The River Wants a Vote

On digital democracy, marginalised voices, and widening the conversation

The spark of this film was a chatbot.

I've been interested in digital democracy and civic technology for a long time, and especially in the work of Audrey Tang, Taiwan's former digital minister. For years I've been trying to develop ideas for how to make films about this space, and it's not easy, because all the things I find most exciting about it are happening on computers — people writing code, coming up with mechanisms, building interfaces for council meetings. It's not visually exciting. It's not obvious where the story is.

But then, reading Tang and Glen Weyl's book Plurality, I came across an example called the river chatbot. "Say Hi to the River" — a chatbot developed in Taiwan around 2023 by Shu-Yang Lin and the team at PDIS, Taiwan's Public Digital Innovation Space, from a prompt by Audrey Tang to think about the most marginalised voices and how technology could bring them into democratic conversation. The idea was simple: if you're going to make a decision that affects a river, why not bring the river into that conversation? Why not include it in the democratic discussion about that decision?

I thought the idea was such a beautiful and poetic gesture in the direction of a richer and deeper democracy. The actual implementation was super limited — it was essentially a ChatGPT wrapper, really more of an art project or a concept than anything usable in the real world. But I got caught on this idea of giving a voice to a river through technology, and that's how the whole project started.

From there, as we began to develop the film, we started to realise there was a really interesting paradox at its centre. On one side, there are these fascinating pilot projects and experiments in giving voice to rivers through technology — the chatbot, yes, but also scientists using sensor arrays and machine learning to build real-time pictures of a river's health, tracking where pollutants are coming from; artists experimenting with hydrophones and generative sound and data visualisation. In Sheffield, we found the River Storyteller platform, built by Johnny Douglas and his team at Dark Matter Labs, which combines live environmental sensor data with poems, memories, and stories contributed by local residents. It doesn't claim to speak for the River Don — it holds scientific data and human testimony and cultural memory in tension, layered together, none fully authoritative on its own. It's a genuinely civic tool for relating to the river, and a quieter, more honest model than the chatbot's confident synthesis.

But at the same time, the data centres and semiconductor factories upon which this AI technology relies are draining huge amounts of water from rivers, causing enormous damage to ecosystems and to the communities that live alongside them. In Taiwan, TSMC consumes over 150,000 tonnes of water per day. In Spain, activists under the banner Tu Nube Seca Mi Río — "Your Cloud Dries My River" — are fighting Amazon's proposed data centres in drought-stricken regions. In Memphis, Tennessee, xAI installed dozens of unpermitted gas turbines to power its supercomputer in a neighbourhood already suffering from decades of industrial pollution. The technology that promises to give rivers a voice depends on infrastructure that is silencing them. That tension became central to our film.

There is also a legal dimension to this that predates the technology. The global Rights of Nature movement has been building the case for non-human participation in governance for decades. New Zealand granted the Whanganui River legal personhood in 2017, with representatives appointed jointly by the Māori iwi and the Crown. Ecuador, Colombia, Quebec, and most recently the Colorado River Indian Tribes have all found ways to grant rivers legal standing. In the UK, Fang-Jui Chang — a designer at Dark Matter Labs and one of the key figures in our film — has been appointed as one of the first official "River Representatives" for the River Roding, carrying the river's interests into local council meetings. All of these are steps toward widening who gets a voice in democratic decisions. The chatbot is another step in the same direction, taken from the technological side rather than the legal one. But legal personhood is always a translation — it takes a relationship that existed in one register and renders it in another, and something is inevitably lost. The Māori, for whom the Whanganui is a tupuna, an ancestor, had always understood their obligation to the river; the Te Awa Tupua Act expressed that understanding in terms the Western legal system could process. The danger, with any of these frameworks, is that the symbol of inclusion becomes a substitute for the real thing — that having appointed a representative, a council feels it has "included" the river, and stops there.

When we think about AI and democracy, one of the most interesting advances is the ability of these systems to ingest not just hard data — not just scientific measurements, the kind of data that previous systems could compute — but also, to some limited extent, soft data. Poems. Music. Cultural history. Memory. Sound. What we are beginning to have is a system that can start to understand something of the web of interconnected experiences that make up an actual complex ecosystem. That capacity is still limited, of course, and it could be criticised as naive — as a human projection of voice onto an entirely non-human thing. But democracy has always been limited too. One vote every four years is about one bit of information every four years. That's a simplification, and we do bring more than that into our democratic process already, but the fact remains that we rarely think about how much technology could enable richer, faster, deeper engagement between a population and the systems that govern it.

In our film, we try to imagine a river AI that is the best possible version of this. A system that is learning not only the hard scientific data being fed through cutting-edge sensor networks, but also the cultural memory, the history, the poems, stories, and music. We try to imagine a system that is holistically integrating all of the ways of listening, all of the ways of understanding the river, in order to most fully represent that river in human society. A translator, of sorts. And yes, this comes with very obvious limitations. It could be criticised as totally missing the point of what real embodied experience is, of what it means to actually stand in the water and sing to the river the way the Paiwan and Amis communities in Taiwan do. But I think what we want to show in the film is that yes, this is limited — yes, it is a human projection of voice onto something non-human — but nevertheless it is a step in the right direction. It is better than what we had before, which was little to no attempt at listening to the river in all its complexity.

The film is still in development. We haven't shot it yet. But the direction is clear to me: a film about widening the circle of who gets heard, using every tool available — legal, technological, cultural, scientific — and being honest about the limits of each one while still insisting that the attempt matters. (We're also experimenting with making the film itself more democratically — using an open-source collaborative model inspired by the Gov4Git protocol that was used to write Plurality. The early framework is on GitHub.)

Democracy has always been a technology for aggregating voices. The question has always been whose voices count. The river chatbot is a small, imperfect, poetic gesture toward an answer: everyone's. Including the river's.

Back to 'Beneath A Hundred Streams'

Back to Writing Index