While general purpose scientific software has enjoyed great success in industry and academia, domain specific scientific software has not yet become well-established in many disciplines where it has potential. Based on a survey of the literature as well as the authors’ experiences contributing to Econ-ARK, a structural modeling toolkit for Economics, we argue that this is due to the well-documented skills gap that prevents researchers, publishers, and professors from making the most of the opportunities afforded by scientific software. When researchers professionalize their code, it enables more cumulative progress in research and facilitates technology transfer. When publishers release interactive computational artifacts, it enables constructionist learning of the material. When students are trained in software engineering, they can participate fully in the reproduction of their scientific field. This is especially the case for fields where scientific knowledge is represented in software code, as in the case of Economics. The skills gap will not be closed until software engineering is considered a core skill for the discipline. Software engineering should be reconceived as a research method.

Keywords:computational methodcomputional thinkingconstructionist learningresearch software engineering