Hortatory operations and the Colloquium: Modeling a human-machine system using knowledge engineering techniques.

Abstract

People who operate complex automated systems typically work in a monitoring or supervisory role rather than one involving direct manipulation and command. Research models of control centers generally assume that operators have extensive real-time information about the system and significant levels of influence over it. However, such models have limited applicability in conditions where restricted influence and insufficient information are commonplace. Furthermore, they are less pertinent when jurisdiction over the system is shared among a network of heterogeneous organizations. This study addresses these two distinctive limitations in particular. The term hortatory operations is proposed as a paradigm to describe circumstances where administrators rely heavily on advice and encouragement, rather than direct control, to influence system behavior. In many cases, they may also be working with inadequate or ill-suited information about activity in the system. The paradigm extends the purview of the supervisory operations model, which is frequently used in the human factors literature. Hortatory operations accepts that communications among system components are somewhat haphazard and ad hoc, rather than being dependent upon extensive and well-understood command-and-response protocols. The paradigm is particularly helpful in modeling work situations that involve groups of loosely-cooperating organizations. The study also develops a comprehensive model--termed the Colloquium--of this environment. The Colloquium is embodied in the 'network of conversations' among the agent organizations. Information about the system under administration is shared among Colloquium agents using activity or status reports. Several examples of such environments--in transportation, financial crimes enforcement, and computer network administration--are discussed. The study includes an experimental investigation of the effects of sharing information between Colloquium agents working in freeway traffic management. The responsiveness of a decision-aiding tool, which relies solely upon numeric methods to identify exceptions in traffic conditions, is first measured. It is then supplemented with human operators' routine practices and maxims, which are encoded as knowledge-based heuristics. These are used to introduce into the analysis such qualitative information as can be obtained from other Colloquium agents. The result is a significant improvement in the responsiveness of the integrated system.