Farther Than One Can See

A group can see farther than any one of its members — not by stacking up their eyes, but by grinding a lens none of them could grind alone. This essay is about that lens: how a group comes to know its world and act in it, and why the same instrument that widens its sight can be ground down to tunnel vision.

The question, and why it isn’t a library

Ask what a community knows and the mind reaches, almost at once, for a store of some kind. A library, an archive, a database — the sum of the facts its members carry, gathered in one place and looked up when needed. It’s a natural picture, and it’s the wrong one. A store of facts just sits there: it knows nothing, sees nothing, does nothing until someone reads it. What a we knows isn’t a stockpile but an activity — a lens the group grinds and looks through, an instrument for seeing and reaching that lets it notice what no member could notice, and act where no member could reach.

So this essay makes one claim about that knowing, before we get to how a we should choose with it: a group’s knowledge and capability are a real, emerging thing — the methods-model of a self made of selves — and they are not the sum of what its members know. Get that wrong and the questions downstream get answered wrong. You’ll go looking for a we’s understanding in the heads of its cleverest members and miss it; you’ll try to make it wiser by piling on facts, and only harden it instead. What a we values was the first essay’s subject; how a we decides is the last essay’s. This one sits between them, on the organ both of those depend on — how a we knows its world, and what it can do there.

Farther than one can see

Start with the capability itself, because it’s the plainest place to see that the collective isn’t just the sum.

A warship can’t be navigated by any sailor aboard it. Edwin Hutchins spent a deployment watching a crew bring one into port and found that no single person on the bridge held the knowledge or ran the calculation; the navigating was done by the whole — the team, the charts, the gyrocompass, the bearing log, the practiced handoffs between them — a thinking system spread across people and instruments, no part of which could think the thought alone. The knowing was real, and it sat nowhere in particular. It was the arrangement.

Widen the frame and the same shape turns up at every scale. A craft tradition holds skill that outlives every craftsman who ever practiced it, handed down and built on without slipping back — the ratchet that lets each human generation start where the last one left off instead of from scratch. Science is the grand case: a lens ground over centuries that lets us see the cell, the galaxy, the pathogen — things no bare eye will ever meet, known now by a collective that spreads its labor across rivals and specialists, none of whom holds the whole. This is emergence in the register of capability, and it’s worth being careful with the word, as the first essay was: the group’s power isn’t our failure to add up what the members separately know. It’s a real capacity of members-arranged-just-so — the whole has powers the scattered parts wouldn’t; a failure of aggregativity, in the philosophers’ phrase, which is a careful way of saying the group is more than the pile.

A lens is the right picture for it, and not only because it magnifies. A lens shows what no eye owns on its own; it’s ground by many hands over long stretches of time; and — this is the part the next sections turn on — once you’ve looked through it, you no longer see the world you saw before.

The lens grinds the grinder

A we doesn’t only build its instrument. The instrument reaches back and reshapes what its members can see and think — the same downward pressure the first essay found in the world of value, now in the world of knowing.

Watch it in any field. A radiologist sees a tumor in a smear of gray where the rest of us see weather; a trained ear hears the flattened third the rest of us hear as merely sad. This isn’t faster reasoning about the same perception — it’s a different perception, and the group installed it. That’s the force of Thomas Kuhn’s unsettling claim that a shared paradigm shapes perception itself, not just the words laid over it afterward: point the trained eye and the untrained eye at the same slide, and they don’t take in the same thing. And most of what the group installs runs below anything a member could put into words — Michael Polanyi’s point that we know more than we can tell, the expert’s judgment arriving a step ahead of any rule she could state. A shared method does the same thing beneath any single skill: double-entry bookkeeping, the controlled trial, the habit of asking for a base rate — each one a technique the group worked out and then pressed into its members, until a trained mind reaches for it without deciding to. What a member can notice at all is bounded by the lens the group ground for her. By the time a practitioner “just sees” what’s there, the freedom in the seeing is real — but its shape was set long before the moment of looking. A we’s knowing lives in its members, as trained perception the we pressed into them — which is exactly why it’s causally real, and exactly why, turned the wrong way, it can blind them.

An engine, not a camera

That reshaping runs across generations, and it’s where this essay carries the most weight. We grind the lens, and then the lens grinds us. Writing didn’t merely record what minds already did; it remade what minds could do — and print, and the network, each remade it again: new collective instruments that reorganized the very cognition that built them.

