A framework for software modelling in social science research

Social science is critical to decision making at the policy level. Software modelling and simulation are innovative computational methods that provide alternative means of developing and testing theory relevant to policy decisions. Software modelling is capable of dealing with obstacles often encountered in traditional social science research, such as the difficulty of performing real-world experimentation. As a relatively new science, computational research in the social sciences faces significant challenges, both in terms of methodology and acceptance. However, there is great potential for computing to aid in the application of scientific thinking to the grave issues facing society. This is particularly true since technological advances and societal change continue to make our lives more complex. Policy decisions can have significant impacts in the lives of those affected; it is imperative we strive to develop novel and effective methods to inform these decisions. This thesis focuses on the interaction of modelling, software development, and experimentation in computational social science research pursued by small teams of interdisciplinary scientists. I present an innovative software development framework designed for this kind of research. By integrating software throughout the research process for both modelling and experimentation, and utilizing a flexible and iterative development model, my framework addresses many pressing issues of computational social science: uncertainty due to lack of data or changing conditions; validation of models; usability; rapid adjustment to changes in direction; facilitating collaboration; and communication of results to peers and stakeholders. Case studies of projects developed using this software modelling framework are used to illustrate and discuss the approach. The case studies span several fields of the social sciences, including Criminology, Geography, Political Science, and Public Health.