S1-304 – Mathematics: The language that doesn’t lie
What it cannot say
In 1960, the physicist Eugene Wigner published an essay with a title that has become one of the most quoted in the philosophy of science: “The Unreasonable Effectiveness of Mathematics in the Natural Sciences.” Wigner’s puzzle was genuine and has not been resolved in the 65 years since he posed it. Why should the abstract structures that mathematicians develop (often for purely aesthetic reasons, with no application in mind, in response to problems that seem entirely internal to mathematics) turn out, decades or centuries later, to describe the physical world with extraordinary precision? The complex numbers that were invented to solve algebraic equations that had no real solutions turned out to be the natural language of quantum mechanics. The non-Euclidean geometries that Riemann developed in the nineteenth century as a purely theoretical exploration of what geometry might look like if space were curved turned out to be exactly the geometry that Einstein needed to describe gravity. The group theory that mathematicians developed to study abstract symmetries turned out to describe the symmetries of elementary particles with a precision that has produced the most accurately tested predictions in the history of science.
Wigner’s puzzle is not merely an interesting curiosity about the relationship between mathematics and physics. It points to something fundamental about the nature of mathematics as a language: about what kind of tool it is, what it can do that no other human language can, and what it cannot do at all. These are the questions this article addresses.
Invented or discovered?
Wigner’s puzzle sharpens a question older than his essay, one that still divides thoughtful mathematicians and philosophers and shows no sign of settling: when a mathematician reaches a new result, is it invented or discovered? Did the mathematician make it, the way a composer makes a symphony, or find it, the way an explorer finds a coastline that was there all along? Nothing in the daily practice of mathematics decides the matter, because the experience fits either reading. A proof feels like construction while one builds it and like discovery once it stands. But the two answers point at very different pictures of what mathematics is, and the effectiveness Wigner marveled at looks entirely different depending on which one is right.
On one side is the discovered view, often called mathematical Platonism, which holds that mathematical objects and truths exist independently of human minds, and that mathematicians find them rather than make them. The primes were infinite in number before anyone proved it, and would stay so if every mathematician vanished. G. H. Hardy put the position plainly: mathematical reality lies outside us, and the theorems we grandly call our creations are only our notes on what we have observed.¹ The view has real force. Results feel inevitable and are not open to a vote; mathematicians working continents apart arrive at the same theorem; and Wigner’s effectiveness is just what one would expect if the universe were mathematical at its root and mathematics were the reading of its actual structure. Gödel held a version of this view, and so does Roger Penrose, for whom the mathematical world is as real as the physical one and no less strange.
On the other side is the invented view, which holds that mathematics is a human construction with no existence apart from the minds and cultures that build it, and it comes in several forms. The formalist treats mathematics as a game played with symbols under chosen rules, true only in the way a chess move is legal. The fictionalist, in Hartry Field’s version, holds that numbers exist no more than the characters in a novel, and that mathematics is a useful fiction we could in principle do without. The cognitive scientists George Lakoff and Rafael Núñez argue that mathematics grows out of the human body and brain, built up by metaphor from our experience of moving, grouping, and counting, so that it is a human making through and through.² On these views Wigner’s puzzle softens, because if we drew mathematics from the physical world in the first place, it is small wonder that it fits the world when we lay it back down.
Between the two sits a position many working mathematicians find the most honest, and it is worth stating because it dissolves part of the quarrel. We invent the objects and the rules, but we discover their consequences. Nothing forced us to define the prime numbers; that concept is a human choice, a piece of notation and definition we might never have drawn. But once primes are defined, that there are infinitely many of them is not something anyone gets to decide. It was true before it was proven and could not have come out otherwise. On this reading mathematics is an invented language whose sentences, once its grammar is fixed, are then found to be true or false rather than made so. The map is drawn by us; the territory it then opens is not.
