Progress towards a development methodology for decision support systems for use in time-critical, highly uncertain, and complex environments

A framework for a design methodology for ecological decision support systems is proposed. Specifically, a methodology is proposed to study and model unfamiliar man-made environments in a manner that supports efforts to create decision support systems. This methodology is demonstrated by the development of a clinical decision support system for neonatal intensive care. The resulting decision support system is shown to significantly improve the diagnostic accuracy of physicians in a simulated intensive care environment. This methodology is applicable to other complex man-made work environments.;Through evolution humans have developed a great ability to work in natural environments that are filled with probabilistic patterns of events. These patterns are what allow us to have meaningful interactions with our environment. However, things are changing people are now finding themselves in artificial man-made environments. These artificial environments are still governed by probabilistic patterns, however, the character of such patterns are not always natural to, and are sometimes hidden from, the people in the environments. Should people evolve new abilities to deal with these new environments? No, what must happen is that these man-made work environments must evolve to take advantage of people's natural ability.;One approach to address this problem is to design decision support systems that match the work environment to the cognitive resources of the people in the environment. The design of these systems is not easy, particularly in the cases were the developer is not an expert in the field. In this case the developer does not understand the probabilistic patterns that are important in the environment. To address these issues a control system theoretic framework is proposed. Appropriate model structures and data collection techniques are discussed. This combination of appropriate model structures and data collection techniques allow for work environments to be modeled in a manner that is useful for the design of decision support systems.;To demonstrate this methodology a clinical decision support system was developed for neonatal intensive care. This system was then empirically evaluated by comparing it to the existing environment. The support system was shown to significantly improve the diagnostic accuracy of physicians. This effect was most consistent with less experienced physicians, every physician in this group performed better with the new support system. Furthermore, the vast majority of the physicians preferred using the new support system.;This dissertation proposes a framework for this modeling methodology and explores several of its components. The ecological decision support systems developed by this framework have the potential to provide significant cost and time benefits in many different fields, for example, clinical decision support systems, executive decision support system, industrial process control. The potential field of applicability is large.