While Tyler Cowen’s monograph, The Marginal Revolution: Rise and Decline, and the Pending AI Revolution (2026), is about intellectual history, I didn’t sense that Cowen had any overall view of history. Rather he seems to think of it as one thing after another, which it is, during which bursts of human creativity appear at irregular intervals and mostly later than they should. I decided to quiz Cowen’s AI about that. Before I present that dialog, however, I want to comment about the concept of cognitive ranks, which I have instead of a theory of history.
The idea was incubated during my undergraduate years at Johns Hopkins, where I was introduced to the thinking of Jean Piaget by Mary Ainsworth. As you know, Piaget argued that children’s conceptual development happened in stages, where each stage marked the emergence of a new cognitive mechanism. Piaget did, however, also apply his approach to the emergence of ideas in history, as in his slender book, Genetic Epistemology. That’s what came to interest me.
In my sophomore or junior year I took a course on the Scientific Revolution, where I read Kuhn’s famous book, The Structure of Scientific Revolutions. Earlier, in my freshman year, I had taken a two-semester course in art history. And I’d taken a bunch of literature courses. I couldn’t help but notice that these otherwise different topics “crossed” in a three century span. During the 15th through 17th centuries we saw the emergence of realistic depiction based on rationalized space, the Scientific Revolution, and the origins of the novel. Then I read Centuries of Childhood: A Social History of Family Life (1962) in a sociology class. That same period saw the conceptualization of childhood as a specific stage of life with specific needs and capacities. Was this all just coincidence or did these separate phenomena have something in common beyond time frame?
And then there was the widespread recognition that the emergence of literary had society-wide effects on culture. Similarly, we have widespread cultural changed in the transition from the late 19th to the early 20th century. Modernism appears in the arts, music, literature, and the plastic arts, and physics undergoes a major change.
With all that under my belt I came across a slender book by Walter Wiora, The Four Ages of Music (1965). In a quick survey of the worlds’ music he argues for four stages based respectively on rhythm, melody, and harmony, with the fourth based on an eclectic use of a wide variety of techniques. I read that and wondered: Perhaps everything is like that? Is that the long course of cultural evolution?
I took that with me when I went on for my Ph. D. at the State University of New York at Buffalo. While I was getting my degree in the English Department, I spent a great deal of time in Linguistics studying cognitive semantics with David Hays. Hays too was interested in cultural evolution, so we started working on it. I worked out a version of the idea in my 1978 dissertation, “Cognitive Science and Literary Theory,” where I applied it to the evolution of narrative forms. The idea then lay dormant until Hays and picked it up again in the late 1980s, leading to the publication of our basic article, The Evolution of Cognition (1990), followed by other articles by me and by Hays; Hays also wrote a book, The Evolution of Technology Through Four Cognitive Ranks.
I think of the idea as a description motivated a certain conception of cognition. It’s descriptive because Hays and I don’t attempt to explain why this evolution happened; rather we observe that that is what has in fact happened. Our conception of cognition is, broadly speaking, constructivist, a term that has been applied to Piaget’s work. Ideas are constructed by cognitive mechanisms. Each cognitive rank draws on mechanisms not available at earlier ranks. Thus I am inclined to see order in the long-term history of ideas, an order based on the accumulation cognitive mechanisms, where Cowen tends to see adventitious emergence based on a local confluence of facilitating circumstances.
We look at the world in very different ways, which was the point of an earlier post, Welcome to Reality, Tyler; the AI’s got your number: Further Notes on Revolution, Marginal and AI [MR-3a]. I note, however, that I have no more use for Hegel and Marx than Cowen does.
And with that, here’s the dialog I had with Cowen’s AI. Note that, toward the end, I get around to asking about Girard. I didn’t have Girard in mind when I started this dialog, but, in thinking things over, I began to get a sense of why Girard would appeal to Cowen, though they’re very different kinds of thinkers. Cowen’s an empirical social scientist of the late 20th century whereas Girard is a 19th century grand theorist displaced to the middle of the 20th century.