The sharpest form of the loop isn’t a friendly one, and it’s the one a framework built on perspectival realism — the view that we always see from somewhere, never from nowhere — has to face head-on. A collective model is, in Donald MacKenzie’s phrase, an engine, not a camera: it doesn’t photograph a world sitting there independently so much as drive one into being. His case is the option-pricing formula that, when it was published, didn’t fit the market it described — its assumptions were false to the trading of the day. Yet traders adopted it, regulators bent the rules to fit it, and within a few years the market had rearranged itself until the formula held. The model didn’t learn the world; it taught the world to match the model.

This is the reflexive power at full strength, and it names the peculiar danger of a we that can act at scale. Perspectival realism says our models are always somebody’s models, answerable to a reality that pushes back and grades them by whether they work. A we strong enough can put that pushback off — can remake the world into the shape of its picture and mistake the echo for confirmation. Reality isn’t repealed; the formula’s world still cracked, in its crashes, once the assumptions finally mattered. But a collective can wall the pushback off long enough to fool itself, and that delay is the seam where widened sight seals shut into a closed loop. The newest lens we’re grinding runs the same risk, and faster — but that’s a later section’s worry, not this essay’s center.

How the many become a model

For a we to act, and not just pile things up, the knowledge scattered across its members has to be gathered into one model it can act on. This is the collective’s representation problem, and it’s where a we’s knowing is either made real or quietly falsified.

The classic mechanism is Friedrich Hayek’s. The knowledge a society runs on is scattered, tacit, and perishable — held in ten million heads as a feel for a local circumstance, and never available to any planner in one place. The price system, Hayek argued, is a way of squeezing all of that into a single traveling signal: a price gathers up what no mind could gather, and lets strangers coordinate without any of them grasping the whole. It’s the channel through which an economy represents what it knows — scattered knowing made actable.

And it shows, in miniature, why that channel is fragile in a way worth naming. Aggregation can eat itself: if a market’s prices already reflected everything known, nobody would be paid to find anything out — so a perfect aggregator destroys the very effort that feeds it, which means a healthy collective model needs friction, disagreement, and duplicated search to keep working: exactly the passing diversity a monoculture throws away. Aggregation can also lie. When members choose in sequence, each watching the last, they’ll sensibly stop consulting their own read and copy the crowd — and the group locks onto an answer while still holding, scattered and unheard, the very knowledge that would have corrected it. And the summaries a we lives by can rot: a measure that becomes a target stops measuring — publication counts get gamed, the number optimized in place of the thing it once stood for. This is the same hazard the essay on holding a we together met at the seams, where a lossy summary passed upward can quietly stop tracking what it summarizes. The channel that makes a we’s knowing collective — price, publication, measurement, the trusted report — is also the place it most easily fails.

The lens turned forward

A we’s model isn’t only a map of what is; its most consequential use is to run it forward — to see what hasn’t happened yet, and to act before it does. This is foresight, and it’s the same widening turned along a different axis: not the lens reaching farther across the world, but farther ahead of it. It’s also the clearest case of all where the group reaches past any member, because no single life is long enough to see a century, and only a we — with records, models, and a memory that outlasts any of its people — can hold a distant horizon in view at all.

Foresight works at two ranges, and a healthy we needs both. Close in it’s tactical: the storm tracked before it makes landfall, the outbreak caught before it spreads, the shortage spotted a week out and covered — seeing far enough ahead to change the next move. Far out it’s strategic: the slow arc of a technology, a climate, a population, read decades ahead, where the consequences land on people not yet born and the only question worth asking is which way the whole thing is tending.

The techniques for it are real and teachable — not crystal balls — and each is a way of grinding the forward lens sharper. A we learns to see ahead by rehearsing several futures instead of betting on one — the scenario method with which a single firm rehearsed the 1970s oil shocks before they hit, while its rivals were caught flat-footed; by disciplining a scatter of experts into a converging estimate, the way the Delphi method does; by running forecasting tournaments that, as Philip Tetlock’s work has shown, expose how poor most confident prediction is and yet prove that careful, humble, often-updated forecasting is a skill you can actually train; by scanning the horizon for the first faint sign of a coming thing, the early-warning systems that turn a spotted famine or contagion into a response in time; and by building models a we can run the world forward inside — the climate projection, the epidemic curve, the systems study that first warned a finite planet of its limits.

Two honesties keep the claim sober. Foresight isn’t prophecy: the future genuinely resists it, and the same work that shows forecasting can be trained also shows how badly most of it is done. And foresight loops — a warning that’s heeded is a warning that looks like it failed, the disaster you headed off indistinguishable from the one that was never coming, so the forecaster who gets it right is often thanked least. Its true opposite is the counter-dynamic stretched along the years: the we that won’t look ahead, that discounts tomorrow down to nothing and bills its costs to people who can’t yet object — coherence bought by narrowing the circle not across space but across time. To turn the lens forward is to widen the reach of concern into the future — toward the generations not yet born, and the world they’ll inherit. How a we comes to actually care about that farther horizon, and not just see it, is Caring Across Time’s job to take up. Here it’s enough that the lens can be turned forward at all.

