Fieldwork with the Interacting Minds Centre (Aarhus)
A few weeks ago, I went to Aarhus, the second-biggest city in Denmark and the biggest Scandinavian University, to attend the first days of an experimental course on AI writing with the UC Davis-based anthropologist of science Joe Dumit. I always wanted to get a sense of how Joe teaches, as his seminal explorations are slowly but surely becoming prominent in our field. I was more than eager to accept his invitation to partake in a collective examination of AI writing informed by his long engagement in the field of science and technology studies, computer science and performance studies. Besides, I was interested in seeing another vastly interdisciplinary workplace at the IMC, where Joe is now enjoying a part-time professorship. Co-founded by anthropologist Andreas Roepstorff in 2012, the Interacting Minds Centre for the Study of Cognition, Communication and Choice is a focusing on the “specific abilities for interaction (that) are key to being human.” Its agenda brings together cognitive scientists, medical anthropologists, neuroscientists, phenomenologists, dancers, and more. They have an ongoing chess game in which a rule is changed at every turn. They have cool borrowed exhibits from the local ethnography museum. They have VR headsets and cooperation with world-class artists. They have an anthropology library with a good proportion of French titles. They have a piano. And now they have Joe Dumit.
How can we make use of the super-abilities of the newest Large Language Models (LLM), instead of pulling away from them? Joe frames his course on AI with the notion of “co-creativity.” The summer teaching programme is also co-creative in the sense that it is funneling the reflections of a number of researchers of the IMC who are currently exploring the possibilities and limitations of LLMs in education and research. With 35 undergrad students from very diverse backgrounds, the course is fully booked. After an introduction by Joe, we dive with the computer scientist Arthur Hjorth into a series of exercises. We challenge each other and the currently most famous algorithm, GPT, using an interface that Arthur is presently developing. The tool helps students to keep track of their prompt iterations with the AI. It gives them the opportunity to reflect on prompting interactions, while collecting the data for the research. The first exercise is rather simple: how do you get with a single prompt to a predefined text with a single prompt? How can you get close to a specific target output? The target output is a short story, involving three pigs and a wolf. And the task isn’t so easy. In the second part of the day, Joe takes over again and points to the creative possibilities opened by the tool: How can we use it to overcome a personal challenge? Joe speaks to us of another of his favourite notions of the moment: slowness. What varieties of slowness are to be supported by the deployment of these new tools? We are invited to develop solutions using ChatGPT to take up one of these challenges.
At the lunch break, Andreas tells Joe and I about an interesting event that is happening at the same moment, an hour away from Aarhus. A summer course at a Buddhist meditation centre, the Gombe Hoejskole, that ambitions to bring together the Norwegian poet and social activist Grundtvig and the teachings of the Buddha.
The first time I read about Andreas was in a newsletter of the Mind and Life foundation. Set up by Francisco Varela just before he passed away prematurely, Mind and Life is fostering investigations on the relations between contemplative practices, neuroscience and society. The summer course at the Gombe Hoejskole is modelled on their retreats, Andreas tells me: a healthy mix of sitting, workshopping, moving and networking.1 “I hope you don’t mind, he also says, I told them about your drawing documentation practice to get you invited too—it would be nice if you could do a few sketches.”
After the class that day, we take a car from the university and we drive through the most beautiful hilly seaside I have ever seen, the Mols Bjerge. Andreas tells us that this land is said to have inspired Tolkien to write about the hobbits’ land, the Shire. The inhabitants of this land, the Molbos, are also reputed for undertaking startling collective actions which defies common understanding (ethnic jokes pass them as a bit stupid). Andreas tells us a typical instance: one has the idea that something must be done to scare away the birds that are eating away the rye. But someone objects that going around the crop would damage it. They end up going around as a group in the rye, carrying a man on a door, who would gesticulate off the bird. I can feel his fascination for the mystery of these interactions that make us truly human.
