203 – Maps, models, and metaphors

The tools we think with

There is a map of London’s Underground railway that has been reproduced more than a billion times. It appears in every station, on every tourist guide, on the phones of every visitor who has ever tried to find their way from Paddington to Borough Market. It was designed by an electrical engineer named Harry Beck in 1931, working in his spare time, and it broke every convention of cartographic accuracy in existence. The distances between stations bear no relationship to the actual distances underground. The curves and bends of the tunnels have been flattened into clean horizontal, vertical, and forty-five-degree lines. The River Thames, which the trains cross at several points, is reduced to a mild decorative squiggle. Geographically, the map is wrong in almost every measurable sense.

It is also, for the purpose of navigating the Underground, essentially perfect. Beck’s insight was that passengers do not need to know where the stations are in relation to the surface streets. They need to know which line to take, where to change, and in which direction to travel. For this purpose, topographical accuracy is not only unnecessary — it is actively counterproductive, because a geographically accurate map of the Underground would be so cluttered with curves, overlapping lines, and inconsistently sized distances that the relevant information would be buried in the noise. Beck’s map works because it omits. Its utility is a direct consequence of its incompleteness.

This is what a model is. Not a copy of the world, not a miniature version of reality, but a representation built for a specific purpose — one that sacrifices fidelity to some features in order to illuminate others more clearly. The Underground map is a model. Newton’s equations of motion are a model. The story you tell yourself about why your last relationship ended is a model. What connects them is the same structural property: each one selects, simplifies, and emphasizes. Each one renders certain features of its subject vivid and clear while leaving others in the dark. And each one, used outside the conditions for which it was built, will mislead you with the same confidence it provides when used correctly.

This article is about the tools we think with — about the maps, models, and metaphors that structure human thought from the most abstract mathematics to the most intimate self-understanding — and about what it means to use them well.

The map that is not the territory

In 1931, the same year that Beck was sketching his Underground map on spare scraps of paper, a Polish-American philosopher and engineer named Alfred Korzybski delivered an address to the American Mathematical Society in which he introduced a phrase that has since become one of the most widely quoted in the philosophy of language.¹ The map, Korzybski observed, is not the territory. The word is not the thing. The name is not the object named.

The observation sounds almost too simple to be worth making. Of course the map is not the territory — everyone knows that a piece of paper is not a city, that the word “fire” does not burn, that the name of a river is not wet. But Korzybski’s point was not about the obvious distinction between symbols and the things they symbolize. It was about something deeper and more consequential: that human beings, uniquely among the animals we know of, have largely lost direct access to the territory and navigate instead through an elaborate, multilayered network of maps. We do not experience the world directly. We experience our representations of it, and we mistake those representations for the thing itself so consistently and so automatically that the distinction disappears from ordinary awareness entirely.

Consider what happens when you read a sentence like “the economy is struggling.” The phrase arrives in consciousness as though it referred to a concrete, observable thing — something you could point to, measure, and assess. But an economy is not a thing in the world in the way that a rock or a river is. It is a model: a conceptual framework that aggregates millions of individual transactions, behaviors, expectations, and institutional arrangements into a single coherent-sounding noun. Different economists, applying different models, will tell you that the economy is struggling for entirely different reasons and will recommend entirely different remedies. They are not disagreeing about the territory. They are using different maps — different frameworks for selecting, aggregating, and interpreting the same vast body of observations — and the maps lead to different conclusions. The appearance of a shared subject of discussion conceals a genuine absence of shared ground.

This is not a special pathology of economics. It is the normal condition of human thought about anything complex. We think in categories, and categories are models: they select certain features of the world as relevant for grouping and ignore others. We think in narratives, and narratives are models: they impose a sequence, a causation, and a meaning on events that might be arranged differently by a different storyteller. We think in theories, and theories are models: they identify which variables matter and which can be safely ignored for the purposes of a given analysis. The categories, narratives, and theories are not the things they describe. They are tools for navigating those things. And like all tools, they work better in some conditions than others, and fail in ways that are sometimes predictable and sometimes catastrophic.

The metaphors inside the thinking

The philosopher George Lakoff and the linguist Mark Johnson argued in their 1980 book Metaphors We Live By that the relationship between language and thought goes considerably deeper than most people assume.² Their central claim was that abstract human thought is not merely described by metaphors but constituted by them — that the conceptual structures through which we understand abstract domains are themselves metaphorical, built from the more concrete and embodied structures that we encounter directly in physical experience.

Time, for instance, is not something we can experience directly as an abstract quantity. We experience it through space: the future is in front of us, the past is behind us, deadlines approach and pass, a long time is a long distance, a short time is a short one. These spatial metaphors for time are so deeply embedded in English — and in most other languages — that they feel like descriptions of the thing itself rather than models of it. But they are models. Other cultures have organized their temporal metaphors differently: some languages place the future behind the speaker (where it cannot be seen) and the past in front (where it can be observed), a mapping that is equally coherent and equally different from ours. The territory — whatever time actually is — remains the same. The maps differ.