The lens that narrows

All of which sets up the fork this essay exists to draw. A we can grind its lens to widen its field, taking in more of the reality it has to act within — or it can grind it down to tunnel vision, fixing on one thing and letting the rest fall dark, buying a clean and tidy picture by refusing to see. From the inside, the two can feel alike; the difference between them is the whole of the matter.

The biggest form of the narrowing is the state’s, in James Scott’s account: to govern, a center has to make its subjects legible, and legibility means flattening the dense, local, practical know-how — the metis — that it can’t measure. The flattening kills the thing. Rationalized forestry cleared the messy undergrowth for neat rows of a single species and produced, a generation later, forest death; the perfectly planned city turned out sterile and hostile to the life it was meant to house. It’s tempting to read this as proof that scale simply requires narrowing — that a we can’t grow capable without mangling what it’s made of. That reading is the counter-dynamic mistaking itself for a law of nature. Scott’s own remedy is to keep knowledge local and plural, answerable and various — which is exactly the nested, many-centered shape the essay on scale defended as the way a we grows without flattening. Legibility-by-amputation is a choice, not a law of scale.

The same move runs through trust itself. An echo chamber isn’t just a member who lacks information — that’s a bubble, and fresh facts can pop it. The echo chamber, by contrast, is a structure that has taught its members to distrust the outside, so that contrary evidence arrives already discredited and the walls only thicken. It’s beautifully coherent, and it’s coherent by not-seeing — the counter-dynamic in the world of knowing, in its purest form: a we made consistent by narrowing what it will let in. Propaganda and manufactured ignorance are the deliberate version: a group grinding down its own sight on purpose. And the newest version is machine-made. A model trained on the world’s variety can be turned into one trained mostly on its own output, and it decays — losing the rare cases first, then settling toward a bland, confident average that has forgotten the world it came from; and at population scale, tools that nudge every writer toward the same dominant phrasing thin out the very variety of mind a collective intelligence runs on.

Set against all of this is the widening lens, and it has a clean signature: a we that can see its own condition responds to it. Famine, Amartya Sen observed, doesn’t happen in a working democracy with a free press — not because such societies are richer, but because a group that can perceive its own hunger is a group that acts on it; the seeing and the response are one motion. Science widens by keeping its model revisable instead of frozen; open knowledge widens by staying answerable; and aggregation can be built to surface disagreement instead of burying it — to bring a genuinely divided public to what it actually shares, rather than to a manufactured average. The test is never how coherent the picture looks. It’s whether the lens is opening or closing.

The honest edges

Three cautions, because the thing is real but it isn’t tidy.

First, where does a we end? A group wired up to its instruments and archives isn’t thereby a single knowing self; coupling isn’t constitution, and you can’t stretch the boundary of a mind to swallow every tool it touches without the word “knowing” going slack. The mark of a genuine collective knower, as opposed to a heap of coupled parts, is the one the book uses everywhere: integration under feedback — the two-way pressure that makes a whole answerable to its parts and its parts answerable to the whole. Short of that, there’s coordination, but no we.

Second, the reflexive power to remake the world is the power to hide from it. A we strong enough to make reality conform can seal itself inside a model that feels like widened sight and is really a closed loop — performativity gone all the way to talking only to itself. Whether-it-works is the only court that can overturn the verdict, and its ruling can be put off but not repealed. The longer the delay, the harder the correction when the world finally refuses to keep pretending.

Third — planted here, and left for the essay that can do it justice — knowing isn’t yet deciding. A we can see clearly and still choose badly; and gathering what its members know into a working model is a different problem from gathering what its members judge into a single verdict. That second problem runs straight into a difficulty the first can sidestep — a perfectly rational membership can produce a group choice no member would endorse — and that difficulty is the doorway to how a we ought to decide. That door is The Course We Carve to open.

What a we knows

Gather it up. A we’s knowing is real, not a stored stock; emerging, not added up; it reaches back and reshapes the members it arises from; it lives not in any one head but in a lens the group grinds and looks through; and, being a methods-model like any other, it’s never finished — a picture still being corrected by a world that pushes back, or else a we that has stopped looking and begun to narrow. That’s how a we comes to know its world and act within it: farther than one can see, exactly to the degree that the seeing stays open.

Which is why the last of these essays is the hardest. A we’s values are real (the first essay’s work), and its knowledge is real (this one’s) — and still neither becomes a choice the we can own until it passes through the one thing left to look at: the craft, and the trap, of deciding together. That’s where the arrow of this small trilogy finally points.