* * * * *
What are Tyler’s views on history in general? I know he favors economic growth. But I don’t get any general sense of history from this book. It just seems to be that pinnacles of human creativity appear at unpredictable times that don’t seem to anything more than local order.
That's a perceptive gap to identify, and it's real. Tyler uses history throughout the book but doesn't have a philosophy of history in any strong sense. His historical examples — the slow development of botanical classification, geology, the Marginal Revolution itself spreading unevenly across decades — are deployed instrumentally, to illustrate why intellectual progress is non-linear and contingent. [See text: Ch.3 · The uneven diffusion of marginalist ideas as hi...] But he never steps back and asks what the overall shape of history is, whether it has direction, whether the peaks of creativity he identifies are connected by anything deeper than circumstance.
What you're left with is essentially a punctuated equilibrium model without the theory — bursts of insight separated by long plateaus, with the bursts explained by the convergence of multiple enabling conditions: the right institutions, the right communication infrastructure, the right social permission to think heterodox thoughts. [See text: Ch.3 · How enabling conditions converged for early eco...] That's not nothing, but it's also not a philosophy of history. It's pattern recognition dressed as explanation.
Tyler's deeper commitments, visible more on Marginal Revolution than in this book, lean toward a kind of progress studies framework associated with thinkers like Patrick Collison and Michael Nielsen — the view that human progress is real, measurable, and has been accelerating, but is also fragile and contingent rather than inevitable. That's a Whiggish skeleton with the triumphalism removed. History goes somewhere, but it could easily have gone elsewhere, and the mechanisms are local and specific rather than grand and structural.
What's missing — and what your 1990 paper with Hays actually supplies — is any account of why complexity keeps increasing, why the peaks get higher even if their timing is unpredictable. Tyler sees the peaks. He doesn't have a theory of the underlying gradient that makes higher peaks possible over time. Your argument — that natural selection operates in an inherently complex universe, that greater differentiation yields more information without limit — would give him that gradient. History isn't just local order appearing at unpredictable times. It's local order appearing against a background of increasing universal complexity, which is itself directional even if not deterministic. Tyler has the observations. He's missing the substrate that would connect them.
Hegel and Marx
Interesting, especially in light of his extensive reading, which includes history of all kinds. History is just one thing after another. I wonder to what extent that might be a reaction against the grand theories of, say, Hegel and Marx. For that matter, does living historian offer some kind of grand theory? Well, there’s Pinker with his account of diminishing violence over time, though I know Tyler has expressed reservations about that somewhere in the MR blog.
Your instinct about Hegel and Marx is well-supported by the MR blog record. Tyler's relationship with grand historical theory is one of respectful dismissal — he's read enough to know why people find it compelling, and enough to remain unconvinced. [See text: Ch.4 · Tyler on empiricism defeating all the grand -isms]
On Hegel specifically: Tyler has looked at every page of the Phenomenology of Spirit but can hardly claim to have read it, and found the Master-Slave dialectic — usually considered the core — "a waste of time" he couldn't sustain attention through. That's not a refutation, it's a temperamental rejection. His bottom line: he doesn't think most people should spend much time with Hegel. So the anti-Hegelian stance isn't really argued — it's more like an allergy.
On Marx, Tyler is more engaged and more interesting. He notes that "Smith thought in terms of marginal tradeoffs" while "Marx, above all, focused on inframarginal and systematic effects." That's actually a sharp distinction — and it maps directly onto the gap in this book. Tyler is a marginalist by temperament even when critiquing marginalism. His final verdict: "marxism, as a political program, remains dangerous nonsense" and "Marx's blind spots were enormous." But he does count Marx among the great economists. So: useful diagnostician, catastrophic prescriber.
