Motus Mentis

Curtis Yarvin, a.k.a. “Mencius Moldbug”, has published the second installment of his five-part “Clear Pill” essay series over at The American Mind. The new essay is about how coordinated, pervasive error enters the national culture in distributed, democratic societies — i.e., without the top-down influence of centralized, authoritarian control.

The essay is long — in true Moldbuggian form, longer than it needs to be — but the gist is this: that, in contrast to authoritarian societies in which public opinion and permissible narratives are controlled and synchronized from above, some other mechanism must be at work to achieve the same effect in democracies. Mr. Yarvin’s essay is an attempt to provide a theory for this.

We read:

Logically, either liberal democracies are inherently free from pervasive error or some other mechanism can cause pervasive error in liberal democracies.

Liberal democracies have no dictator, no center, and no point of coordination. We’re looking for a mechanism which could cause pervasive error without central coordination: Gleichschaltung without Goebbels.

In the 21st-century Western model of government, elite consensus is set by a market for ideas. This consensus narrative drives public policy, which exists within the reality of the consensus. If this market somehow malfunctioned…

Let’s posit that a normal First World civil service will always implement reasonable policy relative to the truth in which it operates. Yet if the market for ideas fails, and generates a delusional narrative, policies that are competent and reasonable within this illusion may prove incompetent and unreasonable in reality.

Good so far. We can see this happening all around us.

Let’s focus, therefore, on the integrity of the public narrative—which is generally developed outside the civil service proper.

On a helicopter there is a part called a “Jesus nut,” because it holds the rotor to the driveshaft. This little steel ring is all that stands between you and the next world. In the modern democracies, markets for truth perform a comparable function. They are all that stand between us and Orwellian mass delusion, homicidal or otherwise.

Anything this important is in a very real way sacred. And indeed many seem to engage their spiritual instincts with these markets. This is understandable but erroneous. Don’t pray to the Jesus nut—inspect it for cracks.

A market for ideas is a machine. The purpose of the machine is to reject error and discover truth. The presence of pervasive error in the output tray of the machine indicates malfunction. This malfunction is an engineering failure and can be debugged as such. No mystical sentiment at all needs to be engaged.

Yarvin asks how, exactly, a crowd-sourced idea market — in which the “crowd” consists of top-tier elites and influencers — detects truth. His opinion is that it does so by using beauty as a proxy for truth:

Every idea market is an aesthetic device. It selects for the most beautiful ideas—according to the taste of its audience. A market measures desire; desire is beauty in action. And by measuring beauty we measure truth, because truth (like my man Keats said) is beautiful.

Seen from an aesthetic perspective, every idea is a story. A marketplace of ideas is a market for stories. An idea can be any kind of story, factual or fictional; even a song or a film. All creative markets are affected by the phenomena we will describe. But the most important idea markets are those which claim to tell a story, both true and truly illustrative, about the real world. Abusing English only slightly, we may term these histories. The definition is broad enough to include all of science—or, as once it was known, natural history.

Our consensus narrative of reality is the output product of the machine which is our market for beautiful histories. From the perspective of our collective epistemic security, the integrity of this mechanism is of the highest possible importance.

It is easy to see the exploitable vulnerability. Is a market for beautiful histories always a market for veracious histories? Yes, because truth is beautiful and lies are ugly. And—no, because other qualities might also be beautiful. These qualities cannot cancel the Keats effect. But if they could overpower it, the result would be a beautiful lie.

Is beauty really what an idea-market must use to detect truth? I’m not so sure, but Mr. Yarvin seems to be “all in” on this:

An idea market can only measure truth by measuring beauty. If beauty is not truth, we assume, beauty is truth plus white noise. Given a sufficiently large audience of high average quality, these individual quirks of taste will cancel out, leaving only the consistent, high-quality truth promised in the brochure.

This assumption had better be true! If deviations from truth are not white noise, but their own coherent signal, averaging will not filter them out at all. These signals will pass through the market and emerge in the output, competing with truth.

Any such competing aesthetic signal is an attack vector on the integrity of any truth market. Only beauty can be measured. Any divergence between truth and beauty can be exploited.

“Only beauty can be measured.” Is that true? (Related question, I suppose: Is it beautiful?) I can certainly agree that beauty can (and does) introduce bias — but to the exclusion of everything else? I doubt it; especially in a crowd-sourced system such as is described here. But to accept for now the article’s argument, maybe bias is enough. There’s certainly no question that the aesthetic appeal of a proposition can, because we are emotional beings, affect its truth-valence. In this way, we might come to prefer a beautiful lie to an ugly truth.

Next comes the question: What can give a proposition emotional appeal? What, in addition to truth itself, might make some of the competitors in the idea-market more beautiful than others? Here Mr. Yarvin draws on the classical Greek concepts of thymos (which I’ve usually seen spelled thumos), pistos, and agape. (Loosely translated, these mean spirited ambition, loyalty/honor, and altruistic empathy, respectively.) That which appeals to any of these will be colored with emotional appeal, and so, when ideas are set loose in the elite’s marketplace, there is a security flaw: we may overrate them enough to successfully bias the market against more truthful ideas.

