Veranstaltungsprogramm

Artificial intelligence (AI) challenges human intelligence and our humanistic self-image. This is justified to a certain extent, but the concerning controversies usually operate with frames of reference that fall short. Humans and technology have always been intertwined in hybrid forms of life. However, the exact nature of this hybridity is misunderstood when inadequate dichotomies between human subjects and technical objects on the one hand are replaced by a totalising conception of a cybernetic information universe on the other hand that reduces everything that exists to this single point of comparison. As the paradigm of digital society, AI is a bearer and expression of such a digital ‘analogism’ (in the analytical sense of Descola): it not only represents a cybernetic scheme for interpreting the world, but also makes this cosmology plausible at the level of practical experience. AI thus deepens and generalises cybernetic conventions and functional patterns of justification that have a long history in industrial society but are reaching a new level of reality with the help of these algorithmic practices and materialities.

The paper asks about the prerequisites for a critical theory of these developments. To regain critical ground, a better understanding of the ontological heterogeneity of the societal modes of existence (Latour) that are assembled in hybrid forms of life is needed. This thesis can be clarified by examining Rosa's theory of world relations, which on the one hand is interested in the diverse ontological relations of the social, but on the other hand establishes an abstract foil for comparison with the principle of resonance, which paradoxically is suitable for justifying the cybernetic expansions of AI as well. This is because a logic of digital resonance between the world and data is inscribed in the training of AI itself. For this reason, this principle does not provide a sufficient foundation for a critical theory of AI. Instead, the concept of friction can be used to focus attention on the heterogeneity of social reality, which on the one hand resists the analogistic subsumption by digital data and algorithms and on the other hand helps to open up (theoretical) sources for the critique of AI.

The emergence of ‘Emotional AI’ – designed to process, interpret and even portray emotional expressions – is currently adding new layers of complexity to the digital transformation. From emotion-recognition systems to AI-driven chatbots and social robots, AI devices are not only responding to (human) emotions, but are now said to display emotional experience, thereby shaping how feelings are expressed and interpreted. These developments challenge theoretical, conceptual and empirical approaches to the inner workings and affective, societal dimensions of digital technology. These advances are often embedded in reductionist frameworks that characterize emotions as universal, biologically pre-defined and exclusively human phenomena (Stark and Hoey 2021). AI is positioned as the mediator or carrier of these emotions, while still operating as an autonomous entity detached from its socio-material, political and affective embedding. However, I argue that such framings limit our ability to critically understand the affective dynamics and underlying mechanisms of AI-driven transformations. Persistent dichotomies between human and nonhuman, experience and computation or technological and socio-political development risk reinforcing essentialist notions and inscribed stereotypes, while overlooking the complexities and power dynamics in which AI systems and models of emotion are embedded.

Against this background, this contribution challenges dichotomous understandings of AI-driven transformations by framing them as sites of contestation emerging with(in) and through affective socio-material relations. Shifting the focus to these transitional spaces – the ‘in-between’ of human and ‘artificial’ emotions – provides a different perspective on the frictional and contingent nature of (Emotional) AI. Drawing on relational accounts of affect (Bennett 2010; von Scheve 2018; Wetherell 2012) and concepts such as "emotive actants" (Stark 2019) and "affective artefacts" (Piredda 2020; Viola 2021), this contribution explores the socio-materially entangled affects and their instabilities and ontological ambiguities which characterize contemporary digitalization. Situating AI within co-constitutive, affective socio-material entanglements could offer alternatives to the dominant frameworks of technological and emotional normativity and new, emancipatory visions of how we live, feel and relate to each other.

