Emergence of Representational Structures in Virtual Agents

Abstract

Autonomy and human-like personality in general is a crucial property of a virtual agent. This paper introduces a framework of self-organised Peircean semiotic processes, which is used to demonstrate the emergence of grounded representational structures in a virtual agent interacting with the environment.

A completed semiosis consists of the three inferential procedures: abduction, deduction and induction, which drive the agent’s logical argumentation. In the abductive phase, the agent observes the environment and describes the nature of a surprising phenomenon on the basis of its anticipations. An analogy between the surprising phenomenon and the agent’s anticipations is attempted, in order to indicate a possible direction of a hypothesis explaining the phenomenon, and the formulation of this hypothesis takes place.