Jevons's Other Machine

In the 1860s, William Stanley Jevons built a machine. Not the marginal utility theory that would make him one of the founders of modern economics. The other one: a logical piano, held today in an Oxford museum, that could perform certain logical operations faster than humans could.

The man who gave economics its most human-scaled idea also built one of the earliest devices for automating reasoning. Marginal utility says that value depends on how much of something you already have, on where you stand, on what the next unit means to you specifically. Both projects lived in the same mind. Jevons wrote that the mind "seems able to impress some of its highest attributes upon matter, and to create its own rival in the wheels and levers of an insensible machine." Tyler Cowen's new book, *The Marginal Revolution: Rise and Decline, and the Pending AI Revolution* (Mercatus Center, 2026), tells the story of the first project. But the second project is the one that won.

The revolution that destroys itself

Here is the shape of the argument. Marginalism is being displaced by machine learning, neural nets, and AI. Not because it is wrong. Because it is no longer necessary. A 2024 finance paper uses 360,000 factors to predict asset returns. No human can hold 360,000 factors in mind. No one pretends to try. Steve Levitt, one of the last great champions of price theory at Chicago, retired from academia and said it plainly: "I think in the marketplace for ideas, I gotta say that the Chicago price theory really has lost."

That is the elegy. But the book's deeper claim is structural. Jevons did not merely launch the marginal revolution; he also launched what Cowen calls the "average revolution," the drive toward statistical measurement and empirical method. Jevons wanted to systematize everything: economics, logic, science, the social world. He wanted to measure, quantify, formalize. That impulse gave economists something to do for a century. It also set in motion the chain of developments (better statistics, then econometrics, then computing, then machine learning) that would one day make the intuitive core of his revolution dispensable. The creator was also the destroyer. The logical piano was not a side project. It was a prophecy.

This is the book's best idea, and Cowen earns it through patient historical work. He traces the parallel between economics and other sciences that were strangely slow to develop: botany before Linnaeus, geology before Hutton, evolutionary biology before Darwin. In each case, the key insight was available long before it was absorbed. Galileo resolved the diamonds-water paradox in 1632; economics did not catch up for over two centuries. The Salamancan theologians stated marginal utility theory clearly. Richard Jennings anticipated Jevons in 1855. In geology, the evidence for an ancient earth was literally underfoot for millennia.

What held these fields back was not a lack of data or equipment. Cowen calls it "seeing around corners": certain ideas are extremely hard to discover, trivially easy to verify once stated. Economic reasoning, he thinks, is "fundamentally counterintuitive." But I think the difficulty is more precise than that. The insights are not hard to think. They are hard to inhabit. They require you to reorganize not just what you believe but how you stand in relation to what you already know, how you respond to it, what you notice and what you overlook. That kind of reorganization is never purely individual. It depends on being surrounded by others who are reorganizing in the same direction, who share your sense of what matters and can check you when you lose it. "The aspects of things that are most important for us are hidden because of their simplicity and familiarity," Wittgenstein wrote. The Salamancans could state marginal utility. They could not see why it mattered, because seeing why it mattered required a reorganization of intellectual life that had not yet occurred.

And what made the reorganization possible, in every case Cowen examines, was not genius alone. It was social infrastructure: professionalization, university chairs, independent income, networks of peers. Linnaeus needed Uppsala. Hutton needed financial independence from the church. Jevons needed Owens College. Darwin needed Lyell, who needed the Geological Survey. The insights were socially hard before they were intellectually hard. They required communities of people who could afford to reorganize their commitments, not just their calculations.

This matters for the AI question, and the book leaves it unfinished. If the breakthroughs of the past required social conditions, not just cognitive capacity, then what does it mean when the next breakthroughs are produced by systems that have no social conditions at all? A neural net does not need a university chair or financial independence from the church. It does not need to reorganize its commitments. It does not, in any recognizable sense, have commitments. The machine that replaces the marginalist is not a better marginalist. It is a different kind of thing entirely.

The elegy as demonstration

Here is what I find most striking about The Marginal Revolution: the book performs what it claims is dying.

His argument is that economic intuition is being replaced by machine pattern-recognition. His method is entirely intuitive. The analogies between economics and botany, geology, evolution: these are not the product of data analysis. They are the product of a mind that reads across fields and notices structural similarities. The examples of "intuitive marginalism" that open the book (why drivers in China sometimes kill the pedestrians they hit, why the homeless prefer expensive cities, why marriages collapse at the margin of approval) work because they ask the reader to occupy a situation from the inside and feel the logic of the margin as a living thing, not an equation.

None of this is what a 360,000-factor model does. The book's achievement is proof that the thing it mourns is not yet dead, at least not in the person mourning it. The elegy is also a demonstration.

Yet here is the question: does this matter? Maybe the book demonstrates only that Cowen personally remains good at something the field no longer needs. Maybe the analogies to botany and geology are engaging but unnecessary; maybe a sufficiently powerful model could have generated the same structural insight without reading Linnaeus or Hutton. Late in the book, Cowen raises exactly this possibility, and he does it with unusual honesty:

"Maybe our intuitions about the world, including the economic world, were never so strong in the first place. Maybe we put so much value on 'intuitive' results, in 20th century microeconomics, as a kind of cope and also security blanket, to make up for this deficiency."

