Quantifying Turing: a systems approach to quantitatively assessing the degree of autonomy of any system

Abstract

This paper describes a method by which the degree of autonomy of a system can be quantified in a manner that allows comparison between systems. The methodology revisits, refines, and extends the contextual autonomous capability (CAC) model proposed by the National Institute of Science and Technology (NIST) by defining three orthogonal system metrics against which the performance of a system may be assessed. During the development of this model, it was recognized that there existed two different but coupled domains of autonomy — the Executive Autonomy describing the degree of independence of a system during the execution of the mission; and the Developmental Autonomy describing the degree of independence of the system during preparation for the mission. The resulting methodology is explicitly developed to be system agnostic such that it could be applied to humans as well as computerized systems. As such, it provides a means of quantifiably comparing the performance of any two systems — including human and computer — that are performing comparable sets of missions. The proposed model is called the system-agnostic quantification of autonomy levels (SQuAL) model.