We arrive at the Gomde, signalled by gorgeous red Tibetan flags. Gomde means temple in Tibetan, Andreas tells us. We are greeted by Jacob. Jacob is a yoga teacher and bodywork practitioner. As I will learn later, he is currently applying for a PhD program at the IMC to study the operating paradigm of a Danish bodywork technique. We hear some singing. We join the seminar. They are assembled into a circle in the courtyard of the old farm. The song is from Grundtvig. I can’t understand a word but Andreas presents me with a printed booklet and invites me to sing along. It feels nice. Colourful prayer flags (“horses of the wind” in Tibetan) are flapping in the Nordic brisk air.
Nikolaj Frederik Severin Grundtvig (1783–1872) was a central figure in the development of “folkehøjskole” (folk high school) movement, which aimed at providing education for adults, particularly those from rural and working-class backgrounds. These schools focus on personal growth, cultural awareness, and active citizenship rather than formal vocational training. Grundtvig emphasized the importance of the “living word” and Christian community over the written Bible, which contrasted with more traditional Lutheran views of the time. He also made vital contributions to Danish literature, including hymns and poetry, and translating old Norse sagas into Danish. The group of fifty people strikes me as the usual crowd of meditation centres across Europe. A nice and welcoming gallery of characters. The sea is a ten minutes’ walk away, Jacob takes me to have a look at it from the top of the hill.
The temple is freshly painted and the mix of bright yellow and red shouts out into a joyful atmosphere. We sit in the back on meditation mats. Joe and I get a little transmitter from Jacob, he will be translating for us in a yellow booth. This workshop is led by Rikke Ulk, an anthropologist and a friend of Andreas. Joe tells me that all anthropologists know each other in Denmark. He gave several keynotes at their “mega seminar”—the community assembles every other year at their foremost anthropology museum, where they present, share and dance together.
Rikke Ulk tells us about places and how to bring climate awareness to society with attention to “where” we are and “where” we come from. Her approach includes diverse projects and initiatives, with a generous amount of podcasts and Instagram stories. In the language of today, she is decidedly multimodal-inclined. We then share in pairs and scribble on post it the distance from home, and what our meaningful tree feels like. After another song outside in the sun, we are offered coffee and shots of juice. I show around that drawing I just did on the iPad, contributing modestly to the happy ambiance. Rikke inspires me further, telling me about her projects with Danish institutions conducted with her company, an anthropological research agency. At Mitte Marie, a progressive centre to rehabilitate substance users with a mental condition, she conducted with her associates the documentation of their many experimental practices.
After the break, it’s the turn of Andreas to lead a workshop inside the temple. He is jumping in to replace another speaker, who couldn’t make it that day. Andreas and us, “his friends,” are invited to take place on red cushions in the front, at the feet of a huge brass statue of the Padmasambhava Buddha. Bring forth the context of his research within the Experimenting Experiencing Reflecting project with Olafur Eliasson, he tells us about the work of Helene Nymann, an artist currently writing her dissertation on memory under his supervision, and of her fascination for the hippocampus, the organ within the brain that seems to bring together the past and future. What should we remember to remember? What kind of ancestors do we want to be? Andreas asks the audience to close their eyes and summon memories of the future.
After an early dinner at the Gombe, Andreas drives us to our next destination: an art and science residence centre called “Earthwise,” led by his friend Christine. The beautiful farm estate hosts musicians, dancers, writers and other shades of artists and activists to help society to “reconnect with the living world.” There we meet fluffy chicken, eggplants grown into a geodesic dome, bees buzzing into their “night song.”
We sit at the table surrounded by music instruments, we eat organic yogurt with a big layer of cream from the farm down the road, and drink an earthly tea in earthen pots. They speak of old friends, new partners and future plans. I grab a string instrument from the window still. It’s a mouth harp, Christine says, and she plays it for us. “Wait, she says, let me introduce you to my Lamborghini.” She disappears in the next room and comes back with a sleek metallic instrument: a mouth harp from Mongolia (see lower right part of the drawing). Last time I played it was at the funeral of a friend, she says. It’s still full of this energy. She brings it to her teeth and transmutes into a rhythmic elemental being, breathing and sounding with springy tones. We want to try. Three more of these instruments appear on the table (the Lamborghini is quickly tucked away) and the three of us attempt to enter into a similar sound trance. We part with hugs and promises of coming again soon. As we drive back to Aarhus, the Nordic summer evening is fading out the landscape into dusk.
