The promises of computational ethnography: Improving transparency, replicability, and validity for realist approaches to ethnographic analysis

This article argues the advance of computational methods for analyzing, visualizing and disseminating social scientific data can provide substantial tools for ethnographers operating within the broadly realist 'normal-scientific tradition' (NST). While computation does not remove the fundamental challenges of method and measurement that are central to social research, new technologies provide resources for leveraging what NST researchers see as ethnography's strengths (e.g. the production of in situ observations of people over time) while addressing what NST researchers see as ethnography's weaknesses (e.g. questions of sample size, generalizability and analytical transparency). Specifically, we argue computational tools can help: (1) scale ethnography, (2) improve transparency, (3) allow basic replications, and (4) ultimately address fundamental concerns about internal and external validity. We explore these issues by illustrating the utility of three forms of ethnographic visualization enabled by computational advances - ethnographic heatmaps (ethnoarrays), a combination of participant observation data with techniques from social network analysis (SNA), and text mining. In doing so, we speak to the potential uses and challenges of nascent 'computational ethnography.'