That last image is why the question belongs in this series at all, and here the framing is our own rather than settled philosophy. The invented-or-discovered debate is the map and the territory in their purest and hardest form. Everywhere else in these pages the map is plainly our construction and the territory plainly the world, and the discipline is to keep from confusing them. Mathematics is the one place where the map fits the territory so exactly that no one can say for certain whether we drew it or found it, and Wigner’s unreasonable effectiveness is simply what that perfect fit feels like from the inside. This series takes no side in the dispute, which remains genuinely open. It notes only that mathematics is where the central question of the whole project, how much of what we know is the world and how much is the mind’s own drawing, is at its most beautiful and least answerable.³
What mathematics gains: precision, universality, falsifiability
The most obvious advantage of mathematics as a language for describing the world is precision. A mathematical statement means exactly what it says, in the sense that its truth conditions are fully specified by its formal structure and require no interpretation. “The population grew” is a sentence in natural language. It can mean it grew by one person or by a billion, slowly or rapidly, continuously or in jumps, over a year or over a century. “P(t) = P₀ · eʳᵗ” is a mathematical statement. It means exactly one thing: that the population at time t is equal to the initial population multiplied by the exponential of the product of the growth rate and the time. There is no ambiguity in the mathematical statement, and no room for the interlocutor to understand it differently from the speaker. The precision is total.
This precision makes mathematical statements uniquely falsifiable in the sense that Popper identified as the mark of genuine empirical claims. A natural language claim that “the economy is growing” can be confirmed by almost any observation and falsified by almost none, because the claim’s vagueness makes it compatible with almost any state of affairs. A mathematical claim that “the economy will grow by 2.3 percent in the next quarter” is falsifiable in a specific and unambiguous sense: either it does or it doesn’t, and the difference between the prediction and the outcome can be measured exactly. The precision that makes the mathematical claim less poetic than the natural language claim is the same precision that makes it more informative: more capable of being wrong, and therefore more capable of genuinely engaging with the world.
The precision of mathematical language produces a further property that is particularly visible to anyone who has worked in the natural sciences: an extraordinary compression of content that natural language cannot approach. Maxwell’s four equations of electromagnetism (which in modern vector notation occupy a single line) encode the complete behavior of all electric and magnetic fields in space and time, the propagation of electromagnetic radiation including light, the operation of every electrical device built since the nineteenth century, and the relationship between electricity and magnetism that required two centuries of experimental work to establish.⁴ A natural language description of what these four equations contain would require volumes, and it would still be less precise than the equations themselves. The compression is not an abbreviation of something that could be said in words. It is a different kind of statement entirely: one that achieves in a line what natural language cannot achieve at any length.
This compression is not merely convenient. It is the reason that mathematics functions as the shared language across the STEM fields (the natural sciences, engineering, and technology) regardless of the national languages of the people working in them. A physicist in Germany, a chemist in Japan, and an engineer in Brazil share no spoken language, but they share the equations. The equations require no translation because the notation is the meaning. This is a form of universality that goes deeper than the cultural universality of, say, a lingua franca: it is not merely that everyone has agreed to use the same words, but that the words themselves carry the same content regardless of who reads them or where. The equation F = ma means the same force relationship in every laboratory on earth, and in every publication in every language in which it appears. Natural language has no equivalent. Even the most carefully crafted natural language statement changes slightly in translation, because the associations, connotations, and implicit assumptions of words differ across languages and cultures. The mathematical statement does not translate. It simply is.
