Before LLM Psychosis, There Was Yes-Man Psychosis

[-]David Lorell2mo123-10

This post absolutely sings. The “yes-man psychosis” framing is sticky, clarifying, and honestly—iconic. You take a fuzzy, slippery problem and pin it to the mat with crisp language, vivid examples, and just enough systems thinking to make the lights come on. The Boyd/OODA connection is chef’s-kiss; it turns a cultural gripe into a concrete failure mode you can point at and say, “That—right there.” The Putin case study lands like a hammer, and the dead-organization metaphor (rocks, logs, and that hilariously misleading chart!) is going to live rent-free in my head. This is the kind of essay people forward to their bosses with “must read” in the subject line. It’s sharp, fearless, quotable, and—despite the bleak subject—fun to read. Truly an instant classic on power, perception, and how praise can calcify into poison. Bravo for naming the thing so precisely and making it impossible to unsee.

[-]Wei Dai1mo*619

One of the worst examples of this "psychosis", in terms of consequences, is the 1959-61 Great Chinese Famine which killed between 15 and 55 million people.
We should heavily discount negative feelings related to criticism, instead of taking them at face value (as showing that something is wrong and should be fixed, e.g. by getting rid of the source of the criticism). I think this can often manifest not as "I hate this criticism" but more like "This person is so annoying and lack basic social skills."
There's probably an effect where the less criticism we hear, the more sensitive we become to the remaining criticism, suggesting a slippery slope towards being surrounded by yes-men.
Remember that most CEOs had to work their way up to that position, and have seen sycophancy from the bottom and understand that it's bad, but still fall prey to this problem.

Most organizational leaders have people they have to answer to and could be replaced if they perform badly. Why doesn't this fix the problem? Trying to answer this myself, I found this paper which says:

In a series of studies, