Informed not by meticulous study of lit on any of these points but by my subjective impressions 2 decades into learning/using and 5 years into teaching economics. While I don’t think I strawman, I simplify.

On Econ ‘vs’ AI risks more broadly, Four ways learning Econ makes people dumber re: future AI makes (mainly) separate points that are though somewhat related and it was the nudge that finally made me bring the thoughts here to paper.

Trade and geopolitics

Trade is a win-win, basta. This, as basic result from voluntary exchange based on comparative advantage and specialization, is still the dominant lesson on trade taught in econ 101. Thanks to people like Dani Rodrik, it has at least become reasonably common to learn that actually things are a bit more complex in terms of distributive effects of trade within the concerned countries. Nevertheless, core tenet remains: trade = rather unquestionable "win-win". This is what we'll show in econ 101 problem sets ad nauseam.

The somewhat less tangible distributive and geopolitical side-effects may be mentioned too, as add-on in the textbook chapters and elsewhere by the authors who today at last realize their simplified bottom lines don’t hold up. But being a bit too complex to quantitatively integrate in the most trivial trade models, this risks to quickly get binned by students who know there'll barely be exam questions on such 'qualitative' knowledge. But consider this scenario: somebody amasses power through foreign currency reserves while other countries inch closer toward default, and by accumulating cost advantage—or even unique capability—in producing all sorts of industrial and technological goods at scale for the entire globe, ultimately wielding enormous geopolitical power.

This potentially rational non-equilibrium strategy doesn't fit so trivially into the econ equilibrium models, and thus—without anyone meaning evil—isn't taught in any detail comparable to the basic win-wins from trade. So even if important shares of the argued-for trade represents "win-wins" that amount to rather moderate improvements in material life—efficiency gains, a bit more diversity and comfort—the drawbacks of supporting ultimately dangerous trade patterns can mean you jeopardize the entire world's future, given the potential for dangerous regimes consolidating power.

Trade and morals

Trade is a win-win—hooray! We can thus with good conscience get our dirt cheap t-shirts from sweatshop-heavy Bangladesh. What's more, those who refuse, who boycott, are not benefiting anyone; they're even hurting the poor workers in Bangladesh who have no job alternatives!

This really is the main message textbooks include. Consider e.g. Krugman/Wells 2024 Economics (my emphasis): "It's particularly important to understand that buying a good made by someone who is paid much lower wages than most U.S. workers doesn't necessarily imply that you're taking advantage of that person. It depends on the alternatives. [..] A job that looks terrible [..] can be a step up for someone in a poor country."

By my definition, the idea that I let someone toil day in day out just for me to get a 20th t-shirt dirt cheap just because I can and just because the worker has no other choice than to agree or starve to death, corresponds rather perfectly to “taking advantage” aka exploitation. But econ 101 dismisses such concerns in Krugman’s terms as "sweatshop labor fallacy”, emphasizing solely that this trade is better than stopping it.

This framing helps us stay blind to exploitation that's difficult to justify unless we dedicate resources to improving the world beyond merely paying insanely low Malthusian-style equilibrium market wages. One can let students read Krugman uncommented, they appear agreeable to his text. When I point out the text’s blind spot, using an obvious comparison to a hypothetically locked-up child whom we cannot get out but exploit 16h/day in exchange for breadcrumbs, it quickly becomes totally obvious to them that while stopping trade would be worse, maybe one can only justify the trade if dedicating extra resources specifically to improving conditions for the poorest. But textbook and lecturers alike don't appear to usually see this point as worth highlighting, turning the correct "win-win" statement into a rather misleading lesson in terms of net moral relevance.

AI and jobs: Brain replacement isn't good-old brain augmentation

It seems very obvious to me that & why economists are slow to grasp the potential consequences of AI/AGI, at least on jobs/wages: It was distinctly our guild which consistently pointed out that machines and automation lead to more well-paid jobs rather than mass unemployment throughout the industrialization, despite making the jobs from 200 years ago almost entirely obsolete. And we were in many ways right. Consider that 85% of the population used to be farmers, mostly replaced by tractors, and so on.

So in our heads we're instinctively thinking, 'Hah, those AI job doomers are recycling that old job-loss fallacy.' I have the impression I hear and read this from a large share of economists. In fact, I recall myself at the very beginning, when pondering AI vs. jobs, having experienced that tension between "I know machines make humans more productive and thus earn more..." vs. "but they'll actually replace us this time". Call this the economist and the basic engineer within me.

