Production and use of documentation in scientific software development

Software is becoming ubiquitous in science. The success of the application of scientific software depends on effective communication about what the software does and how it operates. Documentation captures the communication about the software. For that reason, practices around scientific software documentation need to be better understood. This thesis presents four qualitative empirical studies that look in depth at the production and use of documentation of scientific software. Together, the studies provide evidence emphasising the importance of documentation and shows the handshake between written documentation and the informal, ephemeral information exchange that happens within the community. Four reasons behind the obstacles to producing effective scientific software documentation are identified: 1) the insufficient resources; 2) lack of incentives for researchers; 3) the influence of the community of practice; 4) the necessity of keeping up with the regular advancements of science. Benefits of the process of producing documentation are also identified: 1) aiding reasoning; 2) supporting reproducibility of science; 3) in certain contexts, expanding the community of users and developers around the software. The latter is investigated through a case study of documentation crowdsourcing. The research reveals that there is a spectrum of users, with differing needs with respect to documentation. This, in turn, requires different approaches in addressing their needs. The research shows that the view of what constitutes documentation must be broad, in order to recognise how wide a range of resources (e.g., formal documents, email, online fora, comments in the source code) is actually used in communicating knowledge about scientific software. Much of the information about the software resides within the community of practice (and may not be documented). These observations are of practical use for those producing documentation in different contexts of scientific software development, for example providing guidance about engaging a community in crowdsourcing documentation. i