Verhagen & Kummeneje - Adjustably Autonomous Agents & Decision Making

The Structure of Concern Project compares many theoretical models from many disciplines to the Adizes PAEI model, arguing that they must all be reflecting the same underlying phenomenon. One concern structure model is described below.

Adjustable autonomy refers to the capacity of agents in a multi-agent system to change the degree to which their behavior is constrained by the coalition vs. independently directed. If we sent a fleet of 500 mini-robots to Mars, for example, sometimes we might want them to travel as a ‘herd’ from one region to another, while at other times we might want them to ‘forage’ for scientific samples independently or in groups of two or three. During the travel events we would want them to take a lot of direction from the coalition, but during the foraging events we might basically want them to each do their own thing without straying too far from the dispersed herd. Verhagen and Kummeneje (1999) studied this problem with particular attention to the issue of norms and norm-sharing in multi-agent systems.

Verhagen and Kummeneje identify two perspectives on autonomy that are relevant for agent-based systems programming: abstraction levels and independence. Abstraction levels refer to the degree of control agents have over their own behavior and decision making processes. Independence refers to the agent's degree of independence within a coalition. For autonomy in the first sense of abstraction, the authors refer to a decision-making model drawn from various branches of cognitive science (Verhagen & Smit, 1997; Dennet, 1979; Conte & Castelfranchi, 1995; Werner, 1996). This decision-making model is a very straightforward example of the structure of concern, claiming that decision-making takes place at four separate yet connected levels, listed below in PAEI order:

P – the level of actions
A – the level of plans
E – the level of goals
I – the level of norms

These are the categories within which the degree of control an agent has over its own decisions and actions can be defined.

1. Verhagen, H., & Kummeneje, J. (1999). “Adjustable Autonomy, Delegation and Distribution of Decision Making.” Proceedings of the 1st Internatational Workshop of Central and Eastern Europe on Multi-Agent Systems (CEEMAS 99). (pp. 301-306).
2. Verhagen, H. J. E., & Smit, R. A. (1997). “Multiagent systems as simulation tools for social theory testing.” ICCS and SS Siena, 1997.
3. Dennett, D. C. (1979). Brainstorms: Philosophical Essays on Mind and Psychology (Harvester Studies in Philosophy). Hassocks, Sussex: Harvester Press.
4. Conte, R., & Castelfranchi, C. (1995). Cognitive and social action. London: UCL Press.
5. Werner, E. (1996). “Logical foundations of distributed artificial intelligence.” G. M. P. O'Hare, & N. R. Jennings (Eds.) Foundations of distributed artificial intelligence. New York: Wiley.
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License