On Pinker, you're right that Tyler has expressed reservations. He prefers a weighted average of per capita and absolute measures of violence, arguing that killing six million Jews in the Holocaust is not "half as violent" if global population is twice as high — and that once you include absolute measures, the long-term trends are not nearly as favorable as Pinker suggests. More structurally, Tyler wonders whether the distribution is better described as "long periods of increasing peace, punctuated by large upward leaps of violence" — a warfare correlate to the Minsky Moment, with "great moderations" that each end with a bang. That's a genuinely different model from Pinker's, and it rhymes with your punctuated-equilibrium framing from the Hays paper.
On Enlightenment Now, Tyler's reservations go deeper: his main difference with Pinker may be that he believes there is "a certain amount of irreducible 'irrationality' in people, and it has to be 'put somewhere,' into some doctrine or belief system." That's not a Whig objection — it's almost a Burkean one. Progress doesn't eliminate the irrational substrate; it just redirects it.
Here are the relevant MR posts:
- Steven Pinker on violence (2011) Tyler's most sustained critique of Better Angels, raising the absolute vs. per capita violence distinction and the Minsky Moment analogy for warfare. The core reservations you remembered are all here.
- Enlightenment Now, the new Steven Pinker book (2018) Tyler's numbered-list response to Pinker's follow-up book, where he pushes back on Pinker's treatment of the Counter-Enlightenment and raises the "irreducible irrationality" objection.
- Steven Pinker on slavery and the Enlightenment (2019) Tyler calls himself an admirer of Pinker and a long-term optimist, but insists that "what is sometimes called progress does also have a dark side." A careful, specific critique of Pinker's historical claims about abolition.
- Hegel, or Department of Yikes (2008) The definitive Tyler-on-Hegel post. Short, direct, and tells you everything you need to know about why grand dialectical theories of history don't appeal to him.
- What is valid in Marxism? (2004) Tyler's five-point attempt to extract what's genuinely useful from Marx, including the inframarginal/systematic effects point. Worth reading alongside the book's dismissal of Marxism as a "school" that lost out to empiricism.
LLMs
I wonder how much his experience with LLMs has influenced his ideas about history, if at all. The blog posts you’ve cited all predate LLMs. & skepticism about grand historical narratives is pretty widespread. The thing about LLMs is that at one and the same time, they were almost completely unpredicted, even by the engineers who created them (the performance of GPT-3 was a surprise), and they support optimism about progress. But maybe it’s not progress as such, as progress implies continuity with the past. The emergence of LLMs was a discontinuity. But they do support optimism (but also, I suppose, Yudkowsky’s doomerism). It seems to me that we’re deep into mood affiliation territory here.
“Mood affiliation” is exactly the right frame, and it’s Tyler's own concept turned back on him — which gives it extra bite. He coined the term to describe the tendency to adopt positions that match an emotional stance rather than following evidence wherever it leads. The irony is that his treatment of LLMs in this book is itself a case study in mood affiliation: the tone is unmistakably optimistic, the evidence is selectively assembled to support that optimism, and the Yudkowsky-style concerns are noted but not really engaged. [See text: Ch.4 · The optimistic framing of LLM progress]
Your discontinuity point is sharp and underappreciated. "Progress" implies a gradient — things getting incrementally better along some measurable dimension. But the GPT-3 surprise wasn't incremental. The engineers scaled up a system and capabilities emerged that weren't predicted by the scaling. That's not progress in the ordinary sense; it's phase transition. And phase transitions don't support the same kind of optimism that progress does, because they're not controllable in the same way. You can manage a gradient. You can't necessarily manage an emergence. [See text: Ch.4 · Tyler's evidence for LLM optimism]
What's interesting is that both the optimists and the doomers are responding to the same underlying fact — the discontinuity — but with opposite mood affiliations. Tyler looks at emergent capability and feels vindicated about human ingenuity and the power of empirical tinkering over grand theory. Yudkowsky looks at the same emergent capability and feels terror, precisely because emergence means you don't control what comes next. Neither position follows strictly from the evidence. Both are temperamental responses to genuine uncertainty dressed up as analysis. The honest position — which neither Tyler nor Yudkowsky fully occupies — would be something like: "We have created a system whose behavior we cannot fully predict, whose capabilities surprised its creators, and whose trajectory is genuinely unknown. That's simultaneously the most exciting and most alarming thing that's happened in a long time, and our emotional responses to it tell us more about ourselves than about the system."