Digital transformation narratives barely recede. While past and contemporary proposals distinguish themselves regarding the dominant technological paradigm of the time, they all share that digitalization and now AI-driven transformation is here to stay. Various social areas are considered to need to transform their structures, processes, and knowledge-making digitally. This suggests a binary state of transformation with an envisioned end, while stuck in a continuous transition. I call this simultaneous state a heterogeneous transition to account for the multiple and mostly messy conditions of digital / AI-driven transformations. The paper draws on the concepts of post-digital (Cramer, 2015), friction (Tsing, 2005), and dissonance (Stark, 2009) to account for the heterogeneity and in/stabilities of digital/AI-driven transformation processes in practice. It presents a practice-theoretical case study exploring the use and implementation of computational technologies in architecture and building engineering. The building sector is considered to lag behind in its digital transformation. At the same time, architectural practice and engineering have relied on computer-aided design technologies or computational methods for various tasks over the past decades. Despite that, several new “visions” of the sector’s digital transformation co-exist and co-construct each other (Braun & Kropp, 2023) while neglecting the idea that practitioners already embed digital and computational technologies into their work and decision-making. This asks the question: What entails a digital transformation? How is its fulfillment recognized, and by whom? Currently, the sector is “monopolized” by two dominant and overlapping visions – building information modeling and AI/data-driven design. While they dominate narratives, they fail to account for the nuanced, differentiated modes of digital transformations encountered in everyday practice. Drawing upon interviews and fieldwork in German architecture and climate engineering design offices, I attend to the heterogeneous transitions encountered and performed by actors and organizations. As the case study shows, an actual “fulfillment” of digital transformation would require a stabilization or standardization of what are mostly individual human-non-human-constellations, thereby creating new frictions and dissonances.

In the military context, AI is currently being implemented at various levels – from strategic foresight and wargaming to autonomous weapons and targeting systems. This development can be summarised under the term “algorithmic warfare” (AW). For many military analysts and planners, this ongoing transformation represents a new Revolution in Military Affairs (RMA). While the original concept of an RMA referred to a comprehensive and complex upheaval of military doctrines, strategies, tactics and technologies, the current RMA rhetoric focuses almost exclusively on the “disruptive” power of AI and invokes the need to decisively push forward technological change. Due to an alleged existential threat posed by near-peer adversaries who would not hesitate to use AI without being guided by a “moral compass”, the development of military AI is ultimately not a decision to make. At the same time, however, AW is becoming the subject of a moral controversy in which technoscientific promises of a faster and more precise (and thus less lethal) warfare on the one hand are pitted against fears of the dehumanization of warfare and a loss of so-called Meaningful Human Control over crucial acts of war on the other. However, as I will show, both positions reproduce the technological determinism of the RMA rhetoric on a different level. While proponents derive positive effects of AW directly from the technological properties of AI-enabled systems, their criticism is based on the tacit assumption that the autonomy of human operators and commanders is now overridden by the autonomy of machines that would operate independently of human (moral) reasoning. Following this analysis, I will address the question of what transformations can actually be observed in the course of AW and how these can be conceptualized through an appropriate understanding of military human-machine interactions beyond technological determinism and human autonomy. Finally, I would like to hint at what this implies for the urgent need to regulate the use of AI in the military under International Humanitarian Law or by other means.

This contribution starts from a premise that research is an act of creative in(ter)vention and explores what kind of inter- and trans-disciplinary methodologies might be needed in times of digital transitions that would offer tools to adequately address the complexity of such transitions as well as open possibilities for interventions. Already in 1997 computer scientist Phillip Agre suggested that computing, and specifically AI, is in need of a critical technical practice that would enable engineers to engage with complexity and the sociotechnical aspects of AI. Arguing that computing has the tendency to “reinvent virtually every other site of practice in its own image”, Agre proposed that a critical technical practice would allow for more reflection and epistemic as well as methodological openness on behalf of computer scientists. This tendency to render other practices “in its own image” characterizes many of the digital transitions and AI driven transformations: injecting their own logic of operations into the domains that are digitized, they operate what Deleuze and Guattari would call “diagrammatically", which I will illustrate with examples of simple machine learning algorithms. Taking a new materialist and post-qualitative approach, I will suggest that such diagrammatic operations can be productively frictioned and re-oriented. I will present a notion of “transposition”, drawing from Braidotti’s work, as one way of creating inventive methods (Lury & Wakeford, “Inventive Methods”, 2012). I will describe an example of transposing critical concepts from feminist theory into machine learning systems design and discuss the productive frictions and re-orientations that such a transposition creates. Bringing together AI and machine learning systems design with critical new materialist theories, I will argue for expanding Agre’s notion towards the development of a transdisciplinary critical technical practice.