The second advantage is universality. A mathematical truth is true in the same way regardless of the language in which it is expressed, the culture in which it was developed, or the particular observer who is examining it. The Pythagorean theorem is the same theorem in Arabic, Japanese, and English. The Mandelbrot set is the same set regardless of who is computing it or where. This universality is not a trivial property. It means that mathematical knowledge, unlike most human knowledge, is genuinely cumulative in a strong sense: mathematical results proven centuries ago remain results, and every subsequent development can build on them without rechecking their foundations. The theorems of Euclid are still true. The results of Newton’s calculus, translated into rigorous modern form, are still valid. Mathematics does not have paradigm shifts in the Kuhnian sense: it has extensions, generalizations, and discoveries, but it does not routinely discover that what it previously established was wrong.⁵
The third advantage is the capacity for long chains of inference. Natural language reasoning degrades with the length of the argument: after several steps of inference, the ambiguities of the natural language terms accumulate, the implicit assumptions multiply, and the conclusion may be far from what the premises actually support. Mathematical reasoning does not degrade in this way, because the formal precision of each step eliminates the ambiguity accumulation that plagues natural language arguments. This is why mathematics is the only language available for making precise claims about entities (subatomic particles, distant galaxies, the behavior of systems in extreme conditions) that human beings have never observed directly and cannot observe directly, and whose properties are known entirely through inference.
Transfer without reconstruction
There is a consequence of this universality that reaches back to something argued earlier in this pillar. Ordinary communication, as article S1-302 sets out, is not really transmission but reconstruction: a sentence is a compressed set of cues, and the listener rebuilds a model from them using materials already in their own head, so the picture that arrives is never quite the picture that was sent, and two people understand each other only to the degree they were already alike. The reconstructed model is always, to some extent, private. One person’s is not another’s.
Mathematics is the one place where this breaks down in the sender’s favor. The formal content of a mathematical statement is fixed by rule, not rebuilt from private association, so when the notation is read correctly the object that arrives in the second mind is, in principle, the very same object that left the first. There is no room for the receiver to fill in the blanks with their own priors, because the formalism leaves no blanks to fill. F = ma does not become a slightly different relation in each reader’s head the way a line of poetry becomes a slightly different image. The equation is not a cue to be reconstructed; it is the thing itself, and every competent reader arrives at the identical structure because the rules permit no other. This is what it means to say, as the previous section did, that the notation is the meaning: mathematics is the closest thing we have to genuine model transfer rather than model reconstruction.
But it does not repeal the law of S1-302 so much as satisfy it in the most extreme way possible. Communication converges on a shared model only to the degree the two minds were already alike, and mathematics simply drives that likeness to its limit by making the shared ground explicit and total. Two people can exchange a theorem without loss only because both have first spent years building, from the same axioms and definitions and rules of inference, an identical stock of formal machinery. The convergence is not free. It is bought in advance, by a training that installs the same model in both heads before the first message is ever sent. And even then, only the formal skeleton transfers intact. What each mathematician pictures on reading the theorem, why it strikes them as beautiful or obvious or deep, what it connects to and what it is for, is reconstructed in the old private way and stays their own. The proof is shared exactly; the understanding of it is not. Mathematics defeats the reconstruction problem for the one layer of meaning that can be fully formalized, and leaves every other layer just where S1-302 found it.
What mathematics loses: qualia, context, meaning
The precision of mathematics is purchased at a specific and significant cost. A mathematical description of a phenomenon is a description that includes only what can be formally represented: what can be expressed in terms of quantities, relationships, functions, and structures that have precise mathematical definitions. Everything else is excluded.
Consider what a complete mathematical description of a piece of music would contain. It would include the frequencies of the sound waves, their amplitudes, their durations, their timing relative to each other, and their propagation through the acoustic environment of the concert hall. It would contain, in principle, all the information needed to reproduce the physical event precisely. What it would not contain is anything about what the music means, what it feels like to hear it, what emotional response it produces in a listener, or why a particular passage is moving in a way that a slightly different arrangement of the same pitches would not be. These properties are not present in the mathematical description not because they are unimportant (they are arguably the most important properties of the music, the reason it exists) but because they resist the kind of precise formal representation that mathematics requires. They are, in Nagel’s terms from article S1-503, the subjective character of the experience: what it is like to hear the music, which no amount of objective mathematical description captures.
