Software development as mediated activity: applying three analytical frameworks for studying compound mediation

Field research in software documentation has a tradition of investigating how artifacts (from documentation to online help to interfaces to mundane equipment such as Post-It? notes) mediate or enable workers to perform complex tasks (see for instance [29]). Understanding artifacts and mediation can be key to understanding how well documentation supports work, and consequently, how we might design information to fit work patterns. Yet the field of technical communication has developed or adapted relatively few analytical frameworks for examining compound mediation, the ways that sets of artifacts work together to help workers get their jobs done. Such frameworks are important to understand because they provide us with guidance for investigating the mediatory relationships among artifacts - guidance which has important ramifications for intelligently designing information systems and inserting designed artifacts (such as documentation) into existing systems.In this paper, I use three analytical frameworks - contextual design's work models [4, 5, 23], distributed cognition's functional systems [1, 2, 13, 24], and genre ecologies [25, 26, 27, 28, 30] - to examine observational and interview data from a 1997 study of software developers. The observational study is a 10-week investigation of 22 software developers at work, focusing on how artifacts (such as manuals, code comments, and the code itself) collectively mediated the developers? production and comprehension of code at three units of the same global corporation. The study provides a good case for basing a comparison of the three frameworks because it (a) involves comparing multiple artifacts and complex use of artifacts across the different sites, and (b) uses ethnographic methods similar to those often used by proponents of the three frameworks.By applying the three frameworks to the same study, I illustrate which aspects of compound mediation are illuminated and unexplored by each analytical framework. Based on the comparison, I discuss the strengths and weaknesses of each framework for exploring compound mediation, and I suggest ways in which the frameworks might be coordinated to produce different pictures of work.