How do we communicate with chatbots? How do we talk about our robot vacuum cleaners? And how can art-historical style studies inform our understanding of artificial intelligence? These are questions being addressed by the Uncertain Archives research group, headed by Associate Professor Kristin Veel. By Kristin Veel, associate professor, PhD, Department of Arts and Cultural Studies, University of Copenhagen The production, gathering and communication of data represent a central, global challenge of modern life. We live our lives surrounded by what are effectively gigantic data archives. Data is constantly being gathered about us and by us – in both the public and private spheres – on global search engines, in the health sector, on our fitness trackers and, most recently, by various infection-tracing apps worldwide. It has gradually come to be generally known that gathering these large volumes of data – what can be called a datafication of our lifeworld – not only enables us to see new connections and avoid human error, but also gives rise to a significant number of new uncertainties. Theoretical vocabulary Taking humanistic science as a starting point but forming part of the emerging interdisciplinary field of critical data studies, it is precisely these uncertainties that we in the Uncertain Archives research group are working to understand, not least by drawing on post-structural archive theory. We live our lives surrounded by what are effectively gigantic data archives This is a theoretical field that has long understood archives not as safe deposits in which we store knowledge of the past, but as dynamic entities that are constantly in flux. This approach gives us a theoretical foundation for understanding uncertainties in our contemporary big data archives. In addition, we draw on intersectional feminist theory and critical race theory, which provide valuable tools for identifying and understanding underlying bias, prejudice and pre-understandings. Our work focuses on creating a theoretical vocabulary that will enable us to talk about and understand the uncertainties we are seeing in an increasingly datafied world. We are working to address not only the functional aspects of data-intensive environments but also the signs and bodies that circulate here, and the way in which we encounter them through linguistic narratives or visualisations that interpret data. An example of the importance of artistic research to the work of the Uncertain Archives group: Katrine Dirckinck-Holmfeld’s work The Christmas Report & Other Fragments (2017). Photo: Katrine Dirckinck-Holmfel Interdisciplinary approach Our work depends on close dialogue with multiple branches of science, including computer science and the social sciences, but we also attach importance to working with practising artists. We see ourselves as an international hub for many of the key discussions currently taking place within critical data studies. We have had a number of eminent visiting researchers, including surveillance theorist David Murakami Wood from Queen’s University in Canada; Max Hirsh, Research Professor in Urban Studies at the University of Hong Kong; science historian Orit Halpern from Concordia University, Canada; and, most recently, Marika Cifor from the School of Information, University of Washington, US, who visited us shortly before the outbreak of the coronavirus pandemic to speak about her forthcoming book on the AIDS epidemic, Viral Cultures. This international networking is central to the research group’s work and manifests itself particularly in the book Uncertain Archives: Critical Keywords for Big Data, which will be published by MIT Press February 2021. This brings together 61 contributors, drawing on the group’s work right back to when it was first established in 2015. Our work focuses on creating a theoretical vocabulary that will enable us to talk about and understand the uncertainties we are seeing in an increasingly datafied world Image recognition, chatbots and robot vacuum cleaners Two concrete research projects serve to exemplify our work, namely the ongoing PhD projects. Ece Elbeyi’s projectfocuses on ”conversational agents” – the chatbots that use artificial intelligence to conduct more or less coherent (and meaningful) conversations with us, and with which most of us are familiar in the form of online customer-service chat functions, but which are also increasingly found as standalone applications that aim to boost people’s mental wellbeing by functioning as a friend and conversational partner. We are helping to create the frameworks for culture-based data analysis in the long term Elbeyi is looking specifically at the user’s experience and the sensemaking taking place around the interaction with these forms of conversational agent, and her aim is to understand the social and cultural dynamics behind the interaction and the uncertainties that are generated. Naja Grundtmann is working on artificial intelligence, more specifically image recognition tools, from an art-historical and phenomenological perspective. Neural networks can today be trained to read semantic image content with a surprising level of accuracy. To a certain extent, neural networks can simulate how the human brain processes information, although it is not currently possible to show precisely how they identify and separate elements in an image. In other words, we can understand the architecture of a neural network in technical terms, but it is difficult to decide which aspects of a photograph or a pictured object are critical to classify the content. On the other hand, a number of different techniques make it possible to visualise an approximation of the information that, according to the neural network, is important in recognising motifs belonging to specific categories. Naja’s thesis examines how an art-historical perspective on these visualisations can contribute to a better understanding of artificial perception. Furthermore, we are working with media artist and cultural theorist Kassandra Wellendorf on a project that uses videoethnography to uncover links between humans and machines in everyday life – for example, by looking at how we interact with and talk about the robot vacuum cleaner or robot lawn mower. Ethics and epistemology By drawing on the aesthetic sciences and working in close collaboration with practising artists, the Uncertain Archives group strives to create an innovative and productive research environment in order to understand the uncertainties that other disciplines seem to agree are widespread in large, dynamic data archives. This enables us to facilitate the necessary interdisciplinary dialogue surrounding datafication and dataintensive environments where it is vital to understand the ethical and epistemological implications of new technologies. As such, we are helping to create the frameworks for culturebased data analysis in the long term.