Conspicuous Computing: Software Development as a Knowledge Practice
- Author(s): Erickson, Seth
- Advisor(s): Kelty, Christopher
- et al.
This dissertation is a study of scientific software development based on ethnographic fieldwork in a computational physics lab. I present scientific software development as the ongoing materialization of epistemic objects. Conspicuousness is a term adopted from phenomenology to describe a reflective engagement with software that is distinct from the unreflective, seamless, and transparent relationship that designers often aspire to. To the extent that scientific software matches the trajectory of inquiry, it is frequently conspicuous in this sense—in fact, it must be so if it is to function as a setting in which new insights emerge. The maturity of computational physics makes it an ideal context for understanding how knowledge production inheres in code work and for identifying how recent changes in computational and scientific norms have affected scientific software practices. The field site was a university lab specializing in the development and use of codes for simulating plasma. I characterize the memory practices of the lab as efforts to make the unexpected behaviors of code productive. I describe the lab’s deep-seated ambivalence toward open science initiatives and reflect on the conditions that allow scientists to maintain autonomy and control over their technical practices. Finally, lab members differentiated between production codes, reference codes, frameworks, classroom codes, and toy codes: I present these actor categories as a set of commonplaces that allowed simulation developers to manage the epistemological concerns of software development and the reception of their code in the broader community.