Computation as Context: New Approaches to the Close/Distant Reading Debate

ABSTRACT:The perceived dichotomy between distant and close reading continues to shape conversations about the digital humanities. This debate has functioned as a red herring, overshadowing misreadings of computational work. Drawing on two case studies, I argue that the information produced from computational methods should be understood as a type of context, rather than as data or as a textual "reading." Viewing computational output as context implies that impartial and flawed methods can still supply valuable information. This is a radical departure from the field's current preoccupation with methodological validity. My first case study looks at Google's flawed learning algorithm Perspective. I posit that the tool can be reverse engineered to examine racist and sexist attitudes. I also examine my computational study of direct address in 2,000 Anglophone novels. Despite technical flaws, the project facilitated the textual recovery and close reading of several nineteenth-century African-American novels. As these examples show, the computational analysis of literature produces information that, much like biographical and historical context, is in its own right subjective and incomplete but can be used to provoke further acts of interpretation.