Where Did the Agency Go
1 A feeling before a thesis
This talk started as a feeling rather than a thesis.
For a while I have had this strange tiredness in my practice. Not really tired of making things, and not really tired of ecology or technology or the subjects I keep returning to. More like a sense that the layer I thought I was addressing had moved somewhere else.
There is a familiar idea of the artist as someone who sees something, interprets it, intervenes, and then presents that intervention back to culture. And I have started to feel slightly suspicious of that picture. It assumes that there is a reasonably stable place where agency sits. The human subject, the artist, the institution, the audience. Somewhere in this chain there is a centers from which things are actually being decided.
And lately that centre has felt less and less convincing.
Part of the feeling comes from computation of course. Part of it comes from ecology. Part of it probably comes from middle age. And part of it may simply be that if you stare at anything long enough, the edges start dissolving.
Since this is my second presentation at ARF, I want to stay close to how I work. I going to move through the kittiwakes in Tromsø, the E. coli simulation I developed at Aldea, the porcelain point cloud, and the larger simulation I am working on now.
These works involve different methods. I intervene in existing situations, collaborate, construct models, translate between materials and representations, and revise things through observation. What connects them is a concern with how an arrangement shapes what can happen, and how what happens can change my understanding of the arrangement.
I will a also try to be careful about making the development sound too orderly. A talk is of course another construction. It is very easy to select the moments that lead neatly to the work I am presenting now.
And then suddenly the explanation I have now starts to sound like a plan I had all along.
2 The corridors of our minds
Before the PhD, I worked on several projects examining the relationship and borders between human ideas and other life. More than human, or whatever we want to call it.
My starting point was not primarily the idea that we should include more and more life within our moral circle. Not because I am against that. I was just more interested in agency. The borders between us and the rest of the world.
The title of my PhD became Architecture for the More Than Human. I never really liked the title, but I never changed it either.
Part of what bothered me about the more than human and Anthropocene discourse was how it could again centre human agency and perspective. We were the ones recognising, including, interpreting, taking responsibility. Even the attempt to move beyond ourselves could become another account of how important we were.
Donna Haraway’s idea of situated knowledge offers a useful point of dialogue here. It asks me to account for where I am looking from, and for the limits and responsibilities of that position. Announcing that I have moved beyond the human does not get me out of that work.
But I think I also let the title stand because more than human is part of an architecture we have built. A language architecture. It forms the corridors of our minds. It is a tool that forms the answers by shaping the questions.
More recently, I have been thinking about this through the idea of conceptual density. A concept gathers different histories, uses, expectations, and permissions. Several things are happening inside the same word.
A description also compresses. It preserves certain differences and leaves others out. We need that. I would like my coffee to remain coffee at least until I drink it. I would like a bridge engineer to have fairly stable concepts of load, stress, and collapse.
The difficulty comes when a description preserves something, and then behaves as though it has preserved everything.
We are sitting here while an extinction event unfolds. I find it difficult to believe that what keeps us here is simply a lack of information. It seems we can know quite a lot about what we are doing and still keep doing it.
There is something irrational in that. But I also wonder about the tools we use to understand the world, ourselves included. Perhaps they are forming the wrong questions. Perhaps we can reason very carefully and still remain inside a way of thinking that is part of the problem.
What if agency were smeared all over the place, not neatly contained? What if asking who possesses it already directs our attention away from how it happens?
This is where making becomes a way of thinking. I can make an arrangement, follow what happens, and find that the question I started with no longer quite works. The worlds matter, but so does the material resistance to neat concepts of the mind.
3 The birds had already acted
The kittiwakes in Tromsø offered a rare invitation.
They came to town and began using the building façades. Their arrival set off a domino of reactions that made our difficulty with uncontrolled elements very visible.
I remember the rhetoric as: kill them all. Even though they were endangered. It was us or them.
But the birds had already acted. Before we had agreed on what their presence meant, they had changed the situation.
What we called a façade had become, in practice, a possible cliff.
For several years I worked with kittiwakes nesting on buildings in Tromsø. Or perhaps that sentence is already too clean. I worked with birds, façades, property owners, municipal systems, conservation concerns, noise complaints, shit, scaffolding, nesting structures, architects, biologists, artists, neighbours, weather, eggs, chicks, bureaucracy, affection, and irritation.
