Learning session 16: Ecological Intelligence for AI
Online learning sessions using dialogue to explore the workings, developments, (im)possibilities, and emerging insights of diverse approaches and praxes for Co-creation with the More-than-Human world (Can’t read the full Medium article? All learning session syntheses can also be found on Substack.)
🎧 Prefer listening? Just click the built-in “Listen” play button right below the title at the top of this page
The 16th learning session of the Co-creation with the More-than-Human sandbox, held in September 2026, explored Ecological Intelligence for AI with Anab Jain, co-founder of Superflux, the London design and futures studio she runs with Jon Ardern, while she also holds a position as Professor of Design Investigations at the University of Applied Arts, Vienna.
The session moved through two connected pieces of work. The first, the Ecological Intelligence Agency, was commissioned by the UK Policy Lab and Defra Futures in 2023 to help imagine the future of freshwater decision-making out to 2043. What Superflux built was a speculative, commons-owned AI advocating for the River Roding, just outside London, and for every species entangled with it.
The second, Nobody Told Me Rivers Dream, grew out of what the studio noticed while making the first: sculptural sensors installed along the river Thames, shown at the Design Museum in London’s More than Human: Making with the Living World exhibition in 2025. And moving the question from AI-as-advocate to AI-as-listener.
The session’s dialogue examined what changes when an AI is asked to advocate for a river, what “ecological intelligence” in this context is actually made of, and where the open questions around guardianship, values, and accountability sit once the provocation moves from speculative proposal toward something institutions might one day take seriously.
What follows is a synthesis of key questions, insights, and cross-cutting threads from the dialogue.
🌿 Learning Session 16 (September 2026) — Synthesis
Ecological Intelligence for AI, with Anab Jain (Superflux)
What if AI stopped trying to know everything, and started knowing where to listen?
1. Understanding the work
A future you can step into
Superflux calls itself a studio of archaeologists of the future, working in installations, objects, films, and soundscapes rather than reports. The clearest example of the approach is a London flat, built by Superflux to imagine the home Anab’s son might live in around 2050: ordinary at first impression, until you turned a corner, revealing half of it given over to food production, mycelium grown into computers, a purple fog drifting through the room. No data appeared anywhere in it, no graphs, no mention of climate change, only a radio narrating the morning news.
Several hundred thousand people walked through it. “The future for so many people stopped becoming an idea and actually became a memory. Which is quite powerful, as opposed to this very popular graph we see about climate crisis and global warming.” A graph follows a trend line; the apartment gave people a lived life a graph cannot.
Why the studio reaches for AI at all
Behind the choice for AI sits a more direct claim about what current tools miss: “the relational complexity of what we are facing exceeds any framework’s ability to capture it fully, and it’s not the problem of the framework as much as the nature of where we are right now.” The point isn’t that the tools are badly built. It’s that kinesthetic, sensory, embodied forms of knowing are themselves forms of intelligence, and most frameworks have no room for them.
Why AI for this brief
The brief that led to this work came from the UK Policy Lab and Defra Futures, who issued a call to help imagine the future of freshwater policy in 2043. Superflux’s own long-running interest in rivers, and in the pollution of the Thames specifically, was already there when the call came out.
AI, and in particular the Ecological Intelligence Agency, became the answer to the brief because it was already reshaping policy in this space. Superflux saw a narrow, critical window of imagining other possibilities for AI before that direction hardened into something fixed. Rather than merely stick AI to any app, there was an opportunity to rethink how we do AI, and keep this cone of possibility open. The starting question therefore was blunt: what if AI spoke for our rivers?
What an AI advocate might hold that a human one can’t
Rather than a tidy list of comparative powers, the case made moved into a larger frame: uniting labour, climate, and data justice in something that isn’t extractive. The reference point offered was Te Hiku Media, the Māori broadcaster whose entire te reo Māori speech archive is held under a shared guardianship model that treats data as land, not as a resource.
