Julia Irwin on a History of Artificial Perception
We're thrilled to support our students in their summer research. Read about Julia Irwin's research into the history of artificial perception below!
This summer, with support from BCNM’s summer research fellowship, I returned to Washington, D.C., where I visited archives for several chapters of my dissertation. Titled Patterning Recognition: A History of Artificial Perception, the project investigates the conceptual emergence of pattern recognition in the long twentieth century (1890–2012). I show how concepts of vision from the fields of philosophy, psychology, cognitive science, and artificial intelligence were translated by industrial and military institutions into operable labor programs and computer programs alike. In these contested spaces, institutions instrumentalized the conceptual distance between the categories of “human” and “machine” to pursue political ends, and in doing so they patterned recognition itself.
The second half of Patterning Recognition, chapters three and four, focuses on the CIA’s National Photographic Interpretation Center (NPIC) and its implementation of emerging neural network and computer vision techniques. Because of the formerly classified nature of much of this work, I have compiled and cross-referenced materials from several online databases (e.g., the CIA’s CREST Database) and in-person archives. This summer I did archival research for chapter four, titled “Anomaly Detection,” which traces how computer scientists and the NPIC alike responded to the phenomenon of unidentifiable objects or novel patterns in photographs. Drawing on insights on this topic I had gleaned from Dino Brugioni’s archive last summer regarding his contention with the political perils of misidentification in photographic interpretation, this summer I visited the National Archives textual collection to study the declassified government briefs and internal reports associated with “The Condon Report,” what would later be published as the public-facing Scientific Study of Unidentified Flying Objects. I also visited the University of Maryland, College Park University Archives, where I studied the reports of the University’s Center for Automation Research and the Computer Science Center, both directed by the early computer vision researcher Azriel Rosenfeld. Here, I was collecting materials on Rosenfeld’s work on the pattern-recognition technique of “fuzzy logic.” Chapter four examines the explicit and implicit exchanges between these different academic and government venues, delineating how they contended with the conflicting issues of scientific discovery, national security, and public opinion management through the analytic of misrecognition.
While at the National Archives, I was also able to start research on what will eventually be a fifth chapter to Patterning Recognition, “Decision Augmentation/Decision Automation” as I transition it from a dissertation to a book. For this early exploration, I pulled materials related to J.C.R. Licklider and the range of projects he supported while director of the Defense Department’s Advanced Research Projects Agency in the early 1960s. I was interested in particular in the relations between Licklider’s and his collaborators’ work on early human-computer interaction (HCI) and its relationship with social sciences research projects on cross-cultural communication, behavior management, and persuasion. Even though HCI and social sciences researchers and initiatives rarely came into direct contact with one another, by looking to implicit connections via Licklider, I am investigating how both addressed the phenomenon of bodily automaticity. For example, HCI researcher Douglas Engelbart’s design for the mouse, while focused on the goal of more efficient and pleasurable use of information technology, took as a given that users would develop a near automatic mode of engaging with computer interfaces and that efficient use could be achieved by imperceptibly nudging human behavior. Social scientists, too, were captivated by the potential to imperceptibly alter behavior, especially in global contexts undergoing decolonization. Using this institutional and political context as a backdrop, this chapter will examine the conceptual relationship between imperceptibility and decision making.
I want to express my gratitude for BCNM’s support, which has enabled my ability to obtain important documents I knew I needed as well as to pursue open-ended exploration of archives and incorporate its discoveries. Patterning Recognition has benefited greatly from the archival practice I have developed thanks to research trips made possible by BCNM.