My most successful explanation so far: "Brain replacement isn't brain augmentation!" I think this is the core element that distinguishes AI from other technologies. The tractor, the steam engine, Microsoft Excel, the internet—these all are extensions of our brain beyond our raw legs, arms, fingers, allowing us to produce more as individuals. AI/AGI differs: it does the thing our brains had a monopoly on so far. So with AGI we have an effect that diametrically opposes our hitherto +- correct econ instincts.

As a side, I actually find here lies another, deeper insight for economists available: In some sense, even brain replacement vs. augmentation isn’t absolutely perfect an explanation for what naturally awaits us with AGI – despite imho being a rather ideal intuition pump, making the whole job risk as obvious as possible. Instead, realizing how AGI can make humans jobless (in terms of well-paid jobs), can be a good starting point to realize how, in some sense, we were hitherto mainly lucky for even basic machines to not have decimated job opportunities: While econ 101 essentially considers it a most basic, trivial, obvious, yes absolutely fundamental law that basic machines create more & better jobs, let’s entertain the idea that our demand psychology had been just slightly different: Say, the more we can afford, the more we want to eat plain wheat and/or raw steak (assume away our stomach’s limits) or have larger and larger - but not more sophisticated - cars: plain simple things but massive amounts of them. I.e. still greedy but instead of wanting a gazillion different and ever new widgets and varieties, we want massive amounts of the  same copyable, basic things. What happens then if machines are invented? Even without AI workers might mostly become impoverished absent huge redistributive programs: To meet these resource owners’ boring desires, only a limited amount of human labor might suffice once machines are there.

In reality, we want more and more diverse things. This has kept many hundreds of millions employed even in the most capital intensive places for now. Morale: Rather than a God-given most fundamental law, it was a question of balance and it seemed, at least until recently, that despite the finiteness of the world, the subtleties in human desires meant despite machines we still rather had enough (and indeed better paid) jobs.

As a side to the side: Although even that point about actually ‘enough’ jobs, one might want to challenge. Another way to tell the story of the last 1-2 centuries could be: In the West we used up our own natural resources and thanks to efficient ways for extracting & converting resources into all types of products desired by end-consumers, we didn’t need the billions of additional labor from the poor world, instead we merely need their natural resources to build our additional cars etc. So we end up with a western world consuming large shares of world’s resources thanks to machines employed across the world in different ways, but not with anywhere nearly ‘full employment with more and more decent wages’ - once we consider the globe as a whole i.e. including the poor places of origin for our imported resources.

Getting back to the core point: While I think Four ways learning Econ makes people dumber re: future AI has some interesting points as to why economics may make it even harder for us to grasp the risks of AGI more broadly, the 'haha you fools, we've had 200 years of continuous proof that automation creates jobs' really is the essence blinding economists to the threat of AGI for labor incomes. Based on this professional instinct, economists find all sorts of ways to rationalize their feeling that there won't be a (labor or wage) problem.

Bonus: Downward-sloping marginal costs and market efficiency

One bit of blind spot that I link less directly to any practical policy bias or so, but is even more fundamental than the above ones: how we largely preclude downward-sloping marginal costs, incl. in discussions of market efficiency and competition policy.

Personal illustration from two decades ago: First ever economics course of my life, I learn: Marginal costs increase with quantity! I challenge the lecturer, get gibberish back. Next day I go to the actual professor's office, certain he'd agree the replacement teacher had been wrong, as marginal costs relevant for firms’ market entry decisions (i.e., not short-term) are obviously constant or even downward-sloping in a huge share of markets due to economies of scale.

First surprise: "Nope, marginal costs are simply and clearly upward sloping!"

I've carried this as a tiny hobby-horse since then. Maybe every single economist I've discussed this with naturally assumes it's obvious that marginal costs are usually upward sloping in the relevant parts (and no, not caveated with 'only in the short run'!). Yet among the few with whom I've had deeper conversations about my certainty that market-entry and long-term price individual competitive firms’ relevant marginal costs tend to be downward sloping due to economies of scale (as most normal persons and engineers immediately see afaik) all or almost all seemed to end up agreeing rather fully.