The Ecological Intelligence Agency borrowed that spirit rather than its exact form: a system locally owned by the community and by the river’s own legally recognised entities, able to hold the perspectives of the many smaller, often invisible lives entangled with a river all at once, the insects, the fish, the wider ecosystem, in a way a single human advocate has to compress into one voice.
🌱 Seed:
Speculative design’s power sits in the gap between knowing something and feeling it. A tangible future closes that gap in a way a statistic, however accurate, tends to leave open.
2. Ecological intelligence, and the broader idea of AI
Relationship, not data
Ecological intelligence, in Anab’s account, lives in our relationship with the more than human world, in a long practice of learning to live with the more-than-human world, carried and refined across generations, not in the data an AI happens to be trained on. Later in the dialogue, when several people in the audience described work they called collaborative or relational intelligence, her answer condensed the whole argument into one line: “I would say all intelligence is relational.” Not a case for layering more environmental data onto an AI system that otherwise works the same way underneath, but a claim about what intelligence is in the first place, and what, or who, a system answers to.
A sensory language
Ecological intelligence in that sense acts also as a sensory language, with examples drawn from lived experience: peacocks growing loud and restless before monsoon rain in Gujarat, howler monkeys signalling change in Brazil, the halo around the moon read as weather further north. Communities who say the wind has “died” are picking up a register most people educated inside Western, Cartesian frameworks have simply lost the practice of reading, something the philosopher David Abram has written about as the wind having its own dialect.
Extinction is not the right scale
Foundational AI models have already put a version of this question on the table at the industry level: the idea that a sufficiently powerful system could put humanity itself at risk of extinction. Ecological thinking doesn’t sit outside that conversation, it changes the terms of it.
“Maybe extinction is not the right scale of the question, because it puts us in opposition to the world”, Anab said. “We always take this combative posture, and that is an act of separation.” The deeper issue, in her account, is that enormously powerful systems are being built inside institutions never designed to govern power at that scale, and the extinction framing gives AI more agency than it deserves: these systems don’t appear by fate, they’re financed, designed, and deployed through specific choices made by specific people, which echoes questions about machine agency the scholar Lucy Suchman has been raising for close to forty years.
There’s also a question of whose extinction gets treated as a future hypothetical: for people already facing the flooding in Nepal, something close to it is a present reality, not a scenario. That points to a further question about whose sense of “humanity,” built on ideas of mastery and intellect, is being protected in these debates, even as that same humanity keeps building the most powerful machines in the world.
🌱 Seed:
Existential-risk debates tend to imagine one shared humanity facing one shared threat. Anab’s argument was that both the threat and its distribution are already unevenly shaped by choices, institutions, and geography.
3. The Ecological Intelligence Agency — how it’s imagined to work
The unsolved problem of “I don’t know”
Most AI is built to produce an answer, even a hallucinated one. Building a system able to say it doesn’t know, instead of sounding certain, remains unsolved: “this can only be done if we are able to prototype it and show, then we can have that conversation,” Anab said. “What we are trying to do is, in a small critical window, to show what else AI could be.” The Agency was never presented as a finished answer to that problem, only as a way of holding it open.
What values, and from where
In the working demonstration Anab described, a policymaker places a water sample on a sensor and the system responds as if the river itself were speaking, flagging what it finds: elevated algal levels from wastewater discharge, or a different pollution story with every sample, agricultural run-off, industrial waste, microplastics, heavy metals.
The values behind those responses weren’t explicitly designed in, and Anab was candid about it. Rather, this first version surfaced open-source, locally held data points that already existed, widening what was in front of policymakers rather than resolving into a built decision-making tool, a speculative storytelling exercise aimed at a specific audience, not a working system.
Three questions, wherever it’s applied
The same test recurred whenever the conversation turned to applying this elsewhere, from farmland in India to a river like the Colorado, drained before it reaches the sea: who benefits, who has given consent, and what form does reciprocity take.