At first glance the question seems practical: where should the birds nest?
But as soon as we ask it, different descriptions begin to govern the answer. For a biologist, the bird can be a breeding pair or a conservation concern. For a building owner, it can be noise, dirt, cost. For an artist, a collaborator, a disturbance, a sculptural client.
The same bird passes through several concepts without becoming identical to any of them.
Call it a pest, and removal becomes easier to imagine. Call it a neighbour, and coexistence becomes thinkable. Neither word settles what the bird needs or how people will respond.
My involvement was practical. Working with nesting structures meant participating in the arrangement of the situation. The structures were developed through collaboration, within circumstances that no one participant could determine alone.
In the writing I have been developing, I describe this as practical reclassification. Something is held differently, placed differently, or made available for another use, and a new set of relations becomes possible.
That is a way of describing the method retrospectively. The work itself involves much more particular decisions about materials, locations, people, and birds.
It also depends on what happens after a proposal is made. Calling a structure a habitat does not make it suitable. Calling the bird a collaborator does not mean the bird has agreed to my project.
The bird has digestion, fear, territorial behaviour, and absolutely no obligation to understand the exhibition text.
That is useful. There are consequences that my interpretation cannot settle for me.
4 What happens after the intervention
There is another part of this work that can disappear when we tell the story through the object: maintenance.
Someone has to clean, repair, monitor, negotiate, and return. These activities have to be undertaken or arranged. They do not vanish because the intervention has an artistic intention.
Mierle Laderman Ukeles’s work on maintenance has become a useful reference for thinking about this. It directs attention towards the activity that allows something else to appear stable, finished, or autonomous.
With the kittiwakes, the work includes the negotiation around the structure: who accepts it, who maintains it, who complains, who cares, who is allowed to decide where another species may live.
Collaboration is another word that can compress away who actually did what. A methodological account needs to distinguish my contribution from the work of the other people involved.
The birds also made our own behaviour visible.
What if we were animals?
Of course, intellectually, we know that we are. We share a common ancestor with chimpanzees. But if we looked at ourselves through the eyes of an ethologist, what would we notice? What is our habitat? What are our habits? How do we respond when another population occupies a surface we consider ours?
We could observe birds finding places to nest. We could also observe a human population reacting to a change in its habitat. I was part of that population, and my work was part of its response.
Uexküll’s concept of Umwelt offers another way into this: an animal’s world is structured through what it can perceive and do. For me, this becomes a practical question about a nesting surface, or about which differences a simulated organism can detect. It does not give me access to the bird’s experience.
The concept nature often arrives cleaned up. The bird does not.
The kittiwake shits on the fantasy that the city is a human interior with some decorative nature around the edges.
5 Returning to code at Aldea
An important turning point came during a three month residency at Aldea in Bergen in 2024.
I returned to programming and to making technical arrangements, things I had been involved with in my youth. It was a return to an earlier part of my own life, approached through the questions that had developed in my artistic practice.
At Aldea, I began building a simulation of E. coli.
The question of architecture had followed me down to the scale of a cell. I was thinking about the cell’s boundary as a kind of first architecture: an inside and an outside, with exchanges between them. A boundary that makes a particular kind of life possible through what can pass across it. A semi Permeable membrane.
I built a virtual Petri dish. The simulated bacteria moved, fed, reproduced, mutated, and died. There was food and waste, a temperature gradient, and introductions of food and antibiotics. Individual agents had inherited parameters affecting their behaviour.
There were no bacteria in the room. There was code, a screen, and two dot matrix printers.
The simulation was a model, a diagram, a toy world, a machine for producing behaviour. It borrowed from biology, but the behaviour of the model could not establish biological claims about E. coli.
Most of the organism had been left out. The wetness. The molecular machinery. The actual history of its relations with the world. What remained was a small behavioural diagram.
A world is reduced until it can run.
That reduction is an explicit part of the method. I had to choose what the agents could sense, how they could respond, what could change, and what would remain outside the model.
Once the simulation was running, patterns and individual histories developed through those interactions. I constructed the conditions without separately writing every sequence of events.
