9 September 2015, Logic Tea, Paul Van Eecke

Abstract

The origins and evolution of language remain a largely unsolved mystery. Certain researchers claim that language is innate, genetically encoded, whereas others argue that it is primarily the outcome of a cultural evolution process. But whatever perspective is taken, the fundamental challenge resides in coming up with concrete models and computational validations that show under which conditions particular aspects of language can emerge and evolve.

In this introductory lecture, I will explain the methodology that is used in our research group for modelling cultural language evolution, namely the language game paradigm (Steels, 2011). This approach investigates which cognitive mechanisms and interaction patterns are needed in a population of artificial agents in order to be able to evolve a shared conceptualisation and language. The mechanisms for evolution are the same as in biological evolution: inheritance, variation and selection. The units of selection are here form-meaning mappings (constructions) and criteria for selection include communicative success, expressive power and cognitive effort. In the first part of this lecture, I will present the methodology as a whole and in the second part, I will focus on the language processing tools that were developped for conducting these experiments, in particular Fluid Construction Grammar (FCG).

For more information, please visit the website http://www.illc.uva.nl/logic_tea/ or contact Thomas Brochhagen (t.s.brochhagen at uva.nl), Johannes Marti (johannes.marti at gmail.com), Masa Mocnik (masa.mocnik at gmail.com) or Julian Schloder (julian.schloeder at gmail.com).