Abstract
Artificial Intelligence (AI) is often introduced into scientific practice with a promise: automation. AI is expected to accelerate knowledge production, reduce tedious labor, and free up time for more creative scientific work. Yet, even as AI has made remarkable advances, it rarely enters laboratories as a clean tool that simply replaces existing steps. Instead, it reorganizes scientific practice, reconfigures established roles, and reshapes how scientific knowledge becomes stabilized.
In this talk, drawing on ethnographic research in a biomedical engineering lab in the US, I examine the evolving relations of biomedical researchers and machine learning (ML) systems. With the example of AI-supported image analysis and data pipelines, I show how the roles and “intraactions” (Barad 1996) within these relations are not fixed but are continuously changing due to the introduction of new technological capabilities, the different actors’ intentions and practices, and, also, serendipitous observations and random occurrences. To account for these changes, I introduce the concept of intraversions (Vep?ek 2024): the reconfigurations of human–AI relations over time and in everyday unfolding in which roles, agency, and epistemic responsibilities are continuously redistributed. Intraversions introduces a specific focus on temporal dimensions and processual developments to understand human-AI relations in their becoming (Hultin 2019) and multiple entanglements.
Moving from ML pipelines to recent developments in AI, the talk concludes with an outlook on “AI agents” and how they reshape scientific practice by increasingly actively participating in reshaping their own relations with researchers, tools, and infrastructures.
Bio
Libu?e Hannah Vep?ek is a postdoctoral researcher at Technical University of Munich, affiliated with the TUM Entrepreneurship Research Institute and the Munich Center for Transformative Technologies and Societal Changed (TransforM). The cultural anthropologist and computer scientist completed her doctorate at LMU Munich in the context of the "Playing in the Loop" project (2021-2024) which was funded by the German Research Foundation. Before joining TUM, she was a postdoctoral researcher at the Ludwig Uhland Institute for Historical and Cultural Anthropology at the University of Tübingen. Her work has also included research visits at Cornell University and Toyo University. Her main research areas are digital anthropology, anthropology of technology, science and technology studies, ethics of technology, innovation studies, and digital methods.