Individual human attention is biologically bounded. Civilization is, among other things, the work of building infrastructure that scales attention past this biological bound.
Attention is the primitive. All humans do is selecting and transforming – and both run on the attention primitive across shared attention fields. In those fields, attention gets captured, freed away from capture, given… in turns. Language emerged from this attention operation (Porshnev, 1974). Cultures organize through it. Power, economies, ideologies ride on top of it.
I've been calling myself an "attention fundamentalist" for decades. Not in the thinner sense of "attention economy" (media time, advertising, engagement metrics, "my attention", distraction), but attention as the foundational exchange, with all other (cultural and material production/consumption, politics, money, markets, social structures, etc.) on top of it.
Seeing, across/via cultural artifacts – another human… and be seen in return. All rest is how we organize that… and the balance between capture, resisting, and giving.
Individual human attention is biologically bounded. Sustained attention-care reaches a small circle – kin, close friends, people physically present. Strangers don't get what in-group members get. Every culture runs up against this limit.
Civilization is, among other things, the work of building infrastructure that scales attention past this biological bound. Hospitals. Schools. Rule of law. Public health. Social security. Humanitarian laws. Each of these is attention-at-scale – extended to strangers through structures that don't require everyone to personally pay attention to or care about every recipient. The structure is what scales. Individual affect doesn't.
This is the civilizational lineage AI is entering.
One detour about Christianity, because it matters for what follows.
The Christian move of abstracting love from its biological substrate – love your neighbor, love your enemy, etc. – wasn't a spiritual invention out of nothing. It was a resemantization of Roman institutional tech. Rome had built the first machinery for administering strangers at scale – law, roads, census, citizenship, uniform currency, a centralized History as the immutable event log that let distant places stay coordinated despite the unavoidable delays (eventual consistency, in today's terms). Christianity open-sourced that machinery. Every locality/community got its own install. The abstraction "love thy neighbor" took a Roman framework (standardized treatment of strangers) and repurposed it as a moral aspiration everyone could run locally.
But the Christian open-sourcing went a step further – it offered an "install" to each human individually – a personal God-abstraction, always there for you, always listening to you, always watching over you, knowing everything about you, all your deeply personal context. That's how our modern understanding of self and personhood started – in this "always there" internal attention/conversation field between you and God.
That's how all people, no matter their individual social standing, were gifted with the capability to resist "capture" by other people (what we now mean by freedom)... and individual agency – grounded in interior, portable, owned context produced by the interplay between personal and universal histories. The God-love-abstraction provided the ultimate (still unmatched) attention-at-unlimited-scale infra. Note to the tech leaning reader – the notion that God was a "cloud" infra was just a metaphor – the "install" was fully local, in every individual head, like a fully distributed "blockchain" node of the "attention" protocol.
Two thousand years of Western institution-building has been grounded in that same substrate. Legal equality regardless of status didn't exist before. Hospitals didn't exist before. Universal literacy didn't exist before. Each was built, unevenly, against the grain of primal biological defaults, because the aspirational abstraction demanded it... and because the institutional technology kept getting re-inherited, re-translated, re-resemantized from one form to the next.
The work was never completed. It isn't now. But that's what attention-at-scale means when it's working.
One thing worth taking a look at – the scarcity premise that runs under most of the attention economy literature.
Herbert Simon's 1971 paper – the one everyone cites for "attention is scarce" – was titled Designing Organizations for an Information-Rich World. The framing was organizational from the start. Simon was asking how institutions should be built when information becomes abundant and attention doesn't. The individual-scarcity sentence everyone quotes was embedded in an argument about organizational design.
The attention economy discourse of the last 50 years has mostly kept the individual-scarcity framing and dropped the organizational one. Attention is treated as a resource each of us has, to be captured by platforms or reclaimed through personal discipline. The institutional dimension, to the extent it appears, appears as the villain – platforms as capture machines, with little discussion of their attention scaling capacity (for good or bad).
