Key Takeaways
Look beneath the interface: mathematics is modern life’s hidden infrastructure
The useful part is often invisible. A supermarket checkout, a medical scanner, an online payment, and a satellite navigation system may feel like separate technologies. Underneath, each relies on mathematical ideas and algorithms. Computers have not made mathematics obsolete. They have made calculations cheap enough to apply methods that once seemed impractical, while hiding the work from the people who benefit from it.
Surprise is the point. Mathematics is reasonably effective when it solves the kind of problem it was built for. Its more startling usefulness comes when abstract work migrates elsewhere: number theory helps secure Internet transactions, and topology helps organize kidney transplants. Stewart traces this reach to mathematics’ reality, beauty, generality, portability, unity, and diversity. Together, these qualities let a method developed in one setting become useful in another.
What’s striking is that mathematical infrastructure becomes most visible when it fails. A bad route, a faulty scan, or an insecure payment exposes the hidden system that normally disappears into convenience. This suggests a civic as well as technical lesson: public understanding need not mean everyone mastering advanced equations. It can mean recognizing that important decisions depend on models, assumptions, and people trained to scrutinize them. The book’s claim is strongest when it shows specific transfers between fields. The broad idea that mathematics supports modern life is persuasive, but the examples make it concrete.
Let abstract structures travel beyond the problems that first inspired them
Euler made a map of connections. In 1736, Leonhard Euler answered whether a walker could cross each of Königsberg’s seven bridges exactly once. He reduced the city to regions and bridges, represented as dots and lines, then showed that the bridge arrangement had too many regions connected to an odd number of bridges for such a walk to exist. This launched graph theory, the study of networks of connected points.
The same structure saves lives. In kidney exchanges, each donor-recipient pair becomes a point, and an arrow links a donor to a compatible recipient. Algorithms search the network for non-overlapping swaps and donation chains. By the end of 2017, these methods had identified 1,278 potential transplants in the UK, of which 760 were carried out. Euler was studying a puzzle, not medicine. The reusable idea was the structure of the problem.
The leap from bridges to kidneys is more than a charming historical coincidence. It shows how abstraction can preserve what matters while discarding distracting details. For transplantation, the city’s geography is irrelevant; the pattern of who can connect to whom is decisive. Yet the chosen mathematical representation is not the whole ethical problem. The algorithm can optimize only the priorities supplied to it, such as expected benefit or reliability. Good modeling therefore requires both mathematical skill and explicit judgment about whose needs count.
Judge political rules by their outcomes, not their democratic labels
Boundaries shape representation. Gerrymandering manipulates electoral districts through packing, concentrating opposing voters in a few districts, or cracking, dividing them across many districts so they lose repeatedly. In 2016, Hillary Clinton won 2.87 million more popular votes than Donald Trump, yet lost the US presidency under the Electoral College. District maps create a different problem, but both cases show that one-person-one-vote does not guarantee equal political influence.
No single fairness test suffices. Compactness scores, seat-to-vote comparisons, and measures of wasted votes can flag suspicious maps, but geography and voter distribution can produce unfair-looking outcomes without deliberate manipulation. Mathematical protocols offer another route: opposing parties can take turns dividing and freezing districts, using their rivalry to constrain each other. Even so, results about voting systems show that desirable fairness conditions can conflict. A system needs transparent rules and openly stated trade-offs.
The important distinction is between measuring a map and deciding what fairness means. A compact district is not automatically representative, and proportionality may conflict with other commitments in a political system. Mathematics can expose consequences that rhetoric obscures, but it cannot settle every value dispute by itself. That limitation is a strength of the analysis, not a defect: a clear metric makes disagreement more precise. The public can then debate the standard being used instead of pretending that a complicated rule is neutral simply because it is familiar.
Treat route planning as optimization, not a sequence of nearest guesses
The nearest stop can be a trap. The Travelling Salesperson Problem asks for the shortest tour that visits each location once and returns to the start. Choosing the closest unvisited stop at every turn sounds sensible, but can produce a longer route than a less obvious ordering. The same problem appears in delivery schedules, telescope pointing, DNA sequencing, and the movements of computer-controlled sewing machines.
Hard does not mean hopeless. The problem is NP-hard, meaning no fast exact method is known for every possible case, and mathematicians suspect none exists. Yet sophisticated algorithms have found exact tours through tens of thousands of real locations. When a near-best route is enough, a space-filling curve, a path that passes through every point in a region, can provide a route no more than about 25% longer than optimal. Pigeons in experiments also planned efficient paths among two or three feeders, though that is not a solution to large routing problems.
