Introduction

Agent orientation refers to a software development perspective which has evolved in the past 25 years in the fields of computational agents and multiagent systems. The basic notion underlying this perspective is that of a computational agent, that is, an entity whose behaviour deserves to be called flexible, social, and autonomous. As an autonomous entity, an agent possesses action choice and is at least to some extent capable of deciding and acting under self-control. Through its emphasis on autonomy, agent orientation significantly differs from traditional engineering perspectives such as structure orientation or object orientation. These perspectives are targeted on the development of systems whose behaviour is fully determined and controlled by external units (e.g., by a programmer at design time and/or a user at run time), and thus inherently fail to capture the notion of autonomy.

To date autonomy still is a poorly understood property of computational systems, both in theoretical and practical terms, and among all properties usually associated with agent orientation it is this property being most controversially discussed. On the one hand, it is argued that there is a broad range of applications in complex domains such as e/m-commerce, ubiquitous computing, and supply chain management which can hardly be realized without taking autonomy as a key ingredient, and that it is first of all agent autonomy which enables the decisive features of agent-oriented software, namely robustness, flexibility and the emergence of novel solutions of problems at run time. On the other hand, it is argued seemingly convincingly that autonomy mainly is a source of undesirable and chaotic system behaviour. Obviously, without a clarifcation of these two positions, it is unlikely that agent orientation and agent-oriented systems (having "autonomy" as a real property and not just as a catchy label) become broadly accepted in real-world, industrial and commercial applications.

The purpose of this half-day workshop is to provide a unique forum for discussing the potential and the risks of computational autonomy, and for presenting solutions to the various issues raised by computational autonomy.