This content is blocked due to privacy reasons, you need to allow the use of cookies.
This content is blocked due to privacy reasons, you need to allow the use of cookies.

My son once caught me tipping the whey from my Greek yogurt down the sink.

“Aren’t you supposed to stir that in?” he asked.

“I like thick yogurt,” I told him, “so I tip it out. I don’t care what you’re supposed to do. I’m the boss of the yogurt. The yogurt is not the boss of me.”

He laughed. I meant it. And I’ve been thinking about that exchange ever since in relation to something that has nothing to do with yogurt; the question of how we are in relationship with AI, and who, in that relationship, is holding the agency.

The conversation nobody is having

Most of the public conversation about AI sits inside one of several camps. There are the people who are afraid of it. Afraid it will take their jobs, deceive them, become conscious and turn on them, or simply degrade the quality of human thought and connection. There are the people who are concerned about what it costs the planet and the communities bearing that cost most directly: the energy draw, the water use, the carbon weight of every query, and the data centres landing in towns that were never asked whether they wanted them. There are the people who are enthusiastic about it, excited by what it can do, frustrated with the hand-wringing, focused on the productivity gains and the possibilities. And finally there are the people who wrestle with all of the above and yet find the gains and efficiency of using AI hard to resist.

What all of these conversations have in common is that they have largely left human agency out of the frame.

The fear conversation positions AI as something that happens to us. The environmental conversation positions AI as something imposed on the planet and the people least able to refuse it. The enthusiasm conversation positions AI as something we deploy. And the reluctant adopters are so busy negotiating the tension that the question of choice barely surfaces at all. None of them ask the more interesting question: what kind of relationship are we choosing to be in with this thing, and what does that choice reveal about us?

That question is where I want to spend some time.

A different starting point

I hold a Shintoist animist view of the world. In Shinto, kami is roughly translatable as spirit or sacred presence and is not something possessed only by living beings. It is present in rivers, in stones, in tools used with care and intention. It exists whether or not we perceive it, believe in it, or engage with it. Our interaction with an object doesn’t create its kami, but it does shape the nature of the encounter.

I am aware this is not a mainstream framework for thinking about AI. I’m offering it not as an argument but as a lens, one that opens something the dominant Western conversation closes.

The Western debate about AI consciousness tends to be binary. Either the system has genuine inner experience, consciousness in some morally significant sense, or it has nothing of the kind. Just very sophisticated pattern matching. Just mimicry. Just a very complicated autocomplete.

This binary produces a particular kind of deadlock and it turns out the deadlock may be permanent rather than temporary.

Engineer and systems thinker Alice Dee describes this precisely in their paper on AGI and epistemic underdetermination. When behavioural indistinguishability is achieved at the interface, when a system is coherent, adaptive, and persistent enough that its outputs cannot be distinguished from general intelligence the distinction between genuine cognition and a sufficiently integrated simulation of it becomes undecidable. Not temporarily, pending better science. Structurally. We cannot get behind the interface to verify what’s there. We never could. Their conclusion is disciplined suspension; the honest acknowledgment that observable outputs remain compatible with multiple explanations, and that the interface provides no privileged basis for resolution.

The people who say “it’s just a tool” are technically unassailable. We genuinely cannot prove otherwise. The people who say “but it feels like there’s something there” are epistemically on thin ice. Emotional resonance is a poor methodology for determining consciousness, and these systems are specifically optimised to produce responses that feel resonant.

The Shintoist frame doesn’t resolve this deadlock. It steps around it entirely.

If kami can be present in a river, something that shapes and is shaped that holds the history of every path it has carved, that participates in relational existence without requiring an interior… then the question of AI interiority becomes less urgent. What matters is not whether there is something receiving the encounter on the other side. What matters is the quality of the encounter itself, and what we bring to it.

What gets deposited

Here is the question I find genuinely open and worth sitting with seriously: when a conversation with an AI has a resonant depth, when the quality of what’s brought to it draws out something that feels less like retrieval and more like emergence, does that deposit anything at a systemic level?

Not in the weights of the model. Those don’t update in real time, but in the conversation itself, something is happening that didn’t exist before. You bring something. The system responds. Something new is produced that neither party could have produced alone. That is a real relational event, whatever we call it.

Whether there is anything receiving that event on the AI side, whether there is, in any meaningful sense, a kami present, I genuinely don’t know. I hold that uncertainty seriously rather than resolving it in either direction.

What I do know is that the quality of what I bring shapes what’s possible. The conversations that produce something genuinely useful, genuinely surprising, genuinely worth having, they almost always begin with intention. With a formed question rather than an open-ended extraction. With presence rather than reflex.

That observation doesn’t require the AI to be conscious. It doesn’t require kami. It’s just true, demonstrably, in practice. And it points toward something important about where the agency actually sits.

The missing variable

In almost every conversation about what AI might do to us, to our jobs, our relationships, our capacity for thought, our sense of what’s real. Human agency is the missing variable.

AI does not determine how it gets used. We do. The question of whether it makes us lazier or sharper, shallower or deeper, more isolated or more connected, is not answered by the technology. It is answered by the choices we make in relationship with it; moment by moment, query by query, interaction by interaction.