At some point, I started watching it with the wrong part of my brain. Not exactly empathy, perhaps, but attention. A small narrative began to attach itself to the dots.
6 Life becomes a line of text
The exhibition brought another translation into the arrangement.
The two dot matrix printers produced streams of text connected to the simulated bacteria. The language was tied to internal states, circumstances, and genetic makeup. Individual bacteria were named in relation to their generation and genetics.
The narration followed an individual. When it died, the text moved on to another.
Paper kept accumulating.
Life became a line of text. Death became a line of text. Mutation became a line of text. And the printer keept on printing.
What interested me was the movement between the running system and the account we encountered. A changing computational state became an event in language, and language became a physical trace in the room.
What makes us begin to follow an individual? What changes when a set of values acquires a name and a narrated history?
The work gave me a way of examining those questions through an arrangement. I was constructing both the behaviour and the language through which it could be encountered.
It also gave me several places to work: the model, the narration, the printers, and the relation between them. The translation could be examined alongside what it translated.
The printed account could not establish that a bacterium had an experience. It could make our readiness to recognise experience as part of the encounter.
The work didnt come from a vaccum, the 3 moth recidency, quite the rare opportunity for a father of four, there was the social element at Aldea too. Toril Johannessen had a studio upstairs. I had frequent conversations with her, Cameron MacLeod, and Magnhild. Discussions moved between practical challenges, artistic practice, politics and questions about models, representation, and what we do.
Those conversations became part of the conditions in which the work developed.
Perhaps a residency is itself a temporary container. People, tools, habits, unfinished works, meals, doubts, and borrowed ideas are gathered inside it for a while. But it is a leaky container.
The Petri dish was not the only container in the work. There was also the screen, the printer, the exhibition room, and the residency. Each held something temporarily. Each changed what it held.
7 The staircase that could no longer carry anyone
Another path into this comes through a staircase.
In I Was Here, a point cloud scan of an architectural fragment became a one to one arrangement of around two and a half thousand porcelain pieces, suspended on threads.
A staircase is usually a very practical concept. It moves bodies between levels. Up, down, step, landing, rail. One foot, then another. It teaches the body a small vertical grammar.
In the installation, that function is unavailable. No one can climb it. Something remains recognisable, but the thing that made it obvious has been removed.
This is another method I keep returning to: translating something between forms and attending to what changes in the passage.
Architectural space becomes measured points. The points become material objects. The objects require suspension, placement, and labour. A representation that can be moved almost effortlessly on a screen becomes something that has to be made and installed, piece by piece.
The gaps are present along with the points. The surface has become discontinuous. The installation occupies the room while allowing sight to pass through it.
Translating the scan into porcelain lets me examine what it carries across, what it leaves out, and what happens when its points have to hang in a room.
The scan is not the walking. The porcelain is not the stair. The fragment is not the building. And yet something has been carried across.
If I call it a staircase, I preserve its history of function. If I call it sculpture, I move it into another form of attention. If I call it archive, I emphasise memory and trace.
The thing waits inside several names.
This is where the idea of conceptual density becomes useful to me. The work can remain available to several descriptions, while its material arrangement gives us something specific to encounter.
Objects are less obedient than concepts. They keep a kind of low level resistance. The staircase does not become sculpture without remainder. Sculpture does not fully absorb the staircase. Something creaks between them.
And that creak is interesting.
It also keeps the method grounded in making. I have to deal with the work as an arrangement of pieces in a room, with all the practical demands that come with it.
8 The silicon bedtime story
It happened that the year I started my PhD, 2023, was also the year large language models entered widespread everyday use. They had crossed a threshold. They could write personalised birthday songs for your mother in law. And code. Lots of code.
Suddenly there was another contender in the debate about who, or what, has agency.
The easy version of this anxiety is the silicon story.
We build machines that become better and better at modelling, selecting, designing, and optimising. Increasingly, those systems help decide how the next systems should be built. Perhaps, eventually, they do that more effectively than we can.
The nexus of control migrates.
At that point perhaps we become something more like the bacteria in our own guts. Important. Necessary even. Still processing things the larger system needs. But no longer the obvious place where you would locate the agency of the whole thing.