Is it cope?

This is the most interesting sentence in the book.

Some of what we call intuition probably is cope. The history Cowen himself tells supports this. The early marginalists did not just discover truth; they also built an institution that employed them, gave them status, and organized their professional lives. Marginalism, Cowen notes, was "a full employment project for economists." The commitment to intuitive reasoning was never fully separable from the social rewards of being the kind of person who reasons intuitively.

And the cope reading goes further than professional self-interest. When Levitt mourns price theory, part of what he mourns is a world in which his particular skills were the ones that mattered. When Cowen interviews job candidates and finds they cannot do price theory on their feet, part of what he is testing is whether they belong to his tribe. The Kevin Murphy summer camp at Chicago, which Cowen describes with obvious affection, is a rearguard action: a week-long ritual to transmit a dying craft to young initiates. Blacksmiths grieved too.

So: cope, or something more?

If intuition is a tool for arriving at correct answers, then it is cope. The machine arrives at better answers. The intuition was a workaround for limited computational power, and the workaround is no longer needed. The security blanket reading is correct, and we should fold up the blanket and move on.

But there is another possibility. Maybe some of what we call intuition is not a tool for reaching conclusions but a way of inhabiting the world that conclusions are about. When Jevons walked through the poor districts of London, he was not collecting data. He was reorganizing his relationship to poverty, putting himself in a position where the social implications of marginal utility theory could become real to him, not merely true. When Cowen constructs an example about why a husband gives up trying to please his wife at the margin, the example works because it asks you to feel the logic from the inside, to recognize something about human motivation that you already knew but had not organized. The insight is available before the example. What the example does is make you inhabit it.

If that kind of inhabiting matters, then it is not cope. It is the difference between a system that can tell you the price of everything and one that knows what things cost.

The Jevons Paradox, reversed

The cope question is about individuals. But there is a structural version that the book documents without naming, and it is harder to dismiss.

Cowen discusses the Jevons Paradox: making energy use more efficient increases total energy consumption. He applies it to AI (cheaper chips, more demand). But the version that matters here runs in the other direction. As the tools for understanding the economy become more powerful, the number of people who understand the economy in the inhabitable sense shrinks. More gets done. Fewer people know what is being done or why it matters. The field becomes more productive and less populated.

This is not a complaint about jobs. It is a question about the relationship between a civilization and its own knowledge. Marginalism, as Cowen tells the story, was not just an intellectual framework. It was something closer to a shared form of life: a community that organized its attention in a particular way, that trained its members to see the world through examples and analogies and biographical empathy, that maintained shared senses of what counts as elegant, what counts as surprising, what a good explanation feels like from the inside. The Kevin Murphy summer camp is not a curriculum. It is an initiation into a way of responding to the world together.

That community is now being replaced by systems that produce better predictions without needing any of that. The predictions are better. The community dissolves. And with it dissolves something that no individual can recover alone: the shared attunement that made inhabiting possible in the first place. Jevons did not learn to care about poverty by himself. He learned it in a world where caring about poverty in a particular way, through the lens of marginal analysis, was something a community of people was learning to do together. When the community goes, the caring does not simply transfer to the machines that replaced it. It disappears, the way a language disappears when its last speakers die, even if every word has been recorded.

Jevons would have appreciated the irony. He built the logical piano and he walked through the slums. He wanted to mechanize inference and he wanted to feel the weight of poverty. His mind held both projects without contradiction, or at least without acknowledging the contradiction.

The question the book opens, without quite arriving at, is whether those two projects can keep coexisting. A civilization can outsource the production of knowledge to systems that do not need to see around corners. But can it keep alive the capacity for the kind of seeing that produced the knowledge in the first place? Or does that capacity atrophy once it is no longer economically necessary, the way a muscle atrophies when a machine does the lifting?

I do not know. But I notice that this book works, and that it works because of something specific: the residue of a lifetime of reading across fields, the willingness to feel the weight of the examples, the biographical empathy that lets Cowen see Jevons not just as a contributor but as a person. These are the qualities of a mind that has spent decades inhabiting its subject. No machine currently produces them. Whether that will still be true in five years is a question the book, for all its honesty, does not quite dare to ask.

---

Sources & further reading

Tyler Cowen, [*The Marginal Revolution: Rise and Decline, and the Pending AI Revolution*](https://tylercowen.com/marginal-revolution-generative-book/) (Mercatus Center, 2026). A history of marginalism from Jevons to machine learning, told with exactly the kind of cross-disciplinary intuition it argues is disappearing.

William Stanley Jevons, "On the Mechanical Performance of Logical Inference" (*Philosophical Transactions of the Royal Society of London*, 1870). Jevons's presentation of the logical piano, including his observation that the mind can "create its own rival" in machinery.

Ludwig Wittgenstein, *Philosophical Investigations* (1953). "The aspects of things that are most important for us are hidden because of their simplicity and familiarity" (§129).

Jônadas Techio, ["The Claim Upon the Training Data"](/writing/essays/the-claim-upon-the-training-data) (2026). The first essay in this series, on Cowen's project of writing for AI readers, the fiction layer, and the distinction between knowledge and acknowledgment. The present essay extends that argument: what "The Claim Upon the Training Data" calls attunement and form of life, this essay traces in the specific dissolution of the economist's community of judgment.