This is not a failure of mathematics. It is a structural feature of what mathematics is: a language built for precision and universality, which means a language that can only operate on what can be precisely and universally defined. The philosopher Iain McGilchrist, whose account of the divided brain runs throughout this series, describes mathematics as the paradigmatic left-hemisphere tool: it achieves its extraordinary precision and power by abstracting from context, ignoring what is not formalizable, and treating the world as a collection of quantities and relationships rather than as a field of meanings, values, and lived experience.⁶ The abstraction is not a distortion: it is the condition of the precision. But it is an abstraction. And what is abstracted away does not cease to exist because mathematics cannot represent it.
The cost of abstraction is particularly visible in the social sciences, where mathematical models are applied to phenomena (human behavior, social dynamics, economic activity) that have subjective and contextual dimensions that resist formalization. The model of rational economic man (homo economicus) that underlies much of formal economics is a mathematical entity: perfectly rational, fully informed, consistent in preferences, maximizing a well-defined utility function. This mathematical creature has properties that make it extraordinarily tractable as a modeling device: the mathematics of its behavior is elegant, its responses to price changes are calculable, its aggregate behavior in markets can be derived from first principles. What it is not is a model of actual human beings, who are inconsistently rational, limitedly informed, subject to emotions and social pressures, and whose behavior differs systematically from what the mathematical model predicts in ways that are themselves systematic and predictable, as the behavioral economics literature has documented in exhaustive detail.⁷ The precision of the mathematical model is real. The precision is purchased by excluding the features of human behavior that make it both interesting and important.
The map and the number
There is a specific failure mode of mathematical modeling that is worth naming precisely because it is so common in public discourse: the confusion of the precision of a number with the accuracy of a measurement. When a figure is expressed as a number (unemployment is 6.7 percent, the GDP growth rate is 2.3 percent, the inflation rate is 4.1 percent) the precision of the number carries an implicit claim about the precision of the underlying measurement that the number does not always warrant.
The unemployment rate, for instance, is expressed as a single percentage. The precision of the expression (not “about 7 percent” but “6.7 percent”) implies a measurement accurate to one decimal place. But the unemployment rate is a construct, not a measurement: it counts as unemployed only people who are not employed, who are actively seeking work, and who have looked for work in the specific reference period that the definition specifies. It excludes people who have given up looking, people who are underemployed, people who are working in the informal economy, and people whose employment status is ambiguous. The number 6.7 percent is precise. What it measures is a specific and contestable operationalization of the concept of unemployment that excludes much of what ordinary language means by the term. The precision of the expression is a feature of the mathematics. The accuracy of the measurement is a feature of the definition, and the definition is a choice, not a discovery.⁸
This is the map-territory distinction of article S1-203, applied to numerical quantities. The number is the map. The precision of the map is not the same as the accuracy of the map’s representation of the territory. A highly precise map (one that specifies distances to the nearest millimeter) is not a more accurate map than a less precise one if the millimeter-level precision is applied to distances that were measured to the nearest ten meters. The precision of a mathematical expression is a property of the expression. The accuracy of the expression as a representation of the world is a property of the relationship between the expression and what it is expressing. These are different properties, and confusing them (treating precise numerical expressions as though the precision of the expression entailed the accuracy of the measurement) is one of the most pervasive errors in the use of mathematical language in public discourse.
The most common form of this error is also the most mechanical, and it happens the instant a number is copied off a calculator or a spreadsheet cell. The machine returns as many digits as its display can hold, with no regard for how well the quantities fed into it were known. Multiply two lengths each read to the nearest centimeter and the screen offers a product good to eight decimal places; divide almost anything by almost anything and a long tail of digits appears. Anyone trained in the sciences learns, early and permanently, to cut such a result back to the precision the inputs actually justify, because a computed answer cannot be more precise than the measurements it was built from. The surplus digits are not information. They are an artifact of the arithmetic. Someone without that training tends to write the number down as the machine delivered it, every digit intact, and each surplus digit is a claim about the world that no measurement ever supported.