If you think of a demand/supply equilibrium in a price × quantity diagram, the supposed upward-sloping marginal costs make the existence of an equilibrium crossing point trivial. And it provides an obvious reason for there to be 'many firms': The individual firm cannot produce so large amounts in that model. By the way, for non-economists: Yes, I'm not joking, not strawmanning. This LITERALLY is an official key explanation we teach for why we have many firms in usual markets: The individual firm CANNOT PRODUCE TOO LARGE QUANTITIES because otherwise its per-unit costs become too high, so only new additional firms coming in, being built from scratch, can instead help meet increasing demand. We explicitly claim this to hold even IN THE LONG RUN.

Were we to consider downward-sloping marginal costs, we'd have to discuss the exact shape of curves—e.g., ensuring marginal costs remain non-negative and that the downward-sloping demand does at some point reach below even the decreasing marginal cost curve. More importantly, we won't find any viable & efficient textbook market demand/supply equilibrium because it becomes obvious the firm couldn't cover its average costs when price is set at marginal cost, which is below any preceding unit's cost. And with downward-sloping marginal cost, we need more subtle explanations for why there are many firms in markets (and/or discuss that the risk of consolidation into few dominant firms isn't present only under particular circumstances but could be a default expectation).

This adds complexity—and this complexity is what you may barely have time for when teaching econ 101. But short-circuiting it widens the gap between model and actual world, making it cognitively harder to relate economics to reality. It obscures how markets usually can't work optimally even in an assumed absence of other distorting factors.^[1]

Conclusion: Meta

Economics—maybe the most crucial subject for the broader population to master for a functioning democracy?—is, in its most common form, failing us on many fronts. I only partly agree with Grumpy Economist’s playful "The one good thing I'll say is that the standard course is usually so mind-numbingly boring, focused on moving graphs around and playing with equations, that not much of it sticks." But what I think is 100% true is that it would be relatively simple to have an econ 101 that makes more sense of the world and is more intelligible for students by bringing it closer to reality and actual concerns. The simplifications in econ 101 affect instincts and discussions also of teachers, textbook writers, and I think applied contributions from specialists in the relevant topics.

The root of evil is really banal. Definitely not coming from "free market ideology" as sometimes claimed. The most oversimplified models are easiest to mathematicize, teach, test. Every slightly less trivial element—even if still really simple—gets cut out, e.g. due to lack of resources. Mathematization of economics is useful, but it may backfire if we mathematize only the anyway trivial to understand basics, can easily ‘objectively’ test these in small maths exercises, but leave out the crucial but more subtle and mathematically less obviously & objectively testable elements. It's complacency. Lack of testing whether students and scholars can actually use the models to make sense of the world beyond extremely dumbed-down ‘planet X’ examples instead of actual Earth case studies. Even we professionals end up imbibing these oversimplified models.

I've only illustrated some of the currently most topical issues on trade and AI—plus the imho most hilarious point about the +- universally assumed upward-sloping marginal cost curve. A similar example would be the concept of elasticity. Ever heard of the value of price elasticity of demand for good X? That's how we teach it. That there's an obvious time dimension—i.e., a function of time—with many goods having small elasticity on a day-to-day scale that increases strongly by the time we're at a decadal scale, I don't see systematically reflected. Not in econ 101 for sure, and I think in much applied modeling neither. I’m all for simplifications, but when cutting out the most salient world features at some point you end up creating an irrelevant abstraction that none will or should try to actually use.

On one hand, despite all this, I definitely don't feel one could say it’s economists “fault” in a simple sense. In fact, really weird economics you get when non-economist academics stray into economics, sometimes not even realizing it (engineers doing economic trade-off calculations or life-cycle analysis, philosophers attempting practical philosophy^[2]). On the other hand, it does strike me how regularly in rationalist or EA circles, non-economists grasp many subtler econ topics more straightforwardly than I'm used to, imho 🤔.

1. ^{^}

   The simplification is so naturally assumed in econ 101, I remember distinctly when, despite moving in econ land a large part of my days, I recently heard for the first time a different person talk about how shaky econ 101's foundations for justifying competitive market efficiency are: Glen Weyl trying to explain to Russ Roberts how goods with downward-sloping costs “[want] to be used by lots of people” while the requirement for profits means restricting use in the standard capitalist model. I got the impression that Roberts—theoretically a full pro of the 'marginal' concept— was barely able to make proper sense of Glen’s explanations, despite how trivial it becomes once you've even only briefly thought about the ubiquity of decreasing long-run marginal costs

2. ^{^}

   I love e.g. how Harsanyi points out the absurd implications of the most famous Rawlsian principle.