Arguments, in English, are understood primarily through the metaphor of combat: positions are attacked and defended, claims are shot down, opponents are demolished, criticisms hit home or miss the mark. This is not the only possible way to model argument. Lakoff and Johnson propose an alternative: what if arguments were understood as collaborative construction — as buildings whose walls are raised together by both parties, whose foundations are shared, whose integrity requires contributions from everyone involved? Such a model would make different features of argumentative exchange salient: the importance of building on what the other person has said, the danger of undermining shared foundations, the value of a structure that is robust enough to stand up. The territory — the actual exchange of reasons and objections between human beings trying to understand something — is the same. The map shapes what is noticed and what is not.

This matters because the metaphors we use to model abstract domains are not neutral. They foreground some aspects of the domain and obscure others, and the aspects they obscure are often exactly the ones that would complicate the conclusions the metaphor makes available. The metaphor of the economy as a machine suggests that it can be repaired by identifying the broken parts and fixing them, and that its behavior is deterministic and predictable — both of which are misleading in domains where the components respond to each other and to predictions about their own behavior. The metaphor of the immune system as an army defending the body against foreign invaders captures something real but makes it harder to understand autoimmune conditions, where the army attacks the self. The metaphor of addiction as a moral failing makes treatment approaches that work — harm reduction, medication-assisted therapy — conceptually invisible, because they do not fit the model’s prescriptions. The map is not the territory, and the map shapes what remedies seem available.

The deeper maps

Beyond the metaphors that structure specific domains of thought, the philosopher Charles Taylor has described what he calls background frameworks: the deeper, largely unarticulated structures within which particular beliefs and values make sense.³ A background framework is not a belief — it is not something you can simply choose to revise the way you might revise a factual claim. It is the set of assumptions that makes beliefs intelligible in the first place: what counts as a reason, what counts as harm, what counts as a good life, what kinds of explanation are satisfying and what kinds leave a question unanswered.

Taylor’s insight is that these background frameworks vary across cultures, historical periods, and individuals in ways that are mostly invisible from inside any particular framework. We do not experience our background assumptions as assumptions — we experience them as the obvious, natural, self-evident structure of reality. It is only when we encounter someone operating from a genuinely different framework that the background becomes visible, and even then it tends to be perceived not as a different but equally coherent organization of experience but as a simple failure to understand something obvious.

This is one of the reasons why the most important disagreements — about values, about meaning, about what kind of life is worth living — are so resistant to resolution through argument. When two people disagree within a shared framework, argument can in principle resolve the disagreement by appealing to shared standards of evidence and inference. When two people disagree across frameworks, the disagreement is not at the level of conclusions — it is at the level of what counts as a good argument, what counts as relevant evidence, and what kind of answer would be satisfying. The debate about abortion, for instance, is not primarily a debate about facts. It is a debate about which model of personhood, which conception of bodily autonomy, and which framework of competing obligations should be used to analyze the question — and the frameworks are not themselves subject to resolution by the methods that work within them. Article 208 of this series examines what can be done in conversations where the frameworks themselves differ. What is important to register here is that the frameworks exist, that they are maps of the deepest kind, and that they are largely invisible from inside.

Maps of meaning

There is one further dimension of human model-building that is different in kind from the metaphors and frameworks discussed so far. The psychologist and professor Jordan Peterson has developed, over decades of lecturing and writing, an account of the relationship between human beings and the deep narrative structures — the myths, stories, and archetypes — that constitute what he calls Maps of Meaning.⁴

Peterson’s central argument is that the maps through which human beings navigate the world are not primarily descriptive — they do not primarily represent how things are. They are primarily prescriptive — they represent how to act. A myth about a hero who descends into the underworld and returns transformed is not, in the first instance, a claim about what happens to dead people. It is a model of how to engage productively with difficulty and uncertainty: enter the chaos willingly, endure the worst, and emerge with something valuable that would not have been available without the descent. This model has been refined across many generations in many cultures because it captures something real about the structure of meaningful human experience — something that the more literal and factually precise maps of scientific description do not capture and are not designed to capture.

The distinction between descriptive and prescriptive maps is one of the most important in the series. Scientific models are primarily descriptive: they aim to represent how things are, as accurately as possible, in domains where accuracy can be tested against observation. Narrative and mythological models are primarily prescriptive: they aim to represent how to act, what to value, and what kind of person to become, in domains where those questions cannot be settled by observation alone. Both kinds of map are necessary. The person who has only descriptive maps — who can account for every variable in the evolutionary origin of the maternal instinct but has no model of what it means to love a child — is not better equipped to live a human life than the person who has only prescriptive maps. The map of meaning and the map of mechanism are different tools for different purposes, and confusing them in either direction produces characteristic errors.

