Software testing in a scientific research group

Abstract

Scientific software is more difficult to test than many other software products, but scientists are not usually trained in software engineering techniques. Considering how often software is used to produce scientific results, how can we be sure the predictions made from these results are correct? Software engineering techniques should be useful for computational scientists. The problem is they find it difficult to know how to apply domain-independent techniques to the specific problems they face in their work. Nevertheless, we have discovered scientists use their own intuition to reinvent techniques surprisingly similar to those in software engineering. This seems like a good place to start our training.