Problems of code: ethnographic encounters with the Amazon Forest as a computational model

Abstract

As environmental issues become increasingly datafied, computer code is becoming more important to the quotidian work of earth system scientists. Not only has code become central to their knowledge practices, it has become a way in which social relations travel through global knowledge infrastructures, connecting knowledge, data, and people. This article explores the use of code as material for ethnographic inquiry through the study of how a predictive model about climate change in the Amazon is built at the AmazonFACE modeling laboratory. Based on ethnographic fieldwork, we show that computer code has an ambiguous nature: as both text and process. This duality involving matters of knowledge (how code connects data) and matters of interpersonal relationships (how code connects people) enables it to function as an ethnographic fieldsite. We show that access to source code is a symmetrical problem for ethnographers and modelers alike, producing partial forms of understanding. Based on reading code fragments, we examine how reading, writing, and maintaining code sustain scientific collaborations while simultaneously separating “scientific” contributions from “technical” labor. By following how new and old code are assembled in the model’s source code, we show how careers, knowledge, and relations are made and maintained through computational practices.