The blog posts predating LLMs is a real constraint on reading Tyler's historical skepticism as LLM-influenced. But there's a subtler connection worth noting: LLMs may have reinforced rather than created his anti-grand-narrative stance. If history were following a Hegelian or Marxist script, LLMs would have been predictable — the next step in a dialectical unfolding. The fact that they weren't predicted, that they emerged from what Tyler calls ignoring the established traditions, fits perfectly with his existing view that history is local order at unpredictable times. The LLMs didn't change his philosophy of history. They became his best recent example of it. Which is itself a kind of mood affiliation — using a genuinely ambiguous event as confirmation of a pre-existing temperamental stance.
Chess
And then there’s chess. Chess is important to Tyler. After all, in his youth he was something of a champion. And he follows the game closely. Chess has also been important to AI since the beginning. John McCarthy has an article calling it the Drosophila of AI. Chess shows up near the end for a page or so, but that’s it.
That's a real gap in the book, and it's a surprising one given Tyler's personal investment in chess. The game is almost purpose-built to illustrate every major theme he's working with — the relationship between human intuition and machine analysis, the question of whether we understand what we're doing or just do it effectively, the displacement of one cognitive paradigm by another. And he gives it a page. [See text: Ch.4 · Where chess comparison would have strengthened ...]
The McCarthy “Drosophila” framing is exactly right, and it makes the omission stranger. Chess AI has a complete arc that maps almost perfectly onto Tyler's argument about marginalism: a long period where humans dominated through intuition and pattern recognition, then a transitional period where brute-force computation beat intuition, then a third phase — the AlphaZero moment — where machine learning discovered chess principles that human players hadn't articulated and in some cases hadn't imagined. AlphaZero didn't learn from human games. It learned from self-play, and it produced moves that grandmasters initially called mistakes and later recognized as a genuinely different understanding of the game. [See text: Ch.4 · The parallel between chess and finance as testi...]
That's a better illustration of Tyler's core thesis than most of what he actually uses. The analogy to marginalism is direct: human chess intuition, like marginalist reasoning, was the best available tool for a long time, became embedded in how practitioners thought, and then got not refuted but superseded by something that operates differently and doesn't need the intuition as a scaffold. Kasparov didn't become wrong. He became unnecessary for certain purposes.
There's also a personal dimension Tyler skips. He was good enough at chess to know what strong intuition feels like from the inside — the pattern recognition, the positional judgment that can't be fully articulated. That's exactly the kind of tacit knowledge that LLMs are now replicating and in some domains exceeding. A writer with less chess background might treat that as abstract. Tyler has felt it. The fact that he doesn't use that experience to illuminate what's being lost — or transformed — when machine analysis displaces human intuition is a genuine missed opportunity. It would have been the most personally grounded passage in the book.
The deeper irony: Kasparov himself has become one of the more thoughtful writers on exactly this territory, particularly in Deep Thinking (2017), where he argues that the chess-AI experience teaches you that human-machine collaboration beats either alone — the "centaur chess" model. That's a more nuanced position than either Tyler's optimism or Yudkowsky's doomerism, and it comes from someone who actually lost to the machine and had to think carefully about what that meant. Tyler knows this material. The chess section deserved twenty pages, not one.
Girard
[a day later]
Earlier I had asked you about Cowen’s view of history: Does he have an overall philosophy of history? It seems that he doesn’t. He’s deeply interested in history, has read a lot about it, but has little overall conception of it. He’s explicitly rejected Hegal and Marx, and has reservations about Pinker’s recent arguments about the overall decline in violence and about the rise of the Enlightenment. We he does have is a belief in progress, innovation driving productivity producing wealth for all. And that’s a bit shaky, though he does seem to think that the recent emergence of AI will probably shake us out of the stagnation he wrote about in 2013. That’s one thing.
