Humboldt-Universität zu Berlin - Medienwissenschaft

Vector Models and their intricate ability to measure and project similarity constellations into hyperspace constitute the "Mathematical A Priori'' of current-gen media technology - from word embeddings in Generative A.I. all the way to clustering practices in Viral Epidemiology. Departing from the classical definition of a vector model (Gerard Salton, 1989), our project attempts to analyze the media and the epistemic practices that became encoded in today's approaches to Vector Similarity. Scatterplots, Multidimensional Models and algorithmic techniques of vectorization and clustering began to materialize when the quest to envision complex, multivariate forms of order, pursued in fields such as Biometrics, Criminal Anthropology and Computational Taxonomy, encountered the limitations of paper technology and the human sensorium. By tracing the historical emergence of these Media of Similarity, we will show how the scientific obsession to measure degrees of "family resemblance" and turn natural order into a geometrical object ended up rewriting our very ability to experience similarity, as well as the idea of natural order itself.

Johan Malmsted ist Doktorand an der Universität Umeå im Bereich Medien- und Kommunikationswissenschaften und Mitglied der interdisziplinären Forschungsinfrastruktur Humlab. Er erforscht sonische Medien mit computergestützten und ideengeschichtlichen Methoden.

Sebastian Kawanami-Breu ist Kurator des Signallabors und wissenschaftlicher Mitarbeiter am Lehrstuhl Medientheorien. In seinem Dissertationsprojekt erforscht er kybernetische Neuronenmodelle aus medien- und wissenschaftsgeschichtlicher Perspektive.