic computing enabling cognitive theories to be tested in multi-agent systems. Because such systems enable the simulation of social factors, they can provide a bed-test for social cognitive theory such as provided by (Tomasello, 1999) for example. Furthermore, in multi-agent systems using language, words instead of being processed as arbitrary symbols are used to communicate. Following ordinary language philosophy insight that meaning are related to usage, such multi-agent system could be said to exhibit understanding. In order to show more clearly the advantage of multi-agent models for cognitive science and linguistic study in particular, we have built models to generate the use of demonstrative not present in some early works (Steels and Kaplan, 1999). The models, while quite modest have been able to generate the exophoric usage of demonstrative and serve as an example to implement more complex cognitive processes. Cognitive theories will gain a lot in building their model by considering the importance of embodiment and social environment. Multi-agent models enable such consideration and provide an interesting testbed for computational experiment. Whether these systems will enable the emergence of understanding and intention could not have been resolved in this work but by adopting ordinary language philosophy insight that meaning is related to usage, we can hope that such system using word in relation to complex society settings could be considered as intentional.
