How situated is your agent? A cognitive perspective

Abstract

Software agents are situated in an environment with whichthey interact reactively or in a goal-directed fashion. Generally, suchenvironments do not assume a structure, hence are deemed to be unpredictable. Recent approaches adopt an environment model where artifactsform the building blocks which represent functional components that anagent can exploit for reaching its goals. It has been argued that softwareagents can improve/amend their capabilities at run time through theuse of (new) artifacts as possible means. We argue that such a run timeadaptation by the agents can be realized by creating an appropriate relationship between agent reasoning and the functionality of the artifacts.We have coined the term extrospection to refer to the act of an agentreasoning about the tools. In this paper, we first identify the features ofextrospection, then, we extend the belief, desire, intention (BDI) agentdeliberation cycle to encompass extrospection.