Choice: The key for autonomy

Abstract

Agents must decide, i.e., choose a preferred option from among a large set of alternatives, according to the precise context where they are immersed. Such a capability defines to what extent they are autonomous. But, there is no one way of deciding, and the classical mode of taking utility functions as the sole support is not adequate for situations constrained by qualitative features (such as wine selection, or putting together a football team). The BVG agent architecture relies on the use of values (multiple dimensions against which to evaluate a situation) to perform choice among a set of candidate goals. In this paper, we propose that values can also be used to guide the adoption of new goals from other agents. We argue that agents should base their rationalities on choice rather than search. © Springer-Verlag Berlin Heidelberg 2001.