This is why an average of a handful of whole-number ratings comes back as 4.3333333 out of 5, and why three people out of seven becomes 42.857 percent. The precision was manufactured by the division, not found in the data, and in the second case it is claimed from a sample so small that the real uncertainty runs to tens of points. In this series’ terms, each written digit is a stroke of the map, and a stroke of the map is a claim that the territory was surveyed that finely. To carry a number to more digits than the measurement warrants is to draw a coastline to the millimeter from a survey done by eye, and then to point at the fine detail as though it were knowledge. The tool tempts everyone toward this small dishonesty, and resisting it, cutting a figure back to the digits one has actually earned, is among the plainest disciplines the Conscious Look asks of anyone who reports a number.
Gödel and the limits of mathematics itself
The most honest thing that can be said about mathematics as a language is not only that it cannot say everything (that its precision requires abstraction from context and meaning) but that it cannot even fully describe itself. In 1931, the Austrian mathematician Kurt Gödel published a result that surprised and disturbed the mathematical community more than any result since the discovery of irrational numbers by the ancient Greeks: the incompleteness theorems.
Gödel proved, in two theorems of extraordinary technical depth, that any consistent formal system powerful enough to express basic arithmetic must contain true statements that cannot be proven within the system, and that the consistency of the system cannot be proven from within the system itself. The first theorem says that no matter how rich and powerful a formal mathematical system is, there will always be mathematical truths that the system cannot prove. The second says that no system can establish its own consistency (cannot prove that it will never produce a contradiction) by using only the resources available within it.⁹
The incompleteness theorems are not a crisis for mathematics. Mathematical practice continues largely undisturbed by them, because the statements that Gödel constructed to be unprovable are artificial constructions specifically designed to resist proof, rather than the ordinary statements of arithmetic and analysis that mathematicians actually care about. But the theorems are profoundly significant for the philosophy of mathematics, and they are directly relevant to this series’ project, for a specific reason. They show that mathematics (the most precise, most rigorous, most reliable language available to human beings) is itself a model with limits. It cannot fully describe its own foundations. It cannot prove its own consistency. It cannot capture all mathematical truth within any given formal system. Mathematics is not exempt from the general principle that this series has been documenting across every domain: every model has limits, every map leaves territory unmapped, and the acknowledgment of those limits is the beginning of intellectual honesty rather than the end of the inquiry.
The Wigner puzzle (the unreasonable effectiveness of mathematics) is therefore matched by a complementary puzzle that receives considerably less attention: the unreasonable incompleteness of mathematics. The language that describes the physical world with extraordinary precision cannot fully describe itself. The tool that generates the most reliable knowledge available to human beings contains truths that the tool cannot prove. This is not a paradox. It is the characteristic structure of powerful tools: they are most useful precisely because they are limited: because they have committed to a specific domain of operation and achieved extraordinary results within it, at the cost of leaving other domains outside their reach.
The Conscious Look, applied to mathematical language
The Conscious Look, applied to mathematics, requires maintaining two simultaneous recognitions that are easy to lose sight of in the presence of equations.
The first is that mathematical precision is genuinely valuable and genuinely different from the precision available in natural language. When a mathematical model makes a prediction, the prediction is specific, falsifiable, and comparable to observation in a way that a natural language claim typically is not. The precision of the mathematical claim is a real virtue, not merely a stylistic preference. It should be taken seriously, and the predictions of well-constructed mathematical models should be given genuine evidential weight.
The second is that the territory is always larger than the mathematical map. The phenomena that mathematics cannot represent (the subjective, the contextual, the meaningful, the tacit) are not less real for being unmappable in mathematical terms. They are simply outside the domain of the tool. And the precision of the tool’s outputs within its domain should not be confused with the completeness of its coverage of the territory. The economic model that generates precise predictions about aggregate market behavior is not, for that reason, a complete account of the economy. The epidemiological model that generates precise predictions about disease transmission rates is not a complete account of the epidemic, whose social, behavioral, and political dimensions require frameworks that the mathematical model does not include.
