Evolution of communication with a spatialized genetic algorithm

Abstract

We extend previous work by modeling evolution of communication using a spatialized genetic algorithm which recombines strategies purely locally.Here cellular automata are used as a spatialized environment in which individuals gain points by feeding from drifting food sources and are 'harmed' if they fail to hide from migrating predators. Our individuals are capable of making one of two arbitrary sounds, heard only locally by their immediate neighbors. They can respond to sounds from their neighbors by opening their mouths or by hiding. By opening their mouths in the presence of food they maximize gains; by hiding when a predator is present they minimize losses. We consider the result a 'natural' template for benefits from communication; unlike a range of other studies, it is here only the recipient of communicated information that immediately benefits.A community of'perfect communicators' could be expected to make a particular sound when successfully feeding, responding to that same sound from their neighbors by opening their mouths. They could be expected to make a different sound when 'hurt' and respond to that second sound from their neighbors by hiding.Suppose one starts from a small set of 'Adam and Eve' strategies randomized across a cellular automata array, and uses a genetic algorithm which operates purely locally by cross-breeding strategies with their most successful neighbors. Can one, in such an environment, expect evolution of local communities of 'perfect communicators'? With some important qualifications, the answer is 'yes'.

A Praxical Solution of the Symbol Grounding Problem

“…The AM²s inhabiting a given niche interact with the environment in two ways: they feed and they can hide themselves to avoid the attacks of three kinds of predators -α , β and γ -which put them in three internal states (the reader will find more details about this scenario in (Grim et al, 2001)). Suppose the AM²s involved in this scenario engage in a kind of adaptive language game, such as the guess game (Steels, 2005).…”

Information and Meaning: Use-Based Models in Arrays of Neural Nets

“…Within neural net models in particular, symmetrical rewards for 'successful communication' characterize Werner and Dyer (1991), Saunders and Pollack (1996), and Hutchins and Hazelhurst (1995). 4 In earlier studies we have considered similar environments in which strategy change was by pure imitation of successful neighbours (Grim et al, 2000) or by genetic algorithm crossover with successful neighbours (Grim et al, 2001). In these studies we found similar results, indicating that the forces driving emergence of information transfer are inherent in the environmental structure rather than in the specific mode of strategy change.…”

“…In this sense, language is like a cloud of birds which attains and keeps its coherence based on individual rules enacted by each bird. (p. 2) Another essential aspect of the model offered here is spatialization, carried over from our own previous work in both cooperation and simpler models for communication (Grim, 1995(Grim, , 1996Grim et al, 1998Grim et al, , 2000Grim et al, , 2001. Our community is modeled as a two-dimensional cellular automata array.…”

Making meaning happen

“…In later studies, we explored (1) behaviours that signal an alarm when a predator is present, whether or not the individual is harmed, and (2) models in which twice as many predators as food sources are used. The latter is the environment emphasized here-see also Grim et al (2001). 8.…”