Adaptivity: A Quality Goal for Agent-Oriented Models?

Abstract: If software systems are to behave more intelligently, they will need to be able to adapt to the changing external world. The agent-oriented paradigm offers the potential to build adaptive systems. This paper explores the nature of adaptivity in multi-agent systems by exploring how to represent adaptivity in agent-oriented models. Rather than focussing on mechanisms for individual agents to be highly adaptive, we advocate regarding adaptivity as a quality goal of an overall multi-agent system. Discussing agent-oriented models should help guide where design tradeoffs need to be made. Learning can be approached in a similar way to adaptivity, where learning is viewed as an attribute of the system rather than a fundamental property of individual agents. Examples are given from use of automated teller machines at mobile sites and intelligent information agents