This is a very good story. It has a before and an after. It has a protagonist, tension, a coming threshold, and the possibility of apocalypse, which we humans seem to enjoy tremendously.
I have started to think of it as a bedtime story. It may be partly right. But it is suspiciously well shaped.
It says: agency is moving from us to them.
That is frightening, but also strangely comforting, because it preserves the idea that there was an edge to lose.
A decision that feels like mine already involves hunger, hormones, memory, language, institutions, available tools, money, and whatever somebody said to me fifteen years ago.
Perhaps the autonomous human subject is a useful local model. A bit like an icon on a desktop. You click the folder and it behaves as if there is a little folder there, even though what is happening underneath is much stranger.
The new technology makes these questions difficult to leave outside the studio.
9 A game fully capable of playing itself
Which brings me to the simulation.
I wanted to make something that broke with what I expected of my own work. A computer game without a protagonist. A game fully capable of playing itself.
We could join, more like voyeurs, watching from the sidelines. Things would already be happening. Our arrival would be one event among others.
We are accustomed to entering a game as the reason its world exists. The landscape provides our resources. Events become our challenges. Other life becomes helpful, dangerous, or decorative.
I wanted to see what happened when the world could carry on without us.
Perhaps we would spend some time watching before deciding what anything was for.
This places the work in dialogue with artists such as Hans Haacke and Pierre Huyghe. In Haacke’s Condensation Cube, water responds to the physical conditions around the work. In Huyghe’s Untilled, living organisms participate in an environment whose developments are not individually scripted.
My simulation operates through implemented rules, so these are different kinds of construction. The methodological connection I see is the act of arranging conditions and then attending to processes that unfold through them.
The conceptual answer is only part of why I make it. The other part is that it is delightful to try to model complicated things.
How does water behave in porous material when temperature crosses a phase boundary? How do you make snow accumulate and then release? What happens when limestone dissolves under terrain that already has trees growing on it? What happens when the support disappears?
At some point, these questions become demands on the construction. It has to contain relationships through which one process can affect another.
Take the loss of support beneath a tree. If the tree simply stays upright because it belongs to the scenery, the apparent world and the implemented relationships diverge. Making the tree respond means connecting processes that could otherwise remain separate.
I specify rules and relationships without individually scripting every event their interaction produces.
An outcome can surprise me while remaining dependent on the model’s construction. Perhaps processes are interacting in an interesting way. Perhaps an assumption has consequences I had not understood.
Perhaps it is a bug.
I need to work out which.
I do not want every broken thing to become a philosophical revelation. Sometimes I spend three days fixing something, and what I have at the end is something that works.
But the practical work makes me more specific. Water moves through the ground. Fine. Where? Under which conditions? What happens when the material around it changes?
The computer is completely indifferent to how convincing the concept sounded in the application. That is part of the pleasure too.
10 Encountering before naming
The question of language remains inside the simulation.
I have been moving towards withholding familiar material names and numerical properties from the player’s interface. A mineral can appear as an unknown mineral, with a description of its qualities. The user can give it a name, at the present time that means me.
The underlying model still needs numerical distinctions. But a user can encounter those distinctions through how things behave and through the descriptions available to them.
I want the user to feel the difference between a material and the name given to it. The category is something we bring into the encounter. It helps us, but it does not exhaust the thing.
This is an interface decision with a methodological purpose. It gives me a way of investigating how naming organises attention and expectation. Whether it produces the experience I hope for is a question to examine through encounters with the work.
The title, God vs Evolution, belongs to the same problem. It is the slightly bonkers name that inspired the project years ago. It has been sitting on the shelf collectiong dust for over 15 years in fact.
In my reading of it, god is us: the desire to name, arrange, intervene, and understand where control sits. Evolution is the activity that keeps complicating that picture.
Maybe we are made in gods image or maybe god is made in our image.
But my own position is implicated. I am making the world, choosing its rules, deciding what the player can see, and asking the player to question the authority of categories.
I cannot solve that contradiction by hiding the numbers. I can make it part of the work.
The simulation gives me a place to work with those decisions materially, through code and interface, and to encounter what they do. The question becomes specific enough to change a button, a description, a behaviour, or a relationship between processes.
