A methodology for developing computational implementations of scientific theories

Abstract

Computer programs have become a popular representation for scientific theories, particularly for implementing models or simulations of observed phenomena. Expressing a theory as an executable computer program provides many benefits, including: making all processes concrete, supporting the development of specific models, and hence enabling quantitative predictions to be derived from the theory. However, as implementations of scientific theories, these computer programs will be subject to change and modification. As programs change, their behaviour will also change, and ensuring continuity in the scientific value of the program is difficult. We propose a methodology for developing computer software implementing scientific theories. This methodology allows the developer to continuously change and extend their software, whilst alerting the developer to any changes in its scientific interpretation. We introduce tools for managing this development process, as well as for optimising the developed models.