This is the second part of a pair. Part One, You Cannot Know the Future. You Can Model It., works through perception, prediction and calibration inside a single mind. This one asks what happens when two of those minds meet.
In Part One, I explored an idea that has become increasingly important to me: you cannot know the future; you can model it. The mind does not simply receive reality. It interprets incomplete information, generates predictions, assigns confidence to them, and acts on those predictions. Sometimes those models are well calibrated. Sometimes they are distorted by fear, attachment, previous experience, or simply by the fact that there is not enough information available to know what is actually happening.
But that line of thinking led me somewhere I hadn’t expected. If each of us is continuously building a model of reality, what exactly happens when two people talk? The obvious answer is that they exchange information, but I think something more interesting happens: we exchange perceptions. We give one another temporary access to different ways of organising and experiencing the world.
Once I started thinking about conversation this way, an enormous amount of human behaviour began to look different.
Conversation as perceptual exchange
Imagine that two people watch exactly the same episode of a television series. They have technically received almost identical sensory information: the same dialogue, performances, shots, music, and story. Yet they do not necessarily have the same experience of it. One person notices a visual callback to an earlier episode. Another notices that a character’s behaviour has changed. Someone with a background in filmmaking notices how the editor constructed a particular scene. Someone familiar with the history being referenced recognises an analogy that passes everyone else by.
Afterwards, they talk. One person says, “Did you notice what happened in the background of that scene?” You didn’t. Now you do. The episode hasn’t changed, but your model of the episode has. If you watch it again, you may literally notice something that was perceptually invisible to you the first time.
I think this is part of the reason the desire to discuss art feels so natural. After watching something I love, I often want to read what other people thought about it or talk about it with friends. That isn’t necessarily because I need confirmation that it was good. Often I want to know: what did everyone else see? Someone else’s interpretation can become an additional perceptual lens. In a limited but meaningful sense, they lend me part of their mind.
Expertise creates different perceptual worlds
This becomes even clearer when you spend time with experts. Walk through a forest with a botanist and you may see trees, flowers, moss, and leaves. They may see species, disease, nutrient deficiency, competition for sunlight, succession, fungal relationships, invasive plants, and evidence about the history of the landscape. Walk through a city with an architect and the buildings change. Listen to an orchestra with a composer and the music changes. An experienced doctor can look at a patient and notice signals in details that appear irrelevant to almost everyone else.
The beginner and the expert inhabit the same physical environment, but they don’t entirely inhabit the same perceptual environment. Learning therefore does something much more profound than increasing the amount of information stored in memory. It changes what becomes available to experience.
Knowledge increases the resolution of the world.
Once you learn a distinction, you can start seeing that distinction everywhere. Learn about incentives and organisations start looking different. Learn about evolution and biology changes. Learn about composition and photography changes. Learn about cognitive biases and arguments change. Learn about predictive processing and even your own thoughts begin to look different.
More importantly, these additions aren’t necessarily linear. Each new lens can interact with the lenses you already possess. Learning about game theory after learning about psychology gives you combinations neither subject provides independently. Add evolutionary biology, economics, design, history, or philosophy and the number of possible interpretations grows again. Learning can therefore increase not only the amount you know but the dimensionality of what you are capable of experiencing.
Stories as compressed perception
This also offers a way of thinking about why humans are so obsessed with stories. A story isn’t simply a sequence of events. It is a way for one mind to compress an experience into a form that another mind can partially reconstruct.
I cannot directly experience another person’s childhood, a war they lived through, a relationship they lost, or the society they grew up in. Through narrative, however, I can construct some approximation of their perspective. The reconstruction is obviously incomplete, but it is powerful enough that humans have spent thousands of years building civilisation around it.
Stories allow experience to travel. Someone encounters something, learns from it, compresses the experience into language, and passes it on. Someone else can then update their model of the world without having to live through the original event. At the simplest level this might be, “Don’t eat those berries.” Scale the same basic ability across generations and you eventually get mythology, literature, law, history, science, philosophy, and education.
