Lyman Report: Julia Irwin
We're so pleased to have awarded Julia Irwin the 2022 Lyman Fellowship. Read below how the Fellowship assisted her research this past summer.
Thanks to the support of BCNM’s Lyman Fellowship, I was able to devote my summer to researching and writing two chapters of my dissertation, as well as preparing an article manuscript for submission to a peer reviewed journal. My dissertation, titled Patterning Recognition: A History of Automated Visual Perception, examines the reorganization of optical sensing and formalization of image interpretation in U.S. military and industrial institutions throughout the twentieth century that were the necessary preconditions for today’s computer vision technologies and global surveillance apparatus. The project illustrates how film and photography have, since their emergence, been entwined with both computational processes and the dynamics of embodied perception and bodily difference. Additionally, the dissertation contributes to contemporary discourse on AI bias, offering a historical and theoretical grounding that both troubles how we define “artificial intelligence” and renders today’s black-boxed technologies more legible. Taking cues from the design and uses of machine learning algorithms today, I identify four modalities of vision—kinesthetic sensing, object recognition, anomaly detection, and behavior pattern recognition—and dedicate a chapter to a historiographic analysis of each. I ask, how was human vision conceived as an automatable entity in the first place? Perception itself had to have been patterned, turned into a procedure, in order to eventually be executable by a computer, and I scrutinize the politics of this translation process. My analysis lends particular attention to the social, technical, and (geo)political forces at play in the transitionary period between the Second Industrial Revolution and the Information Age, with a focus on the United States as it perceived of itself on a global and cosmic stage.
This summer I completed the research and writing of chapter one of my dissertation. Titled “Kinesthetic Sensing, the Automatic Factory, and and the Interactive Film and Media Theory of Lillian Moller Gilbreth,” the chapter identifies an early-twentieth-century conceptualization of a visual media interface, sited at the exhibition of cinema intended to train workers to perform factory tasks, which anticipated the cybernetic theories of Norbert Wiener as well as the human-computer- interaction and robotics contexts of today’s media environment. This June, I rounded out my research for this chapter by focusing on Lillian and her husband Frank Gilbreth’s work to identify career paths for disabled WWI veterans. The theories developed by Lillian Gilbreth and practices the couple undertook during this period betray a profoundly contradictory stance on individual difference. They framed the subjective experiences of disabled persons as irrelevant or outright disruptive, just as they treated one’s body and all of its idiosyncrasies as endlessly addressable, and therefore serviceable to an overarching system of productivity—an endeavor the duo termed “automatic invention.” Automation, as framed by industrial psychologist Lillian Gilbreth, involved addressing bodies directly—calibrating the senses, articulating nervous system activity, and ordering the limbs to engage with factory implements according to prescribed sequences—to forge a compatibility with a factory environment whose scale and speed of operations were profoundly incongruous with those of a human form. To complete this research, I continued working with the archivists at Purdue University who oversee the Frank and Lillian Gilbreth Papers and visited the Frank and Lillian Gilbreth Collection at the Archives Center of the National Museum of American History, Smithsonian Institution in Washington, D.C.
Chapter two of the dissertation, titled “Object Recognition, ‘Cratology,’ and the Discursive Field in the Making of Cold War Image Intelligence,” examines the conceptual origins of object recognition as a mode of visual perception which could be interchangeably executed by a human or a machine. Animated by the U.S. DoD’s 2017 announcement of Project Maven, an initiative to incorporate AI into the military intelligence and targeting cycle by way of object recognition algorithms, this chapter historicizes the role of the CIA “photo interpreter” in the 1950s and 60s, and the ways in which the patterns discerned in aerial photographs were bound up with, and overdetermined by, interpretive methods devised specifically for their suitability for automation. Washington, D.C. is also the site of several archives relevant to this chapter, and this summer I also visited both the National Archives as well as the Dino Brugioni Collection at the National Air and Space Museum, Smithsonian Institution. As of August, the writing of this chapter in underway and is to be completed in early September.
I want to express my gratitude to the Berkeley Center for New Media and to the late Professor Peter Lyman. This fellowship support came at a crucial time in the development of my dissertation project, as my research has gained momentum and the writing has now materialized into finished chapters.