The error of applying prescriptive maps to descriptive questions is what produces pseudoscience: the insistence that the world must be a certain way because of a moral or narrative commitment, regardless of what the evidence shows. The error of applying descriptive maps to prescriptive questions is what produces a different and equally serious confusion: the belief that science can settle questions about how to live, what to value, and what obligations we bear to each other — when in fact these questions require the kind of map that science, by its own methodological commitments, does not produce. The Conscious Look, applied here, requires knowing which kind of map a given question requires and resisting the temptation to apply the wrong kind because it is the one currently to hand.

The maps we build to live in

Every map discussed so far in this article was made for navigation — built to help its users find their way through a territory that exists independently of the map. The Underground map does not determine where the stations are. Newton’s equations do not determine the paths of planets. The myths and archetypes that Peterson describes do not determine the structure of meaningful human experience; they are attempts to represent a structure that was already there. All of these are maps of territories that push back — territories that correct the map when the map is wrong, that impose costs when the navigator follows a line that does not exist, that insist on their own reality regardless of what the representation says.

The digital environments that now constitute a significant portion of daily life for hundreds of millions of people — and the majority of waking experience for a growing fraction of the young — have a property that distinguishes them from every previous human technology: they are not maps of a pre-existing territory. They are territories built to match the map. The social world of a platform is designed around the map of human social psychology that its engineers have constructed — around the finding that validation from strangers triggers the same neurological response as validation from friends, that outrage spreads further than reflection, that visual beauty can be separated from embodied presence, that the feeling of achievement can be delivered without the friction that normally produces it. These are not incidental features. They are the product, optimized by billions of data points and continuous refinement, for maximum engagement. The platform is a territory engineered to feel like the ideal version of the map.

What is omitted in this engineering is precisely what the previous maps could not omit: the resistance of reality. The physical world pushes back. The car breaks down. The climb is harder than expected. The relationship requires negotiation that no amount of charm resolves, because the other person has their own intractable needs and their own model of what is happening. The argument has consequences that persist after the screen is closed. The mistake cannot be undone by pressing a button. It is in this friction — in the resistance that the territory offers to the map — that most of what matters about human development actually lives. Resilience cannot be built in an environment where failure has no cost. Social skill cannot be developed in an environment where the difficult parts of social life — the misreading, the repair, the slow negotiation of incompatibility — have been optimized away. The capacity to tolerate boredom, uncertainty, and the slow pace of real-world consequence cannot be cultivated in an environment engineered to prevent those states from arising.

The danger this creates is not primarily the danger that is usually discussed — addiction, distraction, the colonization of attention. Those are real but familiar. The deeper danger is epistemological: it is the gradual, largely invisible process by which the map becomes the reference point against which the territory is judged inadequate. The person who has spent formative years in digital environments calibrated for maximum responsiveness, social validation, and consequence-free self-presentation does not merely find the physical world less entertaining. They find it harder to read. The social signals are ambiguous in ways that the digital ones are not. The feedback is slow, indirect, and often absent. The self that emerges in the encounter with actual others — awkward, uncertain, sometimes rejected, often misunderstood — does not match the self that the digital map has been showing them, and the mismatch is experienced not as useful information about the territory but as a deficiency of the territory. The world is the problem. The map is correct.

There is a further and more intimate version of this danger. In digital environments, identity itself becomes a map — a curated, edited, performance-optimized representation of the self, designed to elicit a specific response from a specific audience. This is not new: human beings have always presented different faces in different contexts, and the management of self-presentation is one of the oldest and most sophisticated skills in the social repertoire. What is new is the degree of control, the persistence of the record, and the optimization of the feedback. The digital self can be adjusted until it produces the maximum response. Unflattering images can be deleted. Awkward moments need not be survived — they can be simply not posted. The model of the self that the digital environment reflects back is not the self that encounters friction, makes mistakes, and is seen in its weakness. It is the self that was selected, from among all available versions, as the most likely to be liked. For a person in the stages of development where identity is still being formed, the difference between these two selves matters enormously — and the digital mirror, because it is more polished and more responsive and more flattering than any previous technology, is also more dangerous as a source of self-knowledge.⁵

None of this is an argument against digital life, any more than recognizing the limits of the Underground map is an argument against taking the train. Maps of all kinds are necessary, and the maps that digital environments provide — of information, of social connection, of creative possibility — are genuinely valuable in the domains for which they were built. The argument is the same one this series makes about every other map: know what it omits, know the conditions under which it fails, and resist the temptation to mistake it for the territory it represents. The territory, in this case, is the physical, social, embodied world that exists independently of any screen — the world where the car eventually breaks down, where the other person does not behave as the model predicted, and where the self that emerges is the one that was actually there all along, rather than the one that was curated for approval.