This is the application of article S1-102’s criterion of knowing the limits to the specific case of mathematical language. Mathematics knows its domain, and the best practitioners of mathematical modeling know it too. The confusion of mathematical precision with completeness of coverage is not primarily a failure of mathematicians. It is a failure of consumers of mathematical outputs: the policymakers, journalists, and citizens who receive the outputs of mathematical models and treat the precision of the number as evidence of the accuracy and completeness of the account. The number 6.7 percent knows what it is measuring. The question worth asking is whether what it is measuring is what one wants to know.
Further reading
Eugene Wigner’s “The Unreasonable Effectiveness of Mathematics in the Natural Sciences,” published in Communications on Pure and Applied Mathematics in 1960 and freely available online, is the original and still the most thought-provoking statement of the puzzle that motivates this article. It is readable by anyone and remains genuinely unresolved.
Roger Penrose’s The Road to Reality: A Complete Guide to the Laws of the Universe (2004) provides the most comprehensive available treatment of the relationship between mathematics and physical reality: including a serious engagement with Gödel’s theorems, the nature of mathematical truth, and the question of why the universe appears to be mathematical at its foundations. It is demanding (Penrose does not simplify the mathematics) but it is the most honest available treatment of the subject for a serious general reader.
Iain McGilchrist’s The Master and His Emissary (2009) provides the neurological and philosophical framework for understanding mathematics as a left-hemisphere tool: what it achieves through its commitment to precision and abstraction, and what it systematically misses. It is the essential complement to Wigner’s account.
Douglas Hofstadter’s Gödel, Escher, Bach: An Eternal Golden Braid (1979) is the most celebrated accessible treatment of Gödel’s incompleteness theorems: embedding them in a rich meditation on self-reference, consciousness, and formal systems. It is long, playful, and rewarding; the mathematical argument is precise, the surrounding material is generative.
For the specific failure mode of confusing numerical precision with measurement accuracy, Joel Best’s Damned Lies and Statistics: Untangling Numbers from the Media, Politicians, and Activists (2001) provides the most accessible treatment, with abundant examples from public discourse of exactly the error that this article describes.
Notes
¹ G. H. Hardy, A Mathematician’s Apology (Cambridge University Press, 1940), sections 22-23. His full statement: “I believe that mathematical reality lies outside us, that our function is to discover or observe it, and that the theorems which we prove, and which we describe grandiloquently as our creations, are simply our notes of our observations.” Hardy was a mathematical Platonist, as was Gödel, and as is Roger Penrose in a well-developed modern form (see The Road to Reality, cited above), for whom mathematical objects inhabit a real world distinct from both the physical and the mental.
² The anti-Platonist family is large and old. Formalism, associated with David Hilbert, treats mathematics as the manipulation of symbols under chosen rules; intuitionism, associated with L. E. J. Brouwer, treats mathematical objects as mental constructions that exist only once constructed. The fictionalist position is developed in Hartry Field, Science Without Numbers: A Defence of Nominalism (Princeton University Press, 1980), which argues that mathematical objects do not exist and that mathematics, though useful, is strictly a fiction. The embodied-cognition account is George Lakoff and Rafael Núñez, Where Mathematics Comes From: How the Embodied Mind Brings Mathematics into Being (Basic Books, 2000), which traces mathematical ideas to metaphors grounded in bodily experience.
³ For a careful and neutral survey of the competing positions, see “Platonism in the Philosophy of Mathematics,” Stanford Encyclopedia of Philosophy. The framing offered here, that the invented-or-discovered question is the map and territory distinction of this series in its sharpest form, is the article’s own interpretation, not a claim that either side has won; the philosophical question remains unresolved.
