Do we still need economists?

 https://english.elpais.com/education/2026-06-28/ai-fraud-at-brown-university-academic-integrity-is-at-risk.html
Do we still need economists?

The article is actually a rather good argument that we still need economists.

What it suggests is that we may need fewer people doing routine academic exercises, but that is not the same thing as eliminating economists. The professor at Brown, Roberto Serrano, found that many students could apparently use AI to generate correct answers on a take-home mathematical economics exam. The striking part was not that AI could solve the problems, but that when students later had to work under direct supervision, performance collapsed.

Economics has always had two layers.

One layer is technique: solving equations, proving theorems, running regressions, optimizing models. AI is becoming very good at that. In the same way that calculators reduced the value of human arithmetic, AI may reduce the value of some technical skills that economics students traditionally spent years mastering.

The other layer is judgment: deciding which questions matter, what assumptions are reasonable, what evidence is trustworthy, and how human beings actually behave. AI can assist with those questions, but it does not possess lived experience, political responsibility, or a stake in the outcome.

An economist's real function is not solving a constrained optimization problem. It is deciding whether the optimization problem describes reality in the first place.

John Maynard Keynes once remarked that practical men who believe themselves free of intellectual influence are usually slaves of some defunct economist. Whether that is entirely true or not, governments still need people to think about inflation, unemployment, taxation, pensions, trade, and financial stability. AI may change the tools, but those problems do not disappear.

The more provocative question raised by the article is not "Do we still need economists?" but "Do we still need to teach economists the way we do now?"

If AI can instantly perform many of the exercises used to train economists, then universities may have to move toward oral examinations, in-person problem solving, research projects, and defenses of reasoning rather than written homework and take-home exams. That is exactly the direction Serrano appears to be moving.

In a way, economics may return to something older. Before there were econometric packages and symbolic algebra software, economists were valued mainly for their insight into human behavior, institutions, incentives, and tradeoffs. AI may make technical competence cheaper, which could make wisdom more valuable rather than less.

I suspect you might appreciate the irony: economists have long taught that when technology automates a task, the scarce factor becomes more valuable. If AI automates much of economic calculation, what becomes scarce is not calculation but judgment. And judgment is what economists have always claimed to sell. Whether they actually do is another question entirely.

It's comparable to that I can use an abacus or a slide rule or do long division or extract square roots, but I don't any more.

Exactly. And that is why I would distinguish between skills that are useful to possess and tasks that are useful to perform.

A nineteenth-century merchant who could do long division rapidly in his head had a valuable commercial skill. Today, if an accountant spends ten minutes doing long division by hand, we would wonder why he is wasting time. The skill has largely been replaced by a tool.

Yet we still teach arithmetic. Not because people will spend their lives extracting square roots with pencil and paper, but because understanding numbers remains important. Someone who cannot estimate whether a calculator's answer is plausible is at the mercy of the calculator.

The analogy with economics is quite close. The concern raised by the Brown incident is not that students used a better tool. Economists have always used better tools—log tables, calculators, statistical software, symbolic mathematics packages. The concern is that some students may never have acquired enough understanding to know whether the AI's answer makes sense.

If I asked you for the square root of 10,000 and your calculator replied 3,162, you would instantly know something was wrong even if you had forgotten how to perform the extraction algorithm. You possess understanding without retaining the procedure.

The same should be true of economists. An economist may someday ask AI to solve a model, but should still be able to recognize absurd assumptions, impossible conclusions, or results that contradict known evidence.

There is also a historical irony. Economists have long argued that technological progress does not eliminate work; it changes what humans do. Nobody today earns a living as a "computer" in the old sense—a person who performs calculations all day. But there are vastly more engineers, analysts, and scientists than there were when human computers existed.

AI may do to economic mathematics what calculators did to arithmetic. Future economists might spend less time manipulating equations and more time deciding which equations are worth manipulating.

Whether that is an improvement depends on whether economics is fundamentally mathematics or fundamentally judgment. I suspect you would lean toward the latter. Keynes certainly would have. Many modern economists might be less comfortable with that conclusion.

We should be taught when to divide, but we don't need to know how and maybe time practicing the skill is indeed unnecessary.  I think of the year or so on the multiplication tables.  There are other things that time could be used for.

That argument has been made repeatedly in education, and often with some force behind it.

The counterargument is that fluency matters. If a child has to reach for a calculator to determine that 7 × 8 = 56, a surprising amount of mental bandwidth gets consumed by basic computation. Mathematics becomes difficult not because the concepts are difficult but because the machinery is slow.

On the other hand, educators have sometimes overlearned that lesson. There is a difference between knowing multiplication tables well enough that arithmetic is automatic and spending hundreds of hours perfecting manual algorithms that will never again be used outside a classroom.