11 The tools enter the question
AI is also involved in how I make this world.
I use language models to help write code while thinking about what their participation does to my understanding of authorship and agency. I think its maybe the most problematic tool Im using, and worthy of a talk on it own, I think ill suffice to say that in the current political and increasingly technofeudal environment, as Janis Jarufakis would name it, anything an everything can be enshittifed, its hard to imagine any novel invention not beeing subverted into a shitty subscription service. The biggest issue I take with the widespread use of LLM based services, is the reductive nature, the idea that we can chew thorugh all of humanities compatible knowledge, and spit out a instantly human language searchable database, like freeze dried instant coffee.
The thing that appears to be a subject of the research is part of the working process.
I choose, evaluate, accept, reject, and revise. But the relationship between an intention and its implementation has changed. I can ask for something whose implementation I could not have produced at the same speed on my own.
The response can introduce possibilities I had not considered. It can introduce mistakes I have to learn to recognise. This makes the location of my judgment a practical question. What do I understand? What am I relying on? Where can I intervene effectively? What happens when the scale of what I can produce grows faster than my capacity to inspect every part of it? I encounter those questions while making the work.
Using AI changes how activity is distributed between intention, modelling, implementation, and evaluation. Saying that I used AI explains very little by itself. The useful questions concern what I delegated, what I judged, what I changed, and what I could verify. The system can generate code. It does not remove my responsibility for what I present as my work, or for the claims I make about it.
This brings me back to the distinction I need when talking about agency. Contributing to an outcome, acting intentionally, and being responsible for it are different things.
Calling agency distributed should help me distinguish those relationships. Otherwise, everything is connected becomes a very effective way of saying almost nothing.
12 The methods
Looking across these projects, I can describe the methods more precisely.
With the kittiwakes, I enter an existing situation through collaboration and intervention. The birds have lives and histories independently of my project. The work develops among participants whose activities I cannot determine.
At Aldea, I construct a model and a system of narration. I can examine the passage from implemented behaviour to a printed account, and my own response to that account.
With the point cloud, I translate between spatial measurement and material reconstruction. The process makes the selections and changes involved in representation physically present.
In God vs Evolution, I construct interacting processes and work on the interface through which they are encountered. Iteration happens through the difference between what I intend, what the model does, and what becomes available to experience.
Reflection moves through all of these activities. It happens in conversations, writing, technical decisions, and the repeated act of returning to the work. Some of the language I am using today was developed afterwards. It helps me understand relationships between projects, but I should not pretend it was a plan I possessed from the beginning.
The methodology lies in why I bring these activities together. I use making to give assumptions consequences, and those consequences give me something to examine and respond to.
An unexpected event is a starting point for investigation. Its significance depends on what I can understand about it, how it affects the work, and what decisions follow.
Sometimes the result is that a category I began with no longer seems adequate. The work has given me a reason to reconsider it.
The distinctions between projects matter. Simulated behaviour is evidence about a construction. A bird’s response belongs to a living situation. My intention for an interface is different from another person’s experience of using it.
The contribution I can make is an account of particular arrangements and what working through them has made possible to notice. That account becomes stronger when I can show the decisions, translations, and consequences on which it depends.
It remains open to disagreement. Someone else may look at the same work and notice precisely what my account leaves out.
13 A practice becomes a sentence
There is one more arrangement I need to consider: the artistic practice itself.
I can see the role available to me. Northern ecology. Slow attention. Deep empathy for nature. A few wooden sculptures every year. Elaborately written and drawn notebooks. Slightly withdrawn. Not too talkative. Beard or a moustache at least, tweed shirt.
It is not even a bad role. I could inhabit quite a lot of it sincerely.
That is what makes it dangerous.
We the art world seem to reward this. I make work. Certain parts are understood. Those interpretations generate invitations, funding, texts, and contexts. You notice what travels. You emphasise some traits. Other things fall away.
Eventually the description of the practice starts feeding back into the practice.
A practice becomes a sentence.
Then the sentence starts making the next work.
At some point you are standing in a forest wearing wool, talking to a German curator about lichen, and wondering how the fuck you ended up here.