Rebekah Brita Baglini is teaching the third morning at the AI class. She invites the group to “out prompt” the most popular anti ChatGPT system. Trained in Chicago and in Stanford, she is a linguist specialized in natural language (semantics and pragmatics). Recently tenured in Aarhus University, she co-funded the Center for Language Generation and AI. She presented us with a thorough review of the landscape of university reaction to the LLMs hype. The current model is built around two variables: perplexity and burstiness. Perplexity measures the complexity of the prose, while burstiness checks on the consistency of the sentences. Less complexity and more consistency lead to a high rating on the AI generated probability score. These new tools, which claim to sort out student essays written by an AI, are not only easy to fool. They can also cause a vast array of terrible consequences on the academic system.
“As you may have noticed from my accent, I come from the US. To spice things up, I would like to introduce an element of competition for this exercise.” The participants take up the challenge: we all write up short stories on the “dangers of AI,” which are immediately assessed by the GPTZero algorithm on the platform developed by Arthur. And then we reiterate: how can we out prompt it to get scores to zero while using GPT to generate the essays? The more straightforward techniques used by the students involve asking the bot to incorporate grammar mistakes—probably not the best tactic to get a good grade. The most advanced include prompting the AI to enact a “persona” (i.E. “you are a feminist activist”), requesting unusual styles (i.E. “in the style of Hemingway”—didn’t work so well, not “bursty” enough), or introduce specific juxtapositions of fresh material which the AI will assemble (such as “Donna Harraway and Karen Barad are discussing the dangers of AI”). Kevin wins with a GPT generated essay marked as a 0.0003% probability by the algorithm—yet difficult to say by which means exactly (“Maybe he just got lucky”).
As the demonstration shows, these systems are not only inefficient: they are black boxed and expensive. Administrators love to adopt costly infrastructure which claims to fix wicked problems with straightforward engineered solutions. During the break, Andreas and Joe share with us about their long experience as faculty (and their fieldwork as STS researchers) on that. These models bring the opportunity of moving the authority to evaluate papers from professors to IT departments, who will be the first assessors of the quality of the work. As Rebekah points out, they will inevitably prompt students to write first for the classifiers established by these corporations and what they have labelled as “AI generated text.”
Joe concludes that day with a memorable move: along with slow sentences and simple words, he draws with chalk on the blackboard the crucial matter of it all. Is this really “student writing” vs “AI generated text”? Isn’t there something in between that we need to care for? “We will write a report on what we are doing together, and you will have the opportunity to sign this report as a coauthor.” Because Aarhus University, like many other institutions, are already lining up to adopt the novel surveillance tech, we are here researching collectively and taking a position on an immediate urgency. We—the participants—are given a chance to have an impact on the future of their evaluation system, and of the education at their university. Andreas shares his impressions after the class: the students seemed baffled to hear about their possible involvement. “It shouldn’t be so strange that the essays of the students should shape the future of their own education system,” Joe shares with us.
On the train the next morning, as I drafted this field note, I delved on the richness of a deeply en-sketched approach to interdisciplinarity, the deployment of multiple logics which fold into each other, unyielding pragmatic concerns and darting responses to them. Moreover, it felt a bit like a dreamland. How much of this is real? The Danish down-to-earth and no-bullshit attitude helps to make it highly credible. Also, the food was great.
PS: I tried to ask the AI to co-create this note and it botched the job. In the end, I did use the bit on Grundtvig that it suggested. It will take a longer time to achieve a co-creativity workflow, but I do believe with Joe that the super-ability of the AI—guessing one word at a time—can help us not only to deliver on some of our inventive work but also to understand better what we are and how we write.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)