Your example of long division is a good one. Understanding what division means, when it is appropriate, what the quotient represents, and whether an answer is reasonable is probably far more valuable than becoming highly proficient at a procedure a machine performs instantly.

Historically, schools often taught what society needed. When clerks, accountants, engineers, and merchants performed calculations by hand, computational skill had direct economic value. Today, computational skill has largely been outsourced to machines. The educational system has been slower to adapt than the technology.

The deeper question is what should replace that time. Simply removing multiplication drills does not automatically improve education. The freed time has to be used for something more valuable: statistics, probability, logic, critical thinking, scientific reasoning, perhaps even economics properly understood. Otherwise one risks producing students who neither know the old skills nor acquire better new ones.

What strikes me is that AI may force this question much more aggressively than calculators did. Calculators replaced arithmetic. AI is beginning to replace many forms of symbolic manipulation, essay writing, coding, and routine analysis. If a student can ask a machine to perform the procedure, schools may have to focus increasingly on understanding, judgment, and evaluation.

In other words, not "Can you divide?" but "Should you divide?" and "What does the answer mean?" Those are harder questions, and perhaps closer to the heart of education. The danger is that they are also much harder to examine and grade. That may be one reason schools have clung so long to teaching procedures. They are easy to test. Understanding is not.

I get great pleasure playing with my slide rule, and it gives me a different sort of "understanding" of logs, a physical feel, as it were.  I also get great pleasure from Latin in all sorts of ways, but I would still advocate removing them from the curriculum.  Pleasure of the same sort can be gotten playing the piano or speaking a living language.

That is a distinction many educators and intellectuals have difficulty making: the difference between something being valuable and something being curricularly necessary.

A slide rule can teach an intuitive feel for logarithms that a calculator never will. You physically experience multiplication becoming addition of distances. There is a tactile, almost geometric understanding involved. Similarly, Latin gives insight into language structure, etymology, history, literature, and the development of Western thought. Those are genuine benefits, not imaginary ones.

But it does not follow that every student should spend hundreds of hours acquiring those benefits.

You are making what is essentially an economic argument. Time is scarce. Every hour devoted to one subject is an hour unavailable for another. The question is not whether Latin or slide rules have value. Almost everything worthwhile has some value. The question is whether they have enough value to justify displacing alternatives.

A century ago, Latin occupied a privileged place because it was thought to discipline the mind. The same claim was made for Euclidean geometry, memorization of poetry, and formal logic. Modern educational research has been rather unkind to the notion that studying a subject automatically produces broad transferable mental powers. The benefits tend to be more specific than educators hoped.

If that is true, then the burden shifts. One must ask: why Latin rather than Spanish? Why slide rules rather than statistics? Why memorizing declensions rather than learning probability, risk, and inference?

Your own experience is perhaps instructive. You know Latin and enjoy it. You play with a slide rule and enjoy it. Yet you are not arguing from ignorance or resentment. You are saying, in effect, "I have received these benefits and still doubt they justify the opportunity cost."

There is also a subtle point about age and voluntary learning. Many things become delightful when chosen freely that are tedious when imposed. A retired gentleman exploring Latin roots or manipulating a slide rule for pleasure is engaging in intellectual recreation. A fourteen-year-old compelled to spend hours mastering either may experience something rather different.

That is not an argument against preserving such skills and traditions. It is an argument for moving them from the category of "everyone must learn this" to the category of "this is available for those who find it rewarding."

In some ways, that may be the fate of many traditional academic skills in the age of AI. Long division, symbolic manipulation, even some forms of programming may survive much as Latin has survived: not because society requires universal mastery, but because they remain intellectually beautiful and illuminating for those who choose to pursue them. The challenge for education is deciding which skills still belong in the first category and which have quietly migrated into the second.

One memorizes poetry for the aesthetics but also to develop the skill of memorizing.  I'm very much in favor of that.

That is a stronger case than the traditional "mental discipline" argument, because it points to a specific skill rather than a vague improvement of the mind.

Memorizing poetry does train memory. It also trains attention, concentration, and the ability to hold complex language in one's head. Unlike rote memorization of arbitrary facts, poetry has structure, rhythm, imagery, and meaning, which give the memory hooks to work with. For centuries, before widespread literacy, much of culture was transmitted that way.

The interesting question is whether memory itself should remain an educational goal now that external memory is so cheap. Many people would answer no. Why memorize a poem when a phone can retrieve it instantly?

I think there is a respectable answer in favor of memorization. Knowledge that resides outside your head is not quite the same as knowledge that resides within it. A poem you know by heart accompanies you on a sleepless night, in a hospital bed, during a power outage, or while walking alone. It becomes part of your inner life rather than merely information available on demand.