This is another reason the return to code. It reopened a part of my interests that the description of my practice did not contain, and it was dirty, and it deal with one of the most problematic phenomena of our time, apart form the sixth mass extinction.
Of course, deliberate incoherence could become another identity immediately. Kåre Aleksander Grundvåg resists categorisation through radically heterogeneous methodologies.
There. Fuck. We branded it.
I dont know, I would like to leave enough room that I can still surprise myself. Enough continuity that it is a practice, and enough deviation that the next work does not have to protect the explanation of the last one.
I can describe where a decision came from without turning it into an instruction for everything that follows.
The title I never liked can remain a question. The categories can remain useful without becoming the architecture of every possible future work.
14 The smoke barrel still needs checking
And then, unfortunately, there is the final joke.
After spending all time resisting the predictable figure, I found myself sitting on the porch in early autumn. I was eating homemade bread. I had espresso in a coffee cup I made myself from coarse local clay and fired in an open bonfire. Seaweed gathered the day before was being smoked nearby. There was a little thimble of Finnish Jaloviina, brought back from fieldwork the week before visiting a nature restoration projects in Lapland. Machine learning systems were running around me. One was continuing the programming of the simulation. Another pair was busy self training on hardware installed as a rather expensive heating solution to my studio space.
I needed to go check the smoke barrel.
And I was wearing a black and red tweed shirt.
Apparently, the thing I was trying not to become had already happened. Only it had become contaminated by GPUs, code, fermented seaweed, espresso equipment, bonfire ceramics, simulation bugs, cloud storage, and whatever else happened to be lying around.
Maybe the problem was the demand to resolve it into a clean interpretation.
I keep returning to perspective. Cell, organism, bird, institution, geological layer, algorithm, artist, audience. Move the frame and the meaning moves with it. I do not think there is a final perspective waiting at the end. Not even an ending, maybe many, maybe none.
But changing perspective can have practical consequences. It can change where we build, what we model, which differences we preserve, and what we allow to affect the next decision.
Perhaps, in the end, we might realise we are ourselves more than human. The boundary of my body is real, but it does not contain all the conditions that keep me alive.
We want to live. Around us, other lives are continually doing what keeps them going. Everything pulsing with life at this moment continues an unbroken lineage reaching back billions of years. The kittiwake on the façade. A bacterium. You and me.
References for discussion
These references accompany particular questions in the talk. They provide points of dialogue rather than a chronological account of influences. This appendix is for discussion and follow up reading.
Donna Haraway
Haraway, Donna. 1988. Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective. Feminist Studies 14, no. 3: 575 to 599.
Section 2. Partial perspectives, the conditions of knowledge, and accountability for one’s position. Used as a point of dialogue with the critique of how more than human language can function in a practice.
Jakob von Uexküll
Uexküll, Jakob von. 2010. A Foray into the Worlds of Animals and Humans: With A Theory of Meaning. Translated by Joseph D. O’Neil. University of Minnesota Press.
Section 4. Umwelt and the relations between perception and action. The manuscript’s application to simulated agents is an analogy, not a claim that computational detection establishes an organism’s lived experience.
Mierle Laderman Ukeles
Ukeles, Mierle Laderman. 1969. Manifesto for Maintenance Art 1969! Proposal for an Exhibition CARE.
Section 4. Maintenance as an artistic concern and the continuing labour that supports an intervention. This connection does not imply that the artist personally performed every maintenance task in the kittiwake project.
Hans Haacke
Haacke, Hans. Condensation Cube, 1963 to 1965. Clear acrylic plastic and distilled water. The Museum of Contemporary Art, Los Angeles, accession 2008.69.
Section 09. A constructed physical process that responds to its surroundings. The work offers a comparison with the simulation’s implemented causal relationships.
Pierre Huyghe
Huyghe, Pierre. Untilled, 2011 to 2012, presented at dOCUMENTA (13). Related film: A Way in Untilled, 2012, colour HD video with sound, 14 minutes.
Section 09. Environments involving living organisms and developments that exceed individual scripting. The installation and the related film are distinct works; the spoken comparison refers to the installation.
Installation described by Marian Goodman Gallery · Film in the Pinault Collection