The Conscious Look, applied to maps

The practice this series recommends is not the abandonment of maps. That option is not available — thought without models is not purer thought, it is no thought at all. The practice is the periodic, deliberate examination of the maps themselves: asking what they include, what they omit, what purpose they were built for, and whether the situation currently being navigated is within their domain of valid application.

This examination has a characteristic shape that the series’ diagnostic question captures: what would have to be true for this model to be wrong? The Underground map is wrong whenever someone uses it to estimate walking distances between stations — it was not built for that purpose and gives systematically misleading results when applied to it. The metaphor of argument-as-combat is wrong whenever it makes collaborative truth-seeking harder — when the emphasis on winning a position prevents the incorporation of genuinely good points from the other side. The background framework is wrong, or at least limited, whenever it encounters a genuine challenge from a different framework and responds by rendering the challenge invisible rather than engaging with it.

Beck’s Underground map has a printed legend that tells you it is not to scale. The maps we carry in our minds rarely come with equivalent warnings. The work of identifying what those warnings would say — where the map deviates from the territory, where it was built for a different purpose than the one currently being pursued, and what features of the territory it was designed to leave in the dark — is the work that this series calls The Conscious Look. It is slow, uncomfortable, and never finished. It is also, given that the maps are what we have, the only honest way to use them.

Notes

¹ Alfred Korzybski (1879-1950) introduced the phrase “the map is not the territory” in his address “A Non-Aristotelian System and its Necessity for Rigour in Mathematics and Physics,” delivered to the American Mathematical Society in New Orleans in 1931. The phrase appears in the published version of the address and was subsequently developed at length in his major work, Science and Sanity (1933). Korzybski’s broader project, which he called General Semantics, was an attempt to apply scientific rigor to the study of how language shapes thought and behavior. Though the project never achieved the academic respectability its author sought, the map-territory distinction has passed into common currency in philosophy, cognitive science, and epistemology in ways that would have pleased him.

² Lakoff, G., and Johnson, M. (1980). Metaphors We Live By. University of Chicago Press. The claim that abstract thought is constituted by metaphor, not merely expressed through it, is what makes the book philosophically significant rather than merely linguistically interesting. If the claim is right — and the evidence from cognitive linguistics and neuroscience has generally supported its weaker versions — then the choice of metaphor for an abstract domain is not merely a stylistic decision but a cognitive one: it determines what is thinkable within that domain. The implications for philosophy, political science, and ethics are substantial and largely still being worked out.

³ Taylor, C. (1989). Sources of the Self: The Making of the Modern Identity. Harvard University Press. The concept of background frameworks — what Taylor also calls horizons of significance — is developed most fully in this major work, though it appears throughout his philosophical writing. Taylor’s argument is that the frameworks are not merely cultural overlays on a neutral cognitive architecture: they are partly constitutive of the self, in the sense that who a person is cannot be fully described independently of the framework within which they understand what matters and why. This is the philosophical background for the self-model discussion in article 906 of this series.

⁴ Peterson, J. B. (1999). Maps of Meaning: The Architecture of Belief. Routledge. Peterson’s argument draws on the Jungian tradition of archetype and collective unconscious, on the ethological research of Konrad Lorenz and Nikolaas Tinbergen, on the neuropsychology of Jaak Panksepp, and on the phenomenological tradition of Martin Heidegger and Maurice Merleau-Ponty. The synthesis is ambitious and not uncontroversial, particularly in its treatment of the relationship between biology and culture. The central distinction between maps that describe how things are and maps that prescribe how to act is, however, robust and important regardless of the larger theoretical framework in which Peterson embeds it. The 1999 book is considerably more careful and more fully argued than the popular lectures, which compress the argument in ways that sometimes sacrifice precision for accessibility.

⁵ The psychologist Sherry Turkle has studied the relationship between digital technology and identity development over four decades, beginning with her early work on the psychology of computer use and culminating in Alone Together (2011) and Reclaiming Conversation (2015). Her central observation — that digital environments enable a kind of self-presentation that is more controlled and more edited than any previous medium, and that this control comes at the cost of the vulnerability and unpredictability that genuine self-knowledge requires — is the empirical grounding for the argument made here. The developmental psychologist Jean Twenge’s iGen (2017) provides the generational data: the cohort that grew up with smartphones and social media from early adolescence shows measurable and systematic differences from preceding cohorts in rates of anxiety, depression, loneliness, and — perhaps most instructively — in the experience of the physical social world as more threatening and more difficult to navigate than digital alternatives. The argument made here does not require accepting all of Twenge’s conclusions, which have been contested in the literature. It requires only the more modest claim that a map optimized for engagement rather than accuracy is a map that systematically misrepresents its territory — and that the territory in question is one whose accurate representation matters for human development.