On the Confusion Matrix

Confusion matrices are graphical objects used in machine learning to demonstrate the efficacy of models. They frequently take the form of brightly colored figures adorning scientific reports and technical documents. Guiding readers of these texts through an interpretation of models with their special affordances, such as diagonal lines that draw the eye from one corner to the other, confusion matrices function as graphical arguments. That they instruct interpreters of machine learning how to read the presented methods and models is no surprise, as early forms of these matrices were instrumental to mid-twentieth-century psychological research on teaching children how to read. This essay reconceives of confusion matrices as not only data visualizations but also media objects through the tracing and reconstruction of their material and cultural history

Confusion · Psychoanalysis · Chemistry · Cybernetics and Technology in Society · Data Visualization and Analytics · Digital Humanities and Scholarship · Psychology