Den Bildern auf der Spur. Epistemische Praktiken der Bildverarbeitung und Mustererkennung in Teilchenphysik, Nachrichtentechnik und Informatik in der BRD, 1950–1980.

The dissertation investigates the conditions of pattern recognition and image processing within the context of digital computing and information theory, examining the shifts in epistemic dynamics brought about by these technologies. It provides a diachronic and synchronic comparison of three historiographically framed case studies from scientific research and development. These case studies explore efforts to make the visually perceptible world accessible to machines for the generation of knowledge and the derivation of actions and decisions.

Methodologically, the analysis relies on a critical evaluation of primary sources, with a focus on archival materials, pre- and post-mortem bequests, patents, circuit diagrams, and scientific literature. Oral history interviews and fieldwork further enrich the source base. Drawing on Hans-Jörg Rheinberger's theoretical framework of "experimental systems," historical epistemology offers the theoretical lens through which the source material and the object of investigation are examined.

Through a close reading of the sources and a detailed analysis of technical apparatuses, the dissertation initially treats image processing and pattern recognition as epistemic practices in the production of scientific knowledge. The analysis reveals that these practices were often shaped by similar methodologies to those found in the automation of office, administrative, and industrial functions. Attempts to implement these systems often took place in research and work practices that had already been subjected to rationalization and divided along the lines of gender, class, race, and education. This was accompanied by rhetoric that emphasized the inferiority of female-associated traits in comparison to the computational machine.
To develop a framework for understanding the collective and performative nature and “style” of (scientific) seeing, the dissertation employs the concept of "Denkstil" ("thought style"), as developed by Ludwik Fleck. The findings demonstrate that these practices of seeing could not be fully represented in technical terms. On the one hand, existing workplace conditions played a crucial role in shaping the development and use of these systems. While the systems and their processes were communicated as being universal and objective, they were, in fact, grounded in collective practices. On the other hand, these systems carried their own material, epistemic, and practical dynamics, thereby reshaping scientific work practices.

By tracing the trajectories of these systems, the dissertation highlights a shift in focus from those scientific things (“epistemische Dinge”), whose visual traces were processed by the systems, to the epistemic things that emerged through the development of the systems themselves. This shift underscores the idea that attempts to automate seeing have led to enduring epistemic, technical, and practical path dependencies—many of which continue to influence contemporary discussions surrounding AI, machine learning, and artificial neural networks.