Hypothesis Formation and Qualitative Reasoning in Molecular Biology

My Ph.D. dissertation describes an investigation of scientific reasoning from a computational perspective (Karp 1989).1 The investigation focuses on a program of research in molecular biology that culminated in the discovery of a new mechanism of bacterial gene regulation called attenuation. In the first phase of my work, I performed a historical study of this program of biological research in which I reconstructed the different theories that the biologists possessed at different points in time and then analyzed the differences between these successive theories. In the second phase, I developed a qualitative biochemistry for representing theories of molecular biology. In the third phase of the research, I constructed a machine learning program that solves some of the hypothesis-formation problems that the biologists solved during their research. Hypothesis-formation problems occur when the observed outcome of an experiment conflicts with the outcome predicted by an existing scientific theory. The problem is to modify either the theory or the conception of the experiment such that the prediction agrees with the observation. This statement of the problem assumes that we are able to represent a scientific theory in a manner that allows a simulation program to use the theory to predict experimental outcomes and that allows a hypothesis-formation program to reason about the theory to understand why its predictions are faulty.