On the organisation of agent experience: Scaling up social cognition

Abstract

This paper introduces "micro-scalability" as a novel design objective for social reasoning architectures operating in open multiagent systems. Microscalability is based on the idea that social reasoning algorithms should be devised in a way that allows for social complexity reduction, and that this can be achieved by operationalising principles of interactionist sociology. We first present a formal model of InFFrA agents called m InFFrA that utilises two corner-stones of micro-scalability, the principles of social abstraction and transient social optimality. Then, we exemplify the usefulness of these concepts by presenting experimental results with a novel opponent classification heuristic ADHOC that has been developed using the InFFrA social reasoning architecture. These results prove that micro-scalability deserves further investigation as a useful aspect of socionic research. © Springer- Verlag Berlin Heidelberg 2005.