The concept of autonomy applied to computational agents refers to the ability of an agent to act without the direct intervention of human users in the selection and satisfaction of goals. Actual implementations of mechanisms to enable agents to display autonomy are, however, at an early stage of development and much remains to be done to fully explicate the issues involved in the development of such mechanisms. Motivation has been used by several researchers as such an enabler of autonomy in agents. In this paper we describe a motivational taxonomy, the M-3 Taxonomy, comprising domain, social and constraint motivations, which we argue is a sufficient range of motivations to enable autonomy in all the main aspects of agent activity. Underlying this taxonomy is a motivational model that describes how motivation can be used to bias an agent's activities towards that which is important in a given context, and also how motivational influence can be dynamically altered through the use of motivational cues, that are features in the environment that signify important situations to an agent.
|Title of host publication||Agents and Computational Autonomy|
|Subtitle of host publication||Potential, Risks, and Solutions|
|Editors||M Nickles, M Rovatsos, G Weiss|
|Place of Publication||Berlin, Geramny|
|Number of pages||13|
|Publication status||Published - 2004|
|Name||Lecture Notes in Computer Science|
Munroe, S., & Luck, M. (2004). Agent autonomy through the M-3 motivational taxonomy. In M. Nickles, M. Rovatsos, & G. Weiss (Eds.), Agents and Computational Autonomy: Potential, Risks, and Solutions (pp. 55-67). (Lecture Notes in Computer Science; Vol. 2969). Springer.