Culture, viewed this way, isn’t only a store of information. It is an enormous archive of ways of seeing. Every story says, in some form:
Here is what the world looked like from where I stood.
Language lets us learn from other people’s simulations
This may be one of the stranger powers of language. I can tell you about something that happened yesterday, but I can also tell you about something that might happen tomorrow, something that never happened, a fictional world, a counterfactual world, another person’s beliefs, or even my interpretation of your interpretation of me.
That means the brain no longer has to learn only from direct experience. It can learn from another brain’s model. I experience something and form a model of it. I compress part of that model into words. Those words reach you, and your brain reconstructs an approximation of what I experienced. Your own model can then change despite never encountering the original event.
experience→model→language→reconstruction→changed model
Conversation therefore allows something resembling prediction error to travel between minds. You can tell me, “I think you’re looking at this completely wrong,” and suddenly I have access to a possibility my own model may never have produced. Sometimes you make my perception clearer; sometimes you make it worse. Either way, you have changed the set of hypotheses available to me.
Why disagreement can be valuable
This gives disagreement an interesting role. If everyone around me sees things exactly as I do, the experience may be comfortable, but it often provides relatively little new information. Then someone says, “I think the exact opposite,” and suddenly I have discovered that the same observations can support another model.
I can now investigate where the difference comes from. Maybe they know something I don’t. Maybe we assign different weights to the same evidence. Maybe our underlying assumptions differ. Maybe one of us has experienced something the other hasn’t. Maybe they are simply wrong. Even then, understanding why their model exists can teach me something about the structure of the problem.
An intelligent person who regularly disagrees with you may therefore sometimes be more epistemically useful than someone who reliably confirms your conclusions. Not because disagreement is inherently virtuous, and certainly not because every perspective deserves equal weight, but because difference contains potential information.
Other people increase the variety of the system
This has an interesting implication for groups. A group of ten people isn’t valuable merely because it contains ten times as many working hours. If those people genuinely perceive a problem differently, the group may have access to much more of the problem space than any individual member.
One person sees technical constraints, another sees incentives, another notices user behaviour, another sees regulatory risk, and someone else notices a social consequence nobody had considered. If those perspectives can be integrated without eliminating the differences that made them valuable in the first place, the collective model can become richer than any individual’s.
One way human beings cope with the overwhelming complexity of reality is therefore surprisingly simple: we use other humans. Our perceptual system is, in an important sense, partly social.
And that leads to a question I now find much more consequential.
Who gets to perceive on my behalf?
I cannot directly inspect everything that matters to me. I don’t personally read every scientific paper. I can’t travel to every country, witness every political event, reproduce every experiment, audit every institution, or independently verify every claim I encounter about economics, medicine, technology, war, history, or culture.
So I outsource perception. We all do. We rely on researchers, journalists, experts, friends, institutions, communities, critics, eyewitnesses, and increasingly algorithms. This outsourcing is unavoidable; civilisation could hardly function without it.
But it creates enormous power.
If you can determine which perceptions repeatedly reach large numbers of people, you don’t necessarily need to directly control what those people believe. You can alter the distribution of evidence, interpretations, and experiences from which their beliefs are constructed.
That is a subtler form of influence, and potentially a more powerful one.
Why media power is perceptual power
This is why the familiar claim that control of media can shape public thought contains an important insight, even when it is sometimes stated too crudely. Media doesn’t simply tell us what to think. More fundamentally, it helps determine what enters the perceptual field at all.
Imagine that ten thousand potentially important things happen in a day and a news organisation can meaningfully cover fifty. Which fifty does it choose? Which becomes the lead story? Which photograph accompanies it? Which expert is interviewed? Which quotation becomes the headline? Which contextual fact appears in paragraph two, and which appears in paragraph nineteen? Which events are described as isolated incidents and which are presented as examples of a broader trend?