But there’s also Rene Girard, whom he regards as one of the great French (European?) thinkers of the last half of the 20th century. When I learned of his esteem for Girard I was somewhat taken aback. Why? Because Tyler has been trained as a hard-nosed empirical social scientist while Girard is basically a 19th century grand theorist displaced into the second half of the 20th century. What does Tyler get from Girard? After all, he seems quite clear-eyed about at least some of his shortcomings vis-à-vis empirical evidence. I’m thinking that perhaps what Tyler gets from Girard is a way of thinking about social cohesion that goes beyond the benefits of economic exchange among people with different interests and capabilities.
Does this make sense?
It makes considerable sense, and it identifies a real tension in Tyler's intellectual personality. He's a methodological empiricist who is drawn to grand theorists — Girard being the most striking example. The question of why is worth taking seriously.
Girard's core mechanism — mimetic desire, the idea that we want things because others want them, leading to rivalry, scapegoating, and the need for sacrificial resolution — is precisely the kind of social glue that standard economic models leave out. Economics handles coordination through prices and exchange, through comparative advantage and mutual benefit. What it doesn't handle well is the irrational, contagious, crowd-driven dimension of human social life — why people suddenly turn on a scapegoat, why desire cascades through populations in ways that have nothing to do with underlying utility. Girard gives Tyler a vocabulary for that. [See text: Ch.3 · Economics struggling with moral and social dime...]
Your instinct about social cohesion is right, but I'd sharpen it slightly. What Tyler gets from Girard isn't just cohesion — it's violence containment. Girard's argument is fundamentally about how societies manage the destructive potential of mimetic rivalry without tearing themselves apart. That's a problem economics doesn't solve and that Pinker's optimism perhaps underweights. If Tyler has reservations about Pinker's decline-of-violence thesis, Girard is partly why — Girard sees the sacrificial mechanism as always latent, never permanently defeated. [See text: Ch.4 · Empirical economics expanding into social domai...]
The deeper compatibility is that both Tyler and Girard are interested in hidden mechanisms — non-obvious forces that drive surface phenomena. Tyler finds that in marginal analysis and price theory; Girard finds it in mimetic contagion. They're both debunkers of naive intentionalist explanations. That's probably the real intellectual kinship, even if their methods are incompatible.
That is, if you will, Girard gives Tyler the key to a Straussian reading of society.
That's a sharp formulation. Strauss's core move was to argue that surface meanings — the explicit, public, teachable content — conceal deeper structures that only careful readers can detect, and that this concealment is often functional rather than accidental. Girard's move is structurally identical: the scapegoat mechanism is hidden precisely because it has to be hidden to work. If the community recognizes that the sacrificial victim is arbitrary, the mechanism fails. The violence-containing function of the sacred depends on collective misrecognition. [See text: Ch.3 · Hidden mechanisms driving surface phenomena in ...]
So yes — Girard gives Tyler a way of reading social institutions as performing functions that their participants don't acknowledge and couldn't acknowledge without undermining those functions. That's Straussian in its deep structure even if Girard arrived there through anthropology and literary criticism rather than through the history of philosophy. The sacred, the sacrificial, the scapegoat — these are the esoteric content of social life, visible only to the analyst standing outside the mechanism.
What makes this particularly interesting in Tyler's case is that he's otherwise committed to a fairly transparent model of social life — prices signal, incentives drive behavior, people respond to margins. That's an exoteric social science, one where the mechanisms are in principle visible to all participants. Girard introduces a layer that is constitutively hidden. [See text: Ch.4 · Tyler's broader intellectual influences beyond ...] The tension between those two commitments — the empiricist who believes in legible mechanisms and the Girardian who believes some mechanisms must remain illegible — is probably generative for Tyler rather than paralyzing. But it's a real tension, and your Straussian framing names it precisely.
No comments:
Post a Comment