Academic reputation and scientific software development

Software is increasingly vital to the practice of science, yet we have little understanding of how it interacts with the system of scholarly communication, how it is currently produced and how to improve its production. A promising strategy is the modular strategy for innovation and collaboration, which suggests a mirroring between modular technical and organizational structures will lead to improvements within modules yielding overall system improvement. In 2007 the BLAST bioinformatics software underwent a modular re-design; we examine innovation before and after using interview and publication data. Surprisingly we find little impact: rather than improvements within modules, innovators continue to fork BLAST, releasing code as separate projects described in academic publications. Analysis of interview data supports the argument that this is because academic credit is not as divisible or re-assignable as money; the system benefits of modular improvement are difficult to apportion appropriately. This means that reputation, as a reward for collaboration is effective for motivating development and releasing, but undermines motivations for the integration activity that is crucial to building cyberinfrastructure. This highlights an assumption of the modularity strategy: not only does the product need to be modular (so you can make changes of limited scope), but incentives must also be appropriately modular (so you have reason to make changes of limited scope).