None of those choices require anyone to fabricate information in order to alter the resulting perception. Attention itself is selective. Make one category of event highly salient and another almost invisible, and an audience’s model of the world changes.
A person can quite reasonably infer, “This must be happening everywhere,” when the deeper truth is simply, “This is what my information environment repeatedly showed me.” The distinction between those statements is enormous.
This also makes “Who gets to perceive on my behalf?” a useful question when consuming news. I am not merely choosing a source of facts. I am choosing, to some extent, whose attention I am borrowing: who decides which parts of an impossibly large world are worth placing in front of me.
Social media adds perceptual customisation
The internet makes this much stranger.
Historically, the people whose perspectives you encountered were heavily constrained by geography and circumstance: family, neighbours, coworkers, local institutions, religious communities, newspapers, perhaps a relatively small number of broadcasters. People obviously still selected friends, publications, political groups, and communities, but there were relatively hard limits on how completely you could customise the minds around you. There was unavoidable epistemic friction. You encountered people with views you disliked simply because they were there.
Social media gives us something historically unusual: high-resolution control over whose perceptions repeatedly enter our environment.
The upside is extraordinary. I can listen to an astronaut explain orbit, a surgeon describe an operation, a person living through a conflict, a mathematician explain topology, an expert discuss an obscure historical event, or someone from a culture I may never physically visit. The range of perspectives available to an ordinary person is astonishing. Used one way, the internet may be the most powerful system for perceptual expansion humanity has ever built.
But the same capability has another side. We can select perspectives because they feel compatible with the models we already possess, while algorithms learn which perspectives keep us engaged. This creates two interacting filters: I choose what fits me, and the system learns what affects me and gives me more of it.
The result can be an environment in which our beliefs receive enormous amounts of apparent confirmation while encountering relatively little corrective information.
A custom-built reality can still feel completely real
This is deeper than the usual discussion about “echo chambers.” The important point is that a person’s subjective experience may be entirely genuine.
Spend several hours inside a highly specialised online community and it can easily feel as if “everyone is talking about this.” Perhaps they aren’t. Millions of people may barely know the issue exists. But the person’s information environment really did contain it everywhere. Their conclusion is based on evidence; the problem is that the evidence comes from a radically non-random sample.
Social media therefore allows us to construct something like a custom perceptual organ. But that organ doesn’t necessarily sample reality evenly. It samples a selection of other people’s interpretations, filtered by our preferences, social networks, platform incentives, and recommendation systems.
Our perception of society itself can consequently be generated from a deeply distorted sample of society. We can become highly confident that we understand “what people think” while mostly observing a population selected because it repeatedly generates content we are likely to consume.
A model can receive enormous amounts of confirming evidence without becoming more accurate if the process generating that evidence is itself biased.
Identity makes perspectives harder to update
There is another complication. Perspectives can become attached to identity.
Believing X is one thing. Believing that people like me believe X is something else entirely. Now updating the belief may affect not only my model of reality but my status, friendships, community, moral identity, or sense of belonging.
At that point, belief formation starts to resemble game theory. The mind isn’t only asking, “Is this true?” It may also be asking, “What happens to me if I believe this?”
Communities therefore don’t merely exchange perceptions. They create incentives around which perceptions are acceptable. Certain conclusions receive approval or status; others receive punishment. That can happen in political groups, companies, academic fields, fandoms, religious communities, professions, and friend groups.
A perception can become socially expensive before it ever becomes intellectually implausible.
The paradox of perspective
This brings me to the part I find most beautiful.
At the beginning of learning something, it is easy to imagine that there is simply a correct way to see it. You learn a little and become more confident. Then you learn more, and eventually something peculiar happens: the world doesn’t become simpler. It becomes larger.
You discover exceptions, competing models, different levels of analysis, historical contingencies, unknown variables, and people who know far more than you who nevertheless disagree with one another for intelligent reasons. The number of possible perspectives starts expanding faster than your ability to master them.
