Supporting Scientific SE Process Improvement

The increasing complexity of scientific software can result in significant impacts on the research itself. In traditional software development projects, teams adopt historical best practices into their development processes to mitigate the risk of such problems. In contrast, the gap that has formed between the traditional and scientific software communities leaves scientists to rely on only their own experience when facing software process improvement (SPI) decisions. Rather than expect scientists to become software engineering (SE) experts or the SE community to learn all of the intricacies involved in scientific software development projects, we seek a middle ground. The Scientific Software Process Improvement Framework (SciSPIF) will allow scientists to self-drive their own SPI efforts while leveraging the collective experiences of their peers and linking their concerns to established SE best practices. This proposal outlines the known challenges of scientific software development, relevant concepts from traditional SE research, and our planned methodology for collecting the data required to construct SciSPIF while staying grounded in the actual goals and concerns of the scientists.