Development Cycle Modeling and Risk Calculation
For every development project, a development agent takes on the responsibility for and assumes the risk of completing that project successfully. This is a daunting task because, at the moment, there is no objective and/or quantitative way to calculate the likelihood of success. A prerequisite for such a calculation is a model, a crisp representation, of an end-to-end development process or cycle. Such models are also unavailable. Nevertheless, projects proceed, and often fail. By some measures, for example, software projects fail more than they succeed. The resources lost to failed projects is an enormous motivation to answer two questions: (1) Is a quantitative development cycle model possible? (2) With such a model, is it possible to calculate the likelihood of project success? The research described in this dissertation has responded to those questions with a Statistical Agent-based Model of Development and Evaluation (SAbMDE). The dissertation describes how the model was motivated, conceived, formulated, and applied to answer the research questions. Specifically, SAbMDE calculates the effort needed to complete a project (a cost). It shows how information can be added to a project with tests (a benefit). And it shows that a cost-benefit objective function can guide an agent towards a desired end product. And then, SAbMDE shows that the guiding information can be converted, using Shannon's information formula, into a probability of project success. Finally, the dissertation offers several points of model validation and suggests a number of this development cycle model's transdisciplinary implications.