Get Renilde Becqué’s stories in your inbox
Join Medium for free to get updates from this writer.
🌱 Seed:
Before an AI can advocate for anything, someone has to decide what counts as evidence of a river’s interests. The Agency didn’t answer that question so much as put it in front of policymakers.
4. Nobody Told Me Rivers Dream — sensing and perception
What the sensors do
Three sculptural objects sit along the Thames, built from local materials and running on open-source electronics and Raspberry Pi hardware. A birdsong sensor traces the edges of a heron’s cry or a kingfisher’s whistle. A water-flow sensor, its form arching toward the river, reads tidal rhythm and flow speed. A wind sensor and a weather sensor read atmospheric pattern. They’re deliberately strange to look at rather than disguised as ordinary tech, and more fragile than expected: “they got so beautiful and delicate, in all honesty,” Anab said, “we were like, oh my god, they’re going to collapse in the river on the second day.”
Guardianship, raised and left open
While describing how the sensors work, Anab raised an unresolved question herself, unprompted: who holds the data they collect, and on what terms. “These thorny questions of guardianship, where does this data live, who is the guardian, who is the steward of that data,” she said. “I’m very open to suggestions.” What has taken shape instead is a conviction that not everything gathered this way should become fully legible to a system, that a degree of opacity can be protection rather than a gap to close, given how readily ecological and indigenous knowledge gets extracted once it’s made available at all.
Turning signal into meaning
Turning a microphone that picks up birdsong, or a camera reading sky colour, into something the system can respond to is still a work in progress, built jointly by the small language models and the people feeding them. Open-source pattern recognition can flag a mackerel sky as a sign of an approaching storm; alongside it sits a growing library of local weather lore, gathered from old texts and from people who still hold it.
The river as teacher
None of this stops at collection. Once the sensors and the library have turned a signal into something readable, that reading gets handed back to whoever is standing nearby, which is the reason for building any of this in the first place: a single reversal. Instead of a human prompting the AI, the river’s data starting to prompt the human, offering a poetic reminder rather than an answer. “The river in that sense then becomes our teacher,” Anab said, “and really, as you listen to the river, you start to remember the oldest truth we have: that the river that runs through you runs through me, and you cannot poison the river without poisoning your own blood.” Once that continuity is genuinely felt, the reminders stop being necessary.
🌱 Seed:
Reversing who prompts whom is a small technical choice with a large implication. It treats the river as the one with something to say, and the human as the one who needs reminding how to listen.
5. Reception, stakes, and policy implications
What the Changing Course forum showed
Over eighty policymakers, academics, and NGOs engaged with the Ecological Intelligence Agency at the Changing Course forum. Whether that has shown up in UK policy since is hard to trace from outside of government.
What a consultation actually does
What is clear is that the conversation with the Policy Lab is ongoing, with a real appetite to deepen the thinking. Anab’s broader read on how consultations actually work makes that claim credible rather than hopeful: “every time they do a consultation, what they take back doesn’t change the policy. But what the consultation does is shape public understanding of the thing, and that in turn, eventually, informs policy.”
🌱 Seed:
The gap between a provocation and something a regulator takes seriously isn’t mainly technical. It’s the guardianship and ownership questions, worked through for real.
6. Open questions
Whether this fails
Asked directly whether a version of this work, of “ecological AI”, could be co-opted or diluted into something closer to greenwashing, Anab noted she has seen it before. Co-optation happens and has happened, echoing Mark Fisher’s observation that it’s easier to imagine the end of the world than the end of capitalism.
At the same time, something else is also true: the level of systemic uncertainty we are facing, the level of instability we are likely to be faced with, combined with enough plural stories and the right platforms, still leaves room to shift toward a different way of living. What that shift asks for is a change in lifestyle toward co-ownership and shared assets, and what stands in the way of it, plainly, is loss aversion.
