A Computational Theory of Dynamical Modeling in Scientific Reasoning

The focus of this research is on dynamical modeling in scientific reasoning. Dynamical modeling uses, transforms, and constructs models which capture the causal and structural dynamics of given phenomena. I describe a computational theory of dynamical modeling which utilizes representational and processing languages from artificial intelligence to represent and manipulate the dynamical models used in scientific reasoning practices. My theory uses transformational strategies such as model-based analogy (MBA) and Structure-based Model Transformation (SBMT) to allow the agent to draw on knowledge external to the model to solve problems. By doing so, the agent is able to revise and thus to evolve the model. To carry out this research I have developed a cognitive methodology which employs the following methods: cognitive analyses of historical case studies, expert problem solving protocols, and computational modeling. The thesis of this proposal is that a computational theory of dynamical modeling which includes strategies such as MBA and SBMT is a useful part of a cognitive methodology for studying science. Important computational issues include: the content, organization, and representation of dynamical models, and the nature of the methods used to transform them, e.g. how is knowledge transferred from analog models to target problems, and how are models transformed. Important philosophical issues include: the validity of model-based approaches in scientific reasoning, the validity of computational modeling in the study of science, and the nature of empirical methods in the study of science. The usefulness of my computational theory is demonstrated through a program called ToRQUE which models dynamical modeling in a problem solving protocol.