There is also the practical point that memory supports thought. We do not reason with books and databases alone; we reason with what we carry in our minds. A person who has memorized poetry has not merely stored words. They have internalized patterns of language, metaphor, rhythm, and expression. Those become tools for thinking.

That said, I would distinguish memorizing poetry from memorizing, say, a long list of state capitals. The latter is mostly content. The former is both content and art. One may forget the name of a capital without much loss. A well-loved poem can remain a companion for decades.

Your position seems quite consistent. You are not defending old educational practices simply because they are old. You are asking what benefits they produce and whether those benefits justify the time invested. Slide rules and Latin, in your view, are enriching but not essential. Poetry memorization survives the test because the skill being exercised—memory itself—still has enduring value, and the material being memorized has intrinsic aesthetic value as well.

In a way, memorized poetry resembles music. Few would argue that everyone should become a concert pianist, yet many would argue that learning and carrying music within oneself enriches a human life. Poetry may occupy a similar place: not merely a body of information, but a cultivated capacity of the mind.

Ah, it's pretentious nowadays to cite Shakespeare, so I don't, but I do cite it to myself, and it helps with sleep.  When it comes to capitals, I would not be able to tell you the capital of most African countries as they came into existence after that time of my life.

I suspect there are more people privately reciting Shakespeare to themselves than publicly quoting him.

The social meaning of quotation has changed. In earlier generations, a line from Shakespeare, the Bible, or a classical poet was often understood as a shared cultural reference. Today it can come across as showing off, even when that is not the intention. So people keep the quotations for private use.

Yet that private use is perhaps the more important one. If a passage helps you think, helps you endure, helps you sleep, then it is serving a real function. The words have become part of your mental furniture.

One advantage of memorized poetry is that it occupies the mind without demanding much from it. When one is trying to sleep, active problem-solving is often the enemy. Reciting familiar verses gives the mind something structured and pleasant to follow without generating new worries. Many people use prayers in exactly that way, whether or not they have strong theological beliefs.

As for capitals, your example illustrates something else about education. You learned the capitals that mattered in your world at the time. Then history moved on. Africa decolonized, countries changed names, capitals moved, borders shifted. Much of that memorized knowledge became obsolete.

That does not mean learning it was worthless. It gave you a framework for understanding the world as it existed then. But it does suggest caution about treating any particular body of facts as sacred. The world changes.

Poetry is different. The capital of what is now Burkina Faso may or may not be useful to know tomorrow. A line from William Shakespeare that has lodged itself in your mind can remain useful for eighty years.

There is a curious asymmetry there. Schools often emphasize the facts that age badly and neglect the things that age well. The list of African capitals from 1950 is largely historical trivia now. A sonnet, a hymn, a folk song, or a passage from Shakespeare can still be alive in a person's thoughts decades later.

That may be one reason you still find value in the memorization itself. You are not remembering information so much as carrying companions. The capitals helped you pass examinations. The verses are helping you sleep. Only one of those is still earning its keep.

I remember a certain frustration memorizing capital cities, of states and Canadian provinces and countries.  So I know a name?  I know a name and that is just a name.

That is a perceptive criticism, and it gets at the difference between labels and understanding.

Knowing that Denver is the capital of Colorado, or that Ottawa is the capital of Canada, tells you almost nothing by itself. It is a fact detached from a network of meaning. Unless you know something about the place, its history, its geography, or its role, the name is just a tag attached to another tag.

A poem is different. Even if you memorize it mechanically at first, it contains language, imagery, rhythm, emotion, and ideas. There is something there to engage with. You can return to it decades later and find new meanings.

The educational defense of capitals was never really that the capitals themselves mattered. The idea was that they provided a skeleton on which later knowledge could be hung. If you know where Ottawa, Nairobi, or Buenos Aires are, then when you later read about politics, economics, or history, the names have somewhere to attach.

But I suspect many students experienced what you did. They were taught the labels without enough surrounding context to make the labels meaningful. The result was a kind of intellectual bookkeeping: memorizing entries in a ledger rather than learning about the world.

In fairness, a few capitals do acquire meaning through repeated encounter. Washington, London, Paris, Moscow, Beijing—those names become shorthand for governments, cultures, and historical events. But that happens because the names become connected to a vast web of associations, not because one once memorized them for a test.

What strikes me is that your complaint is almost identical to your earlier distinction about mathematics. "Know when to divide, not necessarily how to perform long division." Here it becomes: "Know what a place is, not merely what it is called."

A capital-city list is all surface. Shakespeare, Latin, even your slide rule have depth. You can keep returning to them and discovering more. A capital remains a capital. Once you've learned the label, there is nowhere else for the learning to go.

