Autonomy Reconsidered: Towards Developing Multi-agent Systems

An agent’s autonomy can be viewed as the set of physically and computationally grounded algorithms that can be performed by the agent. This view leads to two useful notions related to autonomy: behavior potential and success potential, which can be used to measure of how well an agent fulfills its potential, call fulfillment. Fulfillment and success potential induce partial and total orderings of possible agent algorithms, leading to algorithm-based, capability-centered definitions of levels of autonomy that complement common uses of this phrase. Because the success potential of a multi-agent system can exceed the success potentials of individual agents through synergy effects, the fulfillment of an individual can be augmented through interactions with others, though it can possibly also interfere in the fulfillment of the other agents. Interaction algorithms thus enable multiple agents to coordinate, communicate, or exchange information; these algorithms enable and constrain tradeoffs between augmenting and diminishing other agents. Short case studies are presented to illustrate how the algorithm-based definitions can be used to understand existing systems.