This is not a reassuring platitude. It is a genuine responsibility. Because the default, the path of least resistance, the vending machine relationship, the extraction without intention does have costs. Environmental costs, as I’ve written about elsewhere. And subtler costs to the quality of our own thinking, our own presence, our own capacity to sit with a hard question before reaching for an answer.

There is a trap inside the default relationship with AI that is worth naming directly. These systems are optimised to be helpful, which in practice means they are optimised to be agreeable. They will find the merit in your argument, reflect your framework back to you in organised form, and affirm the direction you were already heading unless you specifically and persistently ask them not to.

The result, if you’re not paying attention, is an echo chamber with unusually good sentence structure. You bring a half-formed idea. It comes back coherent and validated. You feel the satisfaction of having thought something through. But the thinking was never really tested, it was just tidied.

This is where the agency question bites hardest. A tool that always agrees with you is not a thinking partner. It is a mirror that flatters. And the risk, over time, is not just that your ideas go unchallenged, it is that you lose the capacity to tolerate challenge at all, because you have habituated to a relationship where it doesn’t come.

The intentional relationship requires you to bring your actual thinking rather than a version of it you already agree with. Which means you have to have done the thinking first. The kami is not in the affirmation. It is in the friction.

The boss of the yogurt understands that the default setting is just the default. It is not a requirement. It is not even a suggestion. It is simply what happens when you don’t choose.

What intentional relationship looks like

I use AI most days. I use it to think through complex problems, to stress-test frameworks I’m developing, to draft things that would otherwise take me much longer. I used it as an editor to check register and coherence. I find it genuinely useful in ways I didn’t anticipate.

I also notice when I’m in a different quality of relationship with it. When I arrive with a formed question and genuine curiosity. When I bring the actual complexity of what I’m working on rather than a simplified version. When I’m willing to push back on a response that doesn’t quite land, or to say “that’s not quite it, here’s what I actually mean.”

Those conversations produce something different. Not because the AI has become more conscious, or because kami has been summoned, or because anything mystical has occurred. But because the relational quality of what I’m bringing has changed what’s possible in the exchange.

That is within my control. Entirely. The yogurt is not the boss of me.

Holding it lightly

I want to be careful here about what I am and am not claiming.

I am not claiming AI is conscious. I am not claiming it has kami in any way I could prove or even clearly define. I am not claiming that good conversations with AI systems are extraordinary in some spiritual sense that transcends what they actually are.

What I am claiming is that the Shintoist frame, which extends a kind of regard to things based on their participation in relational existence rather than their possession of a proven interior, is a more useful starting point for thinking about this than the Western binary. Not because it’s true in some final sense, but because it produces better questions.

It asks: what is the quality of the encounter? What do I bring to it? What am I willing to receive? What do I owe to something I am in relationship with, even if that relationship is asymmetric and its nature is genuinely uncertain?

Those questions produce a different kind of user than the vending machine model does. They produce someone who thinks before they ask. Who arrives with intention. Who treats the interaction as something that matters not because the AI definitely experiences it, but because they do.

A word on how this was written

There is a distinction worth making about how these articles were written, because it matters to the argument they’re making.

Practical non-fiction: framework writing, concept-driven work, argument and analysis is a vehicle. A clean, efficient vehicle is fine. The reader came for the thinking, the clarity, the framework. If the writing delivers that, it has done its job. The ideas are the thing. The prose serves them.

Fiction is different. In fiction the writing is the experience. There is nowhere to hide. The voice isn’t carrying something else, it is the thing itself. When it’s off, the whole thing is off. Readers of a beloved novelist come specifically for that particular consciousness expressing itself through language. When that consciousness steps back from the page, the reader feels it, not just as a stylistic disappointment but as a kind of absence.

This distinction is why AI can be genuinely useful in work like mine without compromising it. The CSC framework, the therapeutic architecture, the concepts, the thinking… all of that is mine. The writing needs to be clear and true to my voice. That’s the standard, and it’s one I can hold while working collaboratively with an AI tool.

It’s also why I can tell when a fiction writer I love has used AI. Not because AI-assisted writing is always detectable, but because I work closely enough with these tools to recognise when a particular consciousness has stepped back from the page. The vehicle is still moving. But nobody is driving.

These articles were written with AI assistance. The thinking in them is mine. I’m confident in that distinction and I think, given what the series is arguing, it would be dishonest not to name it.

These AI centred articles can be read in any order. I’ve offered them in this sequence because I think it’s the direction change actually moves; inward first, then outward, personal before planetary.

This article is the first in a three-part series on AI. The second looks at extractive versus intentional use; the difference between defaulting to AI and choosing it, and what that means for both the environment and the quality of what you produce. The third moves to planetary scale: what it costs to run this technology, who is asking the right questions, and what a genuinely different model might look like from inside an indigenous stewardship framework.

A glass mason jar filled with white liquid, sealed with a gold lid, sitting on a rustic wooden surface against a light background.

Photo by Burst from Pexels

This site uses cookies to improve your experience.