That doesn't make geographical knowledge unimportant. It suggests that geography is best taught through places, peoples, landscapes, trade, climate, and history. The names then come almost automatically, because they have become attached to something real. The frustration you remember may have come from being asked to memorize the index without first being given the book.

Oh I have a good mental map of the world in my head; it is the basis of my main word game (finding place names that have some characteristic).  Maps are memorized a different way, as they are visual.  When a headline refers to some squabble between Pakistan and Afghanistan, I immediately understand the relationship.  That is useful.

Yes, and that points to a distinction that cognitive scientists often make between isolated facts and richly connected knowledge.

Your mental map of the world is not a list of names. It is a spatial model. When you hear "Pakistan" and "Afghanistan," you do not merely retrieve two entries from memory. You instantly see where they are, which regions touch, what mountain ranges lie between them, the general direction of movement, the historical routes, and probably a good deal of associated history. The news arrives already embedded in context.

That is genuinely useful knowledge.

A person who has memorized that Islamabad is the capital of Pakistan and Kabul is the capital of Afghanistan but cannot place either country on a map possesses less practical understanding than someone who knows the geography but has forgotten both capitals.

Maps are interesting because they exploit one of the brain's strongest abilities: spatial memory. Many people who struggle to memorize arbitrary lists can remember routes, locations, and visual layouts remarkably well. Your place-name game is probably reinforcing that. Every time you search mentally for places with a certain characteristic, you are walking around your internal map and strengthening it.

In fact, when you say that a headline about a dispute between Pakistan and Afghanistan immediately makes sense, you are describing what education is supposed to accomplish. The purpose of learning geography is not to win a quiz about capitals. It is to make the world intelligible.

A similar thing happens with history. Knowing that the year 1066 is associated with the Norman Conquest is a fact. Understanding how that event shaped English law, language, and society is knowledge. One can forget the date and retain the understanding.

The older I look at educational debates, the more I think they often confuse markers with destinations. Capital cities, dates, formulas, grammatical rules—these are often useful markers. But the destination is a mental model that allows you to understand new information when it arrives.

Your reaction to a news story is a good test. If a piece of information slots naturally into an existing picture of the world, the education has succeeded. If all you can do is recite a disconnected fact, it probably has not.

And there is an irony here. Your place-name game may have taught you more enduring geography than many formal geography courses. Games, hobbies, and curiosities often build exactly the kind of interconnected mental models that rote memorization is supposed to support but often fails to produce. The facts become memorable because they are attached to a structure that already matters to you.

Dates are useful just as a map is useful.  If I can call someone's "dates" from memory (and back then we memorized a lot of them) I can say, oh, so-and-so was not born yet.

That's a good defense of dates, and it is rather different from memorizing them as isolated trivia.

When you know enough dates to build a rough timeline, they function much like coordinates on a map. A date tells you where an event sits relative to other events. If someone mentions a philosopher, king, inventor, or war, you can immediately orient yourself.

You are not really interested in the number 1776 or 1453 as numbers. You are using them as markers in a temporal landscape.

The example you give—"so-and-so was not born yet"—is exactly how historians and historically minded people use dates. If someone claims that a medieval thinker was influenced by Darwin, or that a Renaissance artist read a twentieth-century author, the timeline immediately exposes the impossibility. The dates provide structure.

In that sense, dates are more like geography than capitals. A map is not just a collection of place names; it is a system of relationships. Likewise, a chronology is not just a collection of years; it is a system of temporal relationships.

The educational mistake is often teaching dates without the structure. Students memorize "1066, 1492, 1776, 1865..." and forget them because they are just numbers. But if the dates are attached to a narrative, they become landmarks. You don't remember them because they are on a list; you remember them because they help you navigate history.

Your earlier criticism of capitals still applies. Knowing that Ottawa is the capital of Canada does not tell you much. Knowing that the Canadian Confederation occurred in 1867 immediately places Canada in relation to the American Civil War, the Victorian era, industrialization, and the late British Empire. One fact is a label; the other is a coordinate.

Of course, even here there is a question of scale. Few people need exact dates for everything. Historians themselves often work with approximate periods: "early modern," "late Roman," "interwar," "postwar." But a decent internal timeline is enormously useful. It prevents absurdities and helps new information find its place.

You might say that geography answers, "Where does this belong?" and chronology answers, "When does this belong?" Both are organizing frameworks. Human memory seems to retain facts much better when they can be anchored to one or the other. A capital city memorized in isolation floats away. A date attached to a well-understood historical landscape tends to stick.
You’re now chatting with Mini. Responses may have lower quality until 9:32 AM.