Neural Networks Acting for Science: Hidden Work and Reconfiguration of Practices at a Thermophysical Lab

Abstract

This article describes the sociotechnical transformations that Artificial Intelligence (AI) instigates within a thermophysical laboratory. Despite the burgeoning interest in how AI will configure our futures, there is still a lack of empirical STS research on AI in Russian sciences (except the issue of ethical considerations), as its pervasive integration into sociotechnical systems is just beginning. The article examines the manifestations of AI as an epistemic tool and methodological intervention, and its impact on reconfiguring laboratory arrangements through an ethnographic analysis of emerging practices. At the core of this work is a case study of a laboratory in the midst of AI integration, with data collected from February to June 2025. Drawing on the fieldwork materials, I seek answers to the following questions: To what extent are “first-wave” STS concepts, developed for studying the social-material practices of the laboratory, applicable to analyzing human-AI engagement with neural networks? How do the modes of collaborative knowledge production within the research group change, if at all? Finally, how is epistemic reliability established for results obtained using black-boxed AI systems, which problematize conventional verification protocols? The case demonstrates that AI is not simply an instrument but an active agent that reconfigures the epistemology of scientific knowledge production.