Sources & further reading

This essay leans on its sources directly, rather than through the book’s per-chapter endnotes. A citation-level pass (Zotero keys) is still owed; the works below were web-verified on 2026-07-07.

Distributed cognition and emergent capability. Edwin Hutchins, Cognition in the Wild (1995), on cognition spread across people and instruments; Andy Clark and David Chalmers, “The Extended Mind” (1998), with Fred Adams and Ken Aizawa’s coupling-constitution caution as the counterweight; William Wimsatt on the failure of aggregativity as the mark of genuine emergence; Michael Tomasello, and Claudio Tennie, Josep Call and Tomasello, on cumulative culture and the ratchet effect; Philip Kitcher, and Michael Weisberg and Ryan Muldoon (2009), on the division of cognitive labor, with the robustness critiques of it (Johanna Thoma, “The Epistemic Division of Labor Revisited,” 2015; Alexander, Himmelreich & Thompson, 2015). (List and Pettit’s group agency is held for the decision essay.)

The method reshaping the member (downward causation). Thomas Kuhn on the theory-ladenness of perception (The Structure of Scientific Revolutions); Michael Polanyi on tacit knowing — we know more than we can tell (The Tacit Dimension, 1966); and Ian Hacking, “The Looping Effects of Human Kinds” (1995, in Sperber, Premack & Premack, eds., Causal Cognition).

Reflexivity — an engine, not a camera. Donald MacKenzie, An Engine, Not a Camera: How Financial Models Shape Markets (2006), and the Black–Scholes case of a model that remade the market it described; on tools remaking the minds that make them, the formulation “we shape our tools and thereafter our tools shape us” (John Culkin, “A Schoolman’s Guide to Marshall McLuhan,” Saturday Review, 1967 — the phrasing is Culkin’s own, glossing McLuhan).

Aggregating distributed knowledge into an actable model. Friedrich Hayek, “The Use of Knowledge in Society” (1945), on the price system as compression of dispersed knowledge; Sanford Grossman and Joseph Stiglitz, “On the Impossibility of Informationally Efficient Markets” (1980), on why perfect aggregation eats its own incentive; Sushil Bikhchandani, David Hirshleifer and Ivo Welch on informational cascades (1992); Goodhart’s law, with Marilyn Strathern’s “‘Improving ratings’: audit in the British University system” (1997) on the measure that stops measuring. Read alongside Coherence at Scale on the lossy summary at the seams.

Foresight — seeing ahead in time. On scenario planning, Pierre Wack on Royal Dutch Shell’s scenarios, which rehearsed the 1970s oil shocks before they arrived (Harvard Business Review, 1985), building on Herman Kahn’s futurism; the Delphi method of structured, iterated expert forecasting (Olaf Helmer and Norman Dalkey, RAND, 1950s–60s); Philip Tetlock on forecasting tournaments — that most confident expert prediction is poor, yet calibrated forecasting is a trainable skill (Expert Political Judgment, 2005; Superforecasting, with Dan Gardner, 2015; the Good Judgment Project); and long-horizon systems modeling — Donella Meadows and colleagues, The Limits to Growth (1972), also cited in the endnotes. Early warning and the seeing-that-prompts-response connect to Sen, below.

The lens that narrows — the counter-dynamic in knowing. James C. Scott, Seeing Like a State (1998), on legibility, high modernism, and the destruction of metis — read as the counter-dynamic, not a law of scale; C. Thi Nguyen, “Echo Chambers and Epistemic Bubbles” (Episteme, 2020); and, at machine scale, Ilia Shumailov and colleagues, “AI models collapse when trained on recursively generated data” (Nature, 2024), with Zhivar Sourati and colleagues, “The Homogenizing Effect of Large Language Models on Human Expression and Thought” (arXiv 2508.01491, 2025). On the widening side: Amartya Sen on famine and a free press (Poverty and Famines, 1981; Development as Freedom, 1999); and the Polis / vTaiwan platform (Christopher Small and colleagues, 2021) and bridging-based ranking (Aviv Ovadya, Belfer Center, 2022 — as in X’s Community Notes).

Within the framework: Chapter 4 — What Works (the methods-model this essay lifts to collective scale); Chapter 5 — Selves Made of Selves (the emergent we — where Michael Levin’s biology of coordinating cells belongs, rendered in AoM’s own terms of the widening context and the reach of concern); Coherence at Scale (nesting as a parallel search, and the Goodhart hazard at the seams); The Tree of Agreement (convergence below the surface); Held Together, Held Apart and Coherent Pluralism (the counter-dynamic, and metis kept alive as coherent pluralism); and the companion essays More Than Its Notes (what a we values) and The Course We Carve (how it decides).