Research

The project aims to develop an approach that offers an aesthetico-epistemological counterpart to existing historiographical, conceptual, and methodological approaches in the emerging field of critical AI studies. To this end, the project seeks to ascertain the fundamental change in the relationship between the sensible and the intelligible through machine learning (ML). That means researching how technical systems dehumanize and rehumanize perception and exploring the implications for the knowledge process. 

What is new about this approach is that it combines media-epistemological questions with media-aesthetic perspectives and that it uses a praxeological approach to focus decidedly on the empirical investigation of current machine-learning practices. Existing media epistemological work primarily aims to open up and evaluate historical bodies of knowledge or to investigate the genesis of the techniques and structuring concepts that produce current knowledge formations. The project, instead, uses praxeological methods that were developed in the context of science and technology studies, to investigate the fabrication of knowledge in laboratories and at aerospace centers (research line 1) and examines how approaches from the field of critical AI studies highlight the reflexive and critical potential of current positions in the field of media art today (research line 2). 

While the first research line focuses on knowledge-generating forms of machine learning, the second theme concentrates on ML in aesthetic practices. The second research line aims to analyze selected works particularly suitable for examining the aesthetic strategies for addressing social, global, and planetary imbalances that ML practices reinforce. The guiding hypothesis is that the artistic works uncover the mystifying discourses surrounding artificial intelligence and offer ways out of practices that sustain inequality. At the same time, some of the works adopt the popular rhetoric of enthusiasm and fear and profit from problematic practices of data collection and annotation. 

The project pursues the strategic research objectives to contribute to the interdisciplinary dialogue with the natural sciences and computer sciences and to undertake a self-reflective assessment of media epistemological research. To this end, the planned work combines insights from critical AI studies, media aesthetics, and science and technology studies.

The overarching goal of the project is to develop a new theoretical and methodological approach in the field of media studies by conducting exemplary media ethnographic laboratory research at the German Aerospace Center and other selected locations of the Copernicus Earth Observation Program in combination with a dispositive and usage analysis of geo-medial platforms such as Destination Earth (DestinE), which collect data for the development of simulation models, so-called digital twins, of the Earth. Specifically, this means to explore how and for what purpose scientists use machine learning in satellite control and in the evaluation of earth observation data to answer the question of how such media techniques generate insights into planetary climate development. The project seeks to corroborate the hypothesis that the knowledge generated by sensor technology and machine learning in remote sensing suggests not just the hypervisibility and calculability of planetary phenomena but, moreover, is organized around a black box. The predominantly automated infrastructures of data collection and data analysis exclude human perceptual and cognitive performance in favor of a fantasy of predictability and prognostics and relegate human interpretative activity to the beginning – the conception phase – and the end – the interpretation phase – of the knowledge process.

The first sub-goal of the project is to explore the infrastructures and media techniques of planetary perception to determine the relationship between the sensible and the intelligible by evaluating the interactions between the aesthetics and epistemology of satellite images and learning algorithms. Based on the insight that sensing and sense-making drift apart here, the project will explore a middle way between a media epistemology that understands artificial intelligence (AI) as ideology and a media epistemology that focuses on AI as a classification tool. 

The second sub-goal is to work out which worldviews are reinforced or undermined by the digital twins of the earth, which simulate different, thematically-oriented future scenarios. The aim is to determine what becomes perceptible and what becomes imperceptible through the suggestion of a planetary oversight and what consequences this has for the self-world relationship. The project claims that the aesthetics of the hyper-visible, which simulation models of the earth imply, anesthetizes the sense of the invisible.

The third sub-goal is to demonstrate the added value that a praxeological-microanalytical approach brings to theory formation in media studies. To this end, the empirically gained insights into the sensory-aesthetic practices of remote sensing are to be linked to the epistemological reflection of the worldviews manifested in the simulation models.