A computer model of reactive planning and implementation strategy in program-constrained decision-making

Although organizational programs are an important influence on the behavior of decision makers in complex task environments, their influence is tempered by other cognitive and situational factors which can lead a decision maker to deviate from the prescribed directives of a program. Using empirical data obtained from two separate domains--Nuclear Power Plant (NPP) operations and Mergers & Acquisitions (M&A)--this research has sought to understand and explain the factors that cause actors to conform to, or deviate from, a prescribed program. This disertation describes a theory of program-constrained behavior derived from cognitive task analysis of observed behavior. The theory explains two observed aspects of program-constrained behavior: (1) actions taken during program execution which might be construed as deviating from the constraints of the program, and (2) arguments constructed by the decision maker to justify and defend such actions. The planning process which produces these behaviors is distinguished by a dichotomy between the reactive planning required to achieve, restore or defend desired goals, and the implementation planning required to execute the reactive plan while avoiding sanctions imposed by the organization in response to observed program violations. Analysis of data describing NPP operator behavior led to development of an initial conceptual and computational theory that explained identified instances of deviations in NPP procedure execution. The initial theory then served as the basis for a more focused exploration of behavior related to corporate Mergers and Acquisitions, where argumentation in defense of program-constrained actions became the focus. This led to an extension of the original NPP-derived model, in part to account for behaviors in M&A, but also to refine the explanatory capabilities of the model in the NPP domain. Finally, having begun in the real-time NPP operations environment, and then having applied lessons learned in that effort to the construction of an enhanced model of post-mortem M&A argumentation, this research project 'closed the circle' by extending the model to incorporate post-mortem argumentation in the original domain of NPP operations. The result of the research is a general computational theory capable of explaining and predicting program-deviating and argumentation behavior across domains.