The progression can look something like this:
I don’t know much ↓ I have learned the answer ↓ I have learned enough to understand how many answers, variables, and perspectives I hadn’t even known existed
There is an interesting connection here to the Dunning–Kruger effect, although the popular version of Dunning–Kruger, usually represented by a neat curve in which ignorant people possess enormous confidence, is much cruder than the underlying research.
The connection that interests me isn’t simply that “people who know less are more confident.” It is that limited knowledge can restrict your awareness of the space of possible knowledge.
If I know only one framework, I don’t merely lack the other nineteen. I may not know that the other nineteen exist.
From inside that restricted model, my framework can therefore feel surprisingly complete.
Once I know twenty, something changes. I begin to understand that every framework reveals some things and obscures others, that there may be perspectives I haven’t encountered yet, and that the number of possible ways of interpreting a complex situation is vastly larger than I originally imagined.
Expertise can produce humility for structural reasons
This kind of humility doesn’t have to be a moral virtue. It can simply be a consequence of possessing a larger model.
A beginner may see three relevant variables. An expert sees thirty. A deeper expert may know that there are another hundred they cannot adequately model. Their increased uncertainty isn’t necessarily evidence that they understand less. It may be evidence that they have become aware of more ways in which they could be wrong.
This gives us a lovely paradox:
The more perspectives you acquire, the harder it becomes to mistake any single perspective for the world itself.
That doesn’t mean all perspectives are equally good. Some models are clearly more predictive, coherent, empirically grounded, or useful than others. The conclusion isn’t relativism. If anything, it is another argument for calibration.
A lens can be extraordinarily powerful without being the landscape.
Perhaps this is what curiosity is really doing
This also gives me a different way of thinking about curiosity. Curiosity isn’t merely the desire to accumulate facts. It is a drive to expand the number of ways in which reality can become intelligible.
- A physicist learns to see forces, fields, systems, and constraints.
- A doctor learns to see symptoms and physiological relationships.
- An artist learns to see form, negative space, colour, gesture, and composition.
- A psychologist learns to see motives, conditioning, attachment, and cognition.
- A child learns that the strange point of light in the sky is a planet.
The physical world hasn’t necessarily changed. The available experience has.
Because each new perceptual lens can combine with the ones already acquired, experience may not increase linearly. It can become combinatorial. Learning one new framework doesn’t simply add one new thing to experience; it can change how hundreds of things already known relate to one another.
Perhaps this is one of the deepest pleasures of curiosity. We aren’t merely accumulating information about the same world.
We are increasing the number of worlds available inside it.
Conversation may be one of humanity’s great perception technologies
This finally brings me back to ordinary conversation.
Perhaps part of the reason a genuinely good conversation feels so pleasurable is that two minds can repeatedly generate useful surprises for one another. I say something you hadn’t considered, so your model changes. You connect it to something I hadn’t considered, so my model changes. That produces another thought in me, which produces another thought in you. Soon the conversation has reached somewhere neither person could have produced independently.
It stops behaving like simple transmission from A to B and becomes a feedback loop:
A changes B→changed B changes A→changed A changes B→…
At its best, conversation is a coupled generative system. Something new appears between the participants that fully belonged to neither of them beforehand.
Perhaps this is why the best conversations can feel so alive. We aren’t simply exchanging facts. We are lending one another temporary ways of seeing.
Stories do this. Art does it. Science does it. Friendship does it. Teaching does it. Travel does it. The internet does it at planetary scale. Journalism does it, which is precisely why deciding who gets to direct our attention matters so much.
Sometimes the borrowed lens distorts. Sometimes it clarifies. Sometimes it reveals something that was previously invisible. Usually it does some mixture of all three.
But once you have genuinely seen the world through another perspective, it can be difficult to return to exactly the world you inhabited before.
And perhaps that is one of the deepest things human beings trade with one another:
Not information, but possibilities for perception.
