Maximilian Pieper

University of Augsburg, LMU Munich

"The dual operation of abstraction and extraction - Towards an environmental sociology of AI"

AI technologies are increasingly characterized in the social science literature on thebasis of an abstraction and extraction paradigm. According to this paradigm,functioning AI technologies are based on an abstracted representation of the world,which is expressed primarily in the form of data. This obscures the material conditions of these technologies on the one hand and favors the extraction of data and resourceson the other. The problem with this conceptualization of AI technologies based on an abstraction and extraction paradigm is that it is related almost exclusively to the digital data foundation, but not to the material resource foundation on which AI technologies are based. This reveals a major problem in social science and, above all, sociological discussions of AI. In the cases in which the actual social embedding of these technologies (rather than their consequences) is critically addressed, it is usually in relation to a bias in the data (e.g. how algorithms reproduce existing socio-economic inequalities) or in relation to the problematic production process of the data (e.g. precarious click-work or violation of privacy when tracking user behavior on social networks). Although these are all relevant areas of research, it seems equally relevant to understand the socio-ecological metabolism of these technologies and thus to penetrate their material side. Here I argue that the abstraction and extraction scheme can be extended. If we understand technologies, following Niklas Luhmann, as 'functional simplifications', it becomes clear how AI technologies, like technology in general, are based on a combination of elements that can be understood as abstract and extracted in the sense that any context of these elements can be ignored as long a sit does not affect their reliable functioning. In this respect, AI technologies cannot solely be understood as extractive simplybecause they are based on abstract data. At the same time, their functioning is based on abstract resources whose ecological damage is ignored, as are the social consequences of their extraction. What is special about AI technologies, however, is that due to their digitality they confront us sociologists with 'data' as an archetypal abstraction, which can also be used to understand classic resource extraction in a new light. At the same time, the data extractivism of AI technologies confronts us with an extractivist dynamic that breaks with the classic divide between the Global North and the Global South.

Bio

I am a doctoral candidate at the international doctoral college "Um(welt)denken - Die Environmental Humanities und die Transformation der Gesellschaft" at the University of Augsburg and the Ludwig-Maximilian University since October 2021. There, I conduct research in the field of sociology of technology and environmental sociology on the phenomenon of technology fetishism and the problematic character of a purelytechnical approach to the multiple environmental crises of the present. The first paper of my dissertation was published in the journal 'Philosophy & Technology' in January 2024.