What she’d build first, given the money
Given an unlimited budget, Anab pointed to a combination of the two projects: the hard-evidence side, because she cares deeply about PFAS levels in UK rivers, paired with the re-attunement work of the sensors, built smaller and more locally powered, rooted in local guardianship, human and more-than-human. Her larger, unfunded dream is an atlas of the sensory languages of the world.
What she doesn’t have an answer to yet
The question Anab is still sitting with wasn’t so much strategic, but personal: “at a very personal level, I would like to swim in a river, that’s partly because I’m a very recent swimmer, so I have a fear of it, but also just to really feel them.” To deepen her own understanding of the ecological wisdom of the Thames specifically. The studio once sketched out building a boat and making a pilgrimage along its full length, from source to mouth, “because the Thames used to be a sacred river, like a lot of rivers where I grew up.”
🌱 Seed:
The most honest answer to “what’s unresolved” wasn’t so much a design gap, but rather a wish to spend more unmediated time with the river the work has spent so much time representing.
Cross-cutting insight threads
Feeling before knowing
Across both projects, the design choice is consistent: make a future or a relationship felt before it’s explained. A graph, a policy brief, or a dataset can inform; a lived object or encounter is what tends to move someone to act.
Relationship, not data volume, is the redefinition
The recurring answer to what makes intelligence “ecological” wasn’t more environmental data layered onto an existing model. It’s the insight that intelligence itself is relational, built through sustained contact with a place and its species, which is what most current AI systems aren’t designed to hold.
The flow reversed, twice over
The Ecological Intelligence Agency reversed who technology serves, putting AI to work for the river rather than extracting from it. Nobody Told Me Rivers Dream reversed who prompts whom, having the river’s data initiate a poetic prompt to the human instead of waiting for a human query. Both are the same underlying move, applied at a different scale.
The same three questions, at every scale
Wherever the conversation moved — the river Roding or Thames, farmland in India, or the Colorado River, drained before it reaches the sea, the test didn’t change: who benefits, who has given consent, and what form does reciprocity take. Offered almost as an aside the first time, it ended up doing more work across the session perhaps than any of the AI-specific design choices, holding the Agency, the sensors, and the wider question of what an AI advocate should be answerable to under one standard.
What stays unbuilt, at every scale
Four separate parts of this work turn out to share the same shape. A system able to say ‘it doesn’t know’ is not yet built. Who holds the sensor data, and on what terms, is not yet settled. Whether the Agency has actually shifted UK policy can’t yet be confirmed. And what ecological wisdom along the Thames specifically requires is still, by design, unfinished. None of that reads as a gap the studio failed to close. It reads as the provocation and a consistent shape: at every scale of this work, from a single sensor to national policy, the honest position is still open rather than settled.
Anab Jain is co-founder and director of Superflux, a design, research, and futures practice based in London, and Professor of Design Investigations at the University of Applied Arts, Vienna. More on the two pieces of work discussed in this session can be found here: The Ecological Intelligence Agency (with UK Policy Lab and Defra Futures) and Nobody Told Me Rivers Dream (Design Museum, 2025).
Closing
These sessions are not designed to arrive at fixed answers, but to surface how different approaches work in practice. Each conversation reveals not only how and where they hold, but also the tensions and questions that remain unresolved, and that continue to invite further inquiry.
Over time, these syntheses form a growing body of insights into what co-creation with the more-than-human world might require, across contexts and practices.
The Co-creation with the More-than-Human sandbox is a collaboration between The Repatterning Collective and Kincentric Leadership.
Our popular monthly online learning sessions use live dialogue & discussion to explore different approaches and praxes for meaningful co-creation with the more-than-human (natural) world.
We also regularly publish Field Dialogues (audio track & synthesis), providing snapshots from a living field — surfacing stories, practices, and insights in the evolving work of engaging the More-than-Human
To receive syntheses of Learning Sessions and Field Dialogues directly in your inbox, find us on Substack, LinkedIn Newsletter, and of course here.
We donate our time and energy to this sandbox. If you value this space, consider co-funding this work by leaving us a tip.