Mr. Yarvin illustrates this, convincingly enough, with a look at climate science — making, along the way, the important point that the same effects of ambition and loyalty-to-the-master that might make us leery of Exxon scientists’ reporting on climate change would also apply to any other scientist affiliated with any institution or faction whatsoever. The only exceptions to this are truly abstract disciplines such as mathematics, whose truths are by their nature orthogonal to society and politics:

[W]e actually know nothing about anything, except math and super-hard science.

The conclusion, then, is that there is a design flaw in the way our distributed system detects truth:

Any despotism is the tyranny of error. No one sensible could possibly mind a benevolent dictator who was also always right.

Distributed systems are hard. It’s amazing when they work at all. We shouldn’t be surprised to see failure modes. But nor should we have to live with, or be ruled by, pervasive error.

So we should admit that distributed despotism is caused by the way power poisons truth markets. Putting a truth market in power is unsound political engineering. A previously reliable machine will start to evolve pretty lies. This is a slow and degenerative process which cannot be reversed.

Putting a church in charge of the government is not putting God in charge of the government. Putting a truth market in charge of the government is not putting truth in charge of the government.

Returning to our original comparison between centralized and distributed despotisms—dictatorship and democracy, platypus and man—it’s remarkable how these very different systems converge on the same “track-aligning” effect, repressing thoughts that challenge the regime, promoting ones that flatter it. Yet distributed evolution does often uncannily mimic central design.

Power is power. Through some pyramid of bureaucrats, the Central Committee ordered Havel’s greengrocer to display in his window their official slogan: “workers of the world, unite.” Do not our stores have such slogans in their windows? I see them every day; don’t you? They even come preprinted in all the right colors.

Yet our slogans are not so standard, nor our cadre so pyramidal. This is no flaw of power, but its dazzling perfection. The entire system, whose objective strength we just compared to the Czech secret police, formally does not exist at all. Very cool.

Yarvin closes with this:

We end (for now) with a paradox on a long horizon. Consider this cycle:

— The intellectual command economy rules. Public opinion is directed by a dogmatic bureaucracy, rife with pervasive error, systematically incapable of changing its mind.

— An unofficial free market for truth evolves. This market cannot be poisoned by power, because it has no power. It develops a higher-quality product than the official narrative.

— A new epistemic elite arises. The old intellectual bureaucracy, smart enough to sense its own inferiority, hands power to the new truth market. A new golden age begins.

— Dogmatic bureaucracy returns. Slowly and inevitably poisoned by power, the once-vibrant civil society slowly ossifies into a dogmatic bureaucracy, evolving more and more pretty lies until pervasive error is again the norm.

Western civilization has been repeating this story over and over for roughly the last half-millennium. At each step in the cycle, there is no clear way to prevent the next.

There’s much more; you should read the whole thing, here. In particular, I didn’t touch on Mr. Yarvin’s discussion of agape, and I have questions about the apparent misfiring of empathy that seems to be a hallmark of the modern West. (I’ll save those for another post, though.)

The most obvious criticism of this model is the one I mentioned above, and is really not a criticism of the idea of the vulnerability of a crowd-powered truth-market to emotional bias: it is the idea that the effect is so overwhelmingly strong — “only beauty can be measured” — that it defeats rational methods of truth-checking our ideas. After all, Mr. Yarvin assumes that it’s obvious to us that much of what drives policy consists of “beautiful lies” — and so it is. How is it, then, that he and we can see them, while “elites” can’t? I suppose his answer would be that we who see them are the ones who are not in a position to be biased toward them by ambition, or loyalty to our particular masters, or the appropriate concept of altruism. And perhaps it is so.

Mr. Yarvin acknowledges that his is a sympathetic position, in that he doesn’t suppose that the affected elites are knowingly, and cynically, promulgating these beautiful lies as lies; he gives them the benefit of the doubt:

A lie is not a fiction; it pretends to be true. All the tilt in the world cannot make a truth market endorse a naked fiction.

Of course, some people are sociopaths. They actually find lies inherently pretty, useful, or both. Sociopaths are rare and hard to wrangle, and it is hard to imagine a pure audience of them. There are always enough sociopaths to fill a niche, never enough to fill a room. So it is hard for sociopaths to sway a market. (If you imagine that there is some party, tribe, class, or nation of sociopaths, this is very sad and you should stop.)

…[Arguments against those who promulgate pretty lies] are too often stated as indictments, implying systemic mens rea. This deep factual error is deeply enfeebling… I, too, dislike progressives. But they are normal people, not evil zombies.

Once you’ve read the article, here’s an exercise: consider the prevailing “beautiful lie” of our age, namely “Diversity is our strength”. How, according to Mr. Yarvin’s model, did the action of thumos, pistos, and agape cause the truth-market machine to produce such a bad result? What was the input — historical, social, or otherwise — that led to such faulty output?