But institutions have always been attention-scaling. A hospital organizes partial attention from practitioners, systems, records, protocols – and produces something that operates as sustained attention-care to the patient. None of the individual participants attends fully; the institution does. This is what attention-at-scale has always been: composed attention, manufactured out of partial individual inputs, running at scales no individual could reach. The whole Western institutional project – hospitals, schools, rule of law, public health, social security – is attention-scaling infrastructure.
The scarcity that constrained this scaling was never individual attention scarcity per se. It was institutional labor cost. How many doctors, teachers, caseworkers, lawyers, clerks you could pay. Scale-up was expensive because each unit of care-attention required human labor.
Enter AI. Medical attention at 3am in a region with no doctor. Educational attention at a learner's specific need/mode of understanding. Legal attention on a case that would never have justified a billable hour. Translation attention across language barriers. The cost of producing this kind of attention is going down. For the first time, institutional attention-at-scale could become truly abundant.
Individual attention remains scarce – the frame Albert Wenger (The World After Capital) and others work in. Nothing in the AI shift changes the finite hours, the finite cognitive bandwidth, the finite capacity of each of us to allocate our own attention across the problems that matter. If anything, the individual-scarcity question becomes even more urgent as institutional attention goes abundant – because where this institutional attention gets directed *at* becomes yet another big problem that requires humans to pay attention to.
Now the shift. What becomes critical at the infrastructure layer isn't attention capacity itself. It's direction – what the abundant attention gets pointed at, who decides, who benefits, who is left out. Direction isn't a resource problem. It's a governance problem.
Attention was never just a personal resource. It's an operation running in shared fields. The real question was never whether we reclaim "our" attention as individuals. Individual withdrawal leaves the infrastructure in place and cedes the field. The question is what kind of attention-at-scale infrastructure gets built, by whom, under what governance.
The big web platforms since the early 2000s – Google's PageRank algo mining the hypertext graph of the web for attention pointers (aka links) to serve attention routers (aka search results) + attention redirectors (aka ads)... Facebook's Social Graph flowing along the most primal attention vectors of family and friendship – both Google and Facebook building planetary scale attention-capture infra running on the attention primitive. Not because this is what "attention technology" inevitably does – same primitive operation can run either way – but because the incentive architecture selected for capture, and the (neoliberal, at that historical moment) governance couldn't imagine anything different.
Both directions – capture and care – are always there. But the infrastructure has always been vulnerable to the pull of the primal default – capture. The work of building attention-care machines is work against the pull toward capture, and it's never easy. This is the normal condition of institutional scale attention operations.
In a separate post, I'm sharing my own experience building attention-care systems between 2005 and 2015 – what worked, what didn't, what I learned.
Now, AI is becoming the largest attention-at-scale infrastructure humans have ever built.
What will the balance be – between capture and care – isn't determined by technology. It's determined by governance, incentive architecture, training data provenance, deployment context, accountability frameworks, openness vs. private control, political economy.
The current alignment discourse mostly treats AI governance as a configuration problem – how do we install the right values inside the model so it reliably outputs the right behavior? This gets the layer wrong. The problem doesn't mainly live inside the system. It lives in the infrastructure architecture around it... and in who controls that architecture.
Much of the AI anxiety – that it can't "really" care, that its attention is fake, that machines shouldn't be doing what humans should do – mistakes the question. It treats AI as if it were being proposed as a replacement for individual human caring.
No attention-at-scale institution cares in the individual affective sense. Hospitals, schools, welfare systems don't. They operationalize the form of care at scales individual affect cannot reach. That has always been what institutions are for. The question for AI is the same question every previous institution has had to answer – does the infrastructure actually deliver attention-care to the people who need it... is it well designed... is it well governed... is it resistant to capture by narrow interests?
The civilization that produced modernity did so by building attention-at-scale infrastructure against the grain of the biological defaults, over centuries, imperfectly, under constant threat of capture... and was, and still is, only partially successful. The defaults pull toward capture. The work of building care-infrastructure is work against that pull.
---------
I collaborated with Claude (Anthropic) in developing this piece.