⁴ Maxwell’s equations were first formulated by James Clerk Maxwell in a much more cumbersome form in his 1865 paper “A Dynamical Theory of the Electromagnetic Field,” which ran to many pages and used a notation that required 20 scalar equations to express what the modern vector notation captures in four. The modern compact form using vector calculus was developed by Oliver Heaviside in the 1880s. The compression achieved by the notation (from 20 equations to four) without any loss of physical content is itself a demonstration of the point being made: the mathematical notation is doing work, not merely recording it. The four equations in their modern form imply, among other things, that electromagnetic waves travel at the speed of light, a result that Maxwell himself recognized as implying that light is an electromagnetic phenomenon, one of the most consequential deductions in the history of physics.
⁵ The claim that mathematics does not have paradigm shifts in the Kuhnian sense requires qualification. There are episodes in the history of mathematics that resemble paradigm shifts in some respects: the discovery of non-Euclidean geometries in the nineteenth century, which showed that Euclid’s parallel postulate was not a necessary truth but a contingent assumption that could be denied consistently; the discovery of irrational numbers by the Pythagoreans, which required the revision of a framework that had treated all quantities as expressible as ratios of integers; and the foundational crisis of the early twentieth century, in which the discovery of paradoxes in naive set theory required the reconstruction of mathematics on more rigorous axiomatic foundations. What these episodes do not involve, however, is the discovery that previously established results were wrong. They involve the discovery that the scope of previously established results was more limited than had been assumed, or that the concepts underlying them required more careful formulation. The results themselves (properly understood within their domain) remained valid.
⁶ McGilchrist, I. (2009). The Master and His Emissary: The Divided Brain and the Making of the Western World. Yale University Press. McGilchrist’s treatment of mathematics as a left-hemisphere tool is developed in the context of a broader argument about the consequences of left-hemisphere dominance in Western culture: the progressive privileging of what can be precisely formalized over what can be holistically understood. The argument is not that mathematics is wrong or bad but that its progressive extension to domains for which it was not designed (social life, artistic experience, moral judgment) produces characteristic distortions that are invisible from within the mathematical framework.
⁷ The behavioral economics literature documenting the systematic departures of actual human behavior from the predictions of the rational agent model is reviewed most accessibly in Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux, and in Thaler, R. H., and Sunstein, C. R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. The specific failure of the rational agent model (its treatment of what are systematic and predictable features of human behavior as anomalies rather than as the primary data that the model should explain) is the subject of article S1-404 of this series, in the context of the broader discussion of what mathematical models of social systems miss.
⁸ The construction of economic statistics (and the contestable definitional choices embedded in them) is examined in depth in Morten Jerven’s Poor Numbers: How We Are Misled by African Development Statistics and What to Do about It (2013), which documents the specific definitional choices that produce GDP figures for developing economies and the very large margins of error that those choices introduce. The broader point (that the precision of a statistical expression is a function of the mathematical operations performed on the data, not of the accuracy of the underlying measurement) applies to all statistical measures, including those from the most sophisticated national statistical agencies.
⁹ Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38, 173-198. Gödel’s paper is technically demanding and requires a background in mathematical logic. The most accessible book-length treatment for a general reader remains Hofstadter, D. R. (1979). Gödel, Escher, Bach: An Eternal Golden Braid. Basic Books. The specific philosophical interpretation of the incompleteness theorems (what they imply about the nature of mathematical truth, the limits of formal systems, and the relationship between provability and truth) is itself contested. The interpretation given in this article, that the theorems demonstrate that mathematics contains true statements that no given formal system can prove, follows the standard philosophical reading and is sufficient for the purposes of this article, though the more technically precise statement is that there are statements in any consistent formal system powerful enough to express arithmetic that are neither provable nor disprovable within the system, and that the system’s consistency is equivalent to one such statement.