Review of the first annual conference of Schaufler Kolleg@TU Dresden on "Artificial Intelligence as a Concept in the Humanities and Social Sciences" (1 - 3 Dec 2021)

The first interdisciplinary online conference of the Schaufler Kolleg@TU Dresden took place from December 1 to 3, 2021, focusing on artificial intelligence as a concept in the humanities and social sciences.

Over the course of three days, 29 researchers dedicated eight panel discussions and two keynote presentations to conceptual issues relating to AI in the humanities and social sciences. The conference centered on how these fields in particular could supplement the current social and political debates surrounding the use of AI by adding conceptual notions to the discourse and considering the cultural effects of using AI technology.

The participants’ vibrant engagement with the topic at hand reflected its pertinence – not just to the work of the Schaufler Kolleg@TU Dresden, as presented by chairman Lutz Hagen in his opening remarks, but also to the clarification of the concept of AI within the various disciplines represented at the conference.

To kick everything off, Ann-Kristin Kühnen moderated the panel on “Sociality with machines,” which comprised three sociological segments. Jan Tobias Fuhrmann analyzed AI from a systems theory standpoint, contrasting its prominent depiction as a self-contained system with a relational understanding of distributed intelligence. In the same vein, Hannah Link established the concept of “circuit intensity” for robotics researchers’ perceptive relation to the predominant AI technologies in their field. In the panel’s final segment, Jonathan Hardt and Maximilian Locher presented their investigation of dynamics in the genesis of social patterns of relation between human and machine.

The second panel, moderated by Carsten Junker and titled “Technopoetologies,” started off with a question about the performative incorporation of AI as a (co-)author of literary texts, exemplified by Pharmako-AI (K Allado-McDowell, 2020). Christian Heck then introduced the concept of adversarial poetry, which posits that the connection between code and political activism can also shape subversive spaces in algorithmically structured environments. The panel concluded with a presentation by Sara Morais dos Santos-Bruss. Her critical view of the historical genesis of the narrative of autonomous machines, which is perpetuated in the paradigm of AI as a technical-neutral entity, referenced the hidden discrimination and precarious working conditions that accompany the development and use of AI systems.

Following Elena Esposito’s keynote presentation titled “Artificial communication” and moderated by Dominik Schrage, the annual conference’s first day wrapped up with an extensive discussion in which the day’s numerous sentiments were digested and refined. The question of an appropriate concept of intelligence for algorithmic systems was addressed. Participants also discussed the extent to which algorithmic communication can coalesce with conventional communication concepts.

The second day began with the “Embedded normativities” panel moderated by Luise Müller. Following Christoph Merdes’ discussion of concepts for justifying algorithmic decisions on the basis of their comparability, Rita Jordan (LINK) discussed how approaches of participative design practices can inform a political theory of democratic AI.

The next panel “Historicity and historizations” was moderated by Rebekka Roschy (LINK). Input from Florian Müller, Helen Piel and Susanne Schregel focused on the relationship of the historical development of AI as a concept, its utopias and present conceptual intentions as well as parallel conceptual designs.

Kerstin Schankweiler moderated the panel “Aesthetics of machine intelligence” to wrap up the day’s academic program. Fabian Offert conceptualized images created by generative adversarial networks as multidimensional sculptures, while Michael Klipphahn (LINK) presented his exploration of AI as a springboard for artistic work and narrative beyond basic technical embedding in art. Miriam Akkermann concentrated on the opportunities for integrating AI in music and the relationship between author and work in consideration of the incorporation of machine practices.

Artists in Residence of the Schaufler Lab@TU Dresden Christian Kosmas Mayer (LINK) and Anton Ginzburg (LINK) led a virtual tour through the exhibition “A&I” (LINK) at the Altana Gallery. The presentation closed with an audience discussion, moderated by the exhibitions curator Gwendolin Kremer.

The conference’s third and final day began with a presentation by Jan Georg Schneider and Katharina Zweig on the reliability of algorithmically generated predictions and their inclusion in societal practice. The critical exploration of the e-rater scoring engine – a tool designed to identify features related to writing proficiency so they can be used for scoring and feedback – led to an extensive and productive discussion moderated by Susann Wagenknecht.

Simon Meier-Vieracker moderated a seventh panel that brought together methods for the automated creation of meaning. Claire Y. Song posed the question of if and how machine learning processes could replace semiological language systems and mediate between them. Returning to this point, Jonathan Roberge and Tom Lebrun explored the epistemological and historical construction of linguistic models based on machine-trained artificial neural networks. Yaoli Du and Nadine Schumann exemplified different avenues to the notion of information as a necessity in the interaction between human and machine using social concepts of cognition.

In the conference’s final panel “Spatial topologies of AI” – moderated by Richard Groß (LINK) – Lukás Likavćan, Patricia Reed and AA Cavia explored topological approaches to AI as well as their impact on philosophical and geopolitical concepts.

Under the moderation of Orit Halpern, Matteo Pasquinelli closed the conference with his keynote address “What kind of knowledge is machine learning? For a political epistemology of AI.” It considered which knowledge models are represented and supported by machine learning. In this regard, the presentation and accompanying dialog centered around the opposition of the paradigms of deductive and inductive logic, symbolic representation and statistic modeling – including from a historical perspective – and the role ideological discourses play in how AI is received and operated.

All the presentations as well as the surrounding discussions demonstrated that AI is more than a rich, empirical item of interest in the humanities and social sciences. It is also an exciting catalyst for theoretical reflection, which in turn can highlight present-day challenges in the interdisciplinary discourse. It became apparent that conceptually engaging with AI does not represent an alienation from the inclusion of algorithms and their applications in society and culture. On the contrary, it opens up a more differentiated view thereof.

Many questions still remain unanswered, as made clear by the overarching, ongoing debate about acceptable concepts of intelligence. Moreover, the jury is still out on the translatability of problems and the definitions of phenomena between technical disciplines and discourses in the humanities and social sciences.