This example separates theoretical difficulty from everyday usefulness. A problem can resist a universal guarantee and still be tractable for the cases people actually face. That distinction matters in public debates about algorithms, where “too hard to solve” can be mistaken for “nothing useful can be done.” Approximation is not failure when the cost of finding perfection is excessive. At the same time, a route that is shortest by distance may not be best by fuel, safety, or fairness. Optimization always inherits the goal it is asked to optimize.
Pure number theory became the lock on global digital exchange
A hard calculation guards the key. RSA encryption turns a message into a number and transforms it using a public rule. The rule depends on multiplying two large prime numbers, which is easy; reversing that multiplication to recover the primes is believed to be difficult. Fermat’s and Euler’s results about arithmetic with remainders help make the transformation reversible for someone holding the secret key. This gives online systems a way to exchange keys and verify digital signatures.
The lock has assumptions. The difficulty of factoring large numbers has not been proved, and careless implementation can create other vulnerabilities. A sufficiently capable quantum computer running Shor’s algorithm could factor numbers fast enough to threaten RSA. That possibility has prompted research into post-quantum cryptography, including methods based on hard problems involving codes, lattices, and special graphs. Number theory, once considered unlikely to have practical uses, now protects ordinary digital commerce.
The case illustrates a useful distinction between a mathematical proof and a security assumption. RSA’s logic can be carefully derived, while its real-world safety still depends on a problem believed to be computationally hard and on sound implementation. That does not make the system unreliable by definition; it makes its basis inspectable and its limits important. The broader lesson is that cryptography is not a permanent lock. Security is a negotiated lead over available computing power, so systems must be designed with replacement in mind.
Imaginary numbers make rotations, signals, and quantum states calculable
A strange number becomes a useful coordinate. The imaginary unit, written i, satisfies i squared equals negative one. Pairing real numbers with multiples of i creates complex numbers, which can represent points on a plane. Multiplying by i rotates a point by a right angle. Engineers use complex numbers to simplify calculations for alternating electrical currents, where amplitude and phase travel together in one compact representation. Quantum mechanics goes further: its basic wave functions require complex values, not merely real numbers.
Three dimensions needed a different tool. William Rowan Hamilton searched for a three-part number system to handle spatial rotations, but no such system could preserve all familiar arithmetic rules. His quaternions use four components and give up commutative multiplication. They proved well suited to representing rotations in computer animation, where spaceships and cameras must move smoothly without accumulating distortions. A mathematical system can be valuable even when its components do not resemble everyday measurements.
These examples challenge the instinct to judge a concept by whether it looks physically familiar. Complex numbers and quaternions work as representations: they make certain relationships easier to calculate while preserving the features that matter. Their success does not require treating them as mysterious substances. It does require care about what a model represents. A convenient coordinate system can simplify a calculation, but the physical meaning belongs to the application and its interpretation, not to the symbols alone.
Transforms reveal hidden structure in scans, photographs, and fingerprints
Changing representation changes what is visible. Joseph Fourier developed tools for describing heat flow through combinations of simple waves. The same general idea lets engineers analyze signals by their frequencies. A related method, the Radon transform, represents an image as projections from many directions. Its inverse reconstructs the original distribution, enabling CT scanners to infer internal structures from X-rays. The medical application arrived decades after the mathematics.
Compression uses selective attention. JPEG transforms blocks of pixels into patterns, then reduces information that human vision is less sensitive to. Wavelets, localized wave-shaped patterns, can preserve important image structure while removing fine detail. The FBI adopted a wavelet method for fingerprint records, achieving compression of at least 15 to 1. These methods do not all do the same job: tomography reconstructs, while compression reduces data. Each changes representation to make a particular task manageable.
The common thread is not a single formula but a strategy: recode information so the desired feature becomes easier to detect or store. That strategy appears throughout science, computing, and measurement. It also makes the losses visible. A compressed holiday photo can tolerate detail being discarded; a fingerprint record cannot casually lose the features used for identification. Every transform therefore embeds a judgment about which information matters, and those judgments should match the consequences of error.
GPS works because geometry, random-looking codes, and relativity cooperate
Your location is inferred from timing. Each GPS satellite sends a precisely timed signal. By comparing its arrival with a locally generated reference, a receiver estimates the satellite’s distance. Each distance defines a sphere around a satellite; the intersections of several spheres narrow down the receiver’s position. Because measurements contain errors, the receiver uses extra satellites and least-squares calculations to find the best estimate.
Physics keeps the clock honest. Satellites carry atomic clocks and transmit carefully designed pseudorandom codes, which help receivers identify signals and measure delays. But the clocks run at different rates from clocks on Earth. Special relativity slows satellite clocks by about 7 microseconds per day, while general relativity speeds them up by about 45. The net difference is about 38 microseconds daily. Without correction, the position error would grow to roughly 10 kilometres per day. Navigation is a working alliance of mathematics and physics.
GPS is a striking example of a system whose accuracy depends on coordinating several kinds of knowledge. Geometry can locate a receiver only if timing is trustworthy; timing is useful only if the signal can be identified; and both depend on correcting for physical effects. The system’s apparent simplicity is an achievement of engineering, not evidence that the underlying problem is simple. It also shows why useful technology can test theory: the relativistic correction is not an optional philosophical flourish when the clocks are part of a navigation system.
Use sequence and shape, not averages alone, to diagnose industrial data
Order can carry the signal. Spring manufacturers could not reliably tell good wire from bad using standard tests. Ian Stewart’s colleague Len Reynolds tried a method for reconstructing patterns from time series: successive gaps between coils became points in a cloud. Good wire produced a compact, roughly circular cloud; uncoilable wire produced a long, thin one. Ordinary averages and variance missed the difference because they ignore the order of measurements. The resulting test replaced days of trial coiling with a more direct assessment.
Shape can reveal more than a picture. A related mathematical tool, persistent homology, tracks which holes in a data cloud remain as the viewing scale changes. It has been used to assess gaps in security-sensor coverage and to detect patterns in industrial and medical data. Both approaches turn data structure into evidence. Neither guarantees meaning on its own: the shape must be tested against the real outcome it is meant to predict.
The spring-wire case captures a practical virtue of modeling: a useful diagnostic need not explain every physical cause. It can be enough to identify a pattern that reliably distinguishes acceptable material from unusable material. That is powerful, but it also places a burden on validation. A cloud shape can correlate with quality in a sample and fail under new conditions. Persistent methods add another caution, since perceived structure depends on scale. Robust analysis asks which features survive changes in measurement, not just which appear in one convenient plot.
An electron-spin model can explain the fractal spread of Arctic melt ponds
A familiar model crossed fields. The Ising model was designed to study how neighboring electron spins create magnetism, and how heating can erase it. On a grid, each site has one of two states, and the model calculates how patterns change with temperature. Near a critical temperature, the arrangement develops structures at many scales rather than one typical patch size.
Sea ice shares a mathematical pattern. Arctic melt ponds also grow and merge into intricate shapes. Researcher Kenneth Golden noticed their resemblance to simulated Ising patterns and adapted the model, replacing electron states with frozen or melted patches of ice. Observations showed pond sizes following a power law, and measured boundary patterns matched important features of the model. The approach is still developing and does not capture every physical process, but it offers a way to connect small-scale pond geometry to how quickly ice melts.
The analogy is valuable because it turns visual resemblance into a testable proposal. Similar patterns alone do not prove that two systems share a mechanism; the comparison becomes stronger when measured statistics also agree. This is a productive middle ground between dismissing a cross-disciplinary hunch and accepting it on appearance. The model’s limits matter too. A sea-ice model must eventually incorporate sunlight, wind, and melting physics, so mathematical portability is a starting point for explanation, not a substitute for observation.
Analysis
What’s the Use? is a tour through applications, but its deeper subject is how knowledge moves. Stewart’s strongest examples follow a recurring path: someone strips a problem down to its relations, builds a general mathematical object, and later someone else recognizes that object in a new setting. The bridge puzzle becomes a network; a network becomes a transplant exchange. A theory of electron spins becomes a model for melting ponds. Mathematics earns its reach not by anticipating every application, but by making structures reusable.
This account also complicates the popular divide between pure and applied research. The distinction concerns motivation more than eventual consequence. An abstract result can be commercially or medically consequential, while a practical model may generate theory that outlives its original purpose. The book therefore makes a persuasive case against judging research solely by immediate utility. Predicting which ideas will travel is difficult precisely because transfer depends on later inventions, institutions, and problems that may not yet exist. A necessary qualification is that mathematical clarity does not equal social neutrality. A voting metric reflects a decision about fairness; a medical optimization reflects a ranking of benefits; a compression method decides what information may be discarded. Formalization exposes consequences and makes reasoning auditable, but humans still choose objectives, acceptable errors, and whose outcomes matter. This is not an argument against mathematics. It is a reason to pair technical transparency with public scrutiny.
The book’s central achievement is to make hidden dependence legible without reducing mathematics to a bag of tricks. Its examples show a discipline that is simultaneously cumulative and surprising. The same capacity for abstraction supports communication, medicine, navigation, and political analysis, yet applications can be fallible, inequitable, or dangerous. Mathematics is not magic, and it is not automatically wisdom. It is a powerful way to represent relationships, derive consequences, and test whether a proposed solution fits the world.
Review Summary
What's the Use? receives mixed reviews, with an average rating of 3.58/5. Many praise Stewart's ability to explain complex math applications in everyday life, covering topics like gerrymandering, medical technology, and climate change. Readers appreciate the humor and historical context. However, some find it too technical or dense for the average reader. Critics note that Stewart sometimes dismisses opposing viewpoints and fails to fully explore math's role in daily decision-making. Overall, it's recommended for those interested in math's practical applications, but may be challenging for non-mathematicians.
People Also Read
Glossary
Unreasonable effectiveness
Unexpected cross-field usefulnessEugene Wigner’s phrase for mathematics’ surprising success in areas far removed from the problems that originally motivated its development. Stewart uses it to describe transfers such as number theory into Internet security, or a model of magnetism into research on melting sea ice. Its central puzzle is why abstract structures so often fit new real-world problems.
Graph theory
Mathematics of connected networksThe study of graphs, which consist of vertices, also called nodes, joined by edges. A graph preserves information about which objects connect, while ignoring details such as their physical shape or position. Euler’s analysis of the Königsberg bridge puzzle helped establish the subject. Graphs now represent problems including kidney exchanges and route planning.
Travelling Salesperson Problem
Shortest visit-every-stop tourThe optimization problem of finding the shortest route that visits each of a set of locations once and returns to the starting point. It models far more than travel, including machine movement and delivery schedules. The problem is NP-hard, so no fast exact method is known for all cases, although practical algorithms solve many large instances.
RSA cryptosystem
Public-key encryption methodA public-key cryptographic system whose security relies on the presumed difficulty of factoring a large number formed by multiplying two secret primes. The public key can encode messages, while a private key enables decoding. RSA is used to exchange keys and verify digital signatures. Its security depends on mathematical assumptions and careful implementation.
Quaternions
Four-part rotation mathematicsA number-like system with four real components, conventionally written using a real part and three additional components. Quaternion multiplication is not commutative, meaning the order of multiplication can change the result. Quaternions provide an efficient way to represent and combine rotations in three-dimensional space, including the movements used in computer graphics.
Fourier transform
Signal-to-frequency conversionA mathematical transformation that represents a signal as a combination of component frequencies. The inverse transformation reconstructs the original signal. Fourier’s work grew from the study of heat flow, but related methods now help analyze audio, images, and other data. Changing to the frequency representation can make patterns easier to identify or manipulate.
Radon transform
Image projections and reconstructionA transformation that records projections of an image from different directions. Its inverse can reconstruct the underlying image from those projections, provided sufficient directional information is available. The method is mathematically related to Fourier analysis and supplies a theoretical foundation for tomography, including the reconstruction of internal body structures in CT scanning.
Wavelets
Localized, scale-sensitive patternsMathematical functions shaped like localized waves or blips. A signal or image can be represented as combinations of wavelets at different positions and scales. Unlike the indefinitely extended sine waves used in Fourier analysis, wavelets can focus on local features. This makes them useful for analyzing data and compressing images while discarding selected fine detail.
Persistent homology
Tracks holes across scalesA computational form of topological analysis that studies how features such as connected components and holes appear and disappear as the scale of observation changes. Features that persist across a broad range of scales are treated as more significant than fleeting ones. The method is used to analyze complex data, including sensor coverage and medical or industrial measurements.
Ising model
Model of interacting two-state sitesA mathematical model of a grid of sites, each in one of two states, that interact with neighboring sites. In its original physical application, the states represent electron spins and the model helps describe magnetic phase transitions as temperature changes. Because its mathematical structure can be reused, variants have also been applied to the formation of melt ponds on Arctic sea ice.
Fuzzy set theory
Degrees of set membershipA mathematical framework in which an item can belong to a set to a degree between zero and one, rather than being only fully included or excluded. This degree is not a probability that membership will occur; it represents how fully the item belongs under the chosen definition. Fuzzy methods can model imprecise measurements and support control systems.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.