On the Organisation of Agent Experience: Scaling Up Social Cognition

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 2 InFFrA that utilises two cornerstones 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.

• Type: Conference
• Year: 2005
• Where: DFKI
• Authors: Michael Rovatsos, Kai Paetow
• Keywords: DFKI 2005, Social Reasoning, Social Reasoning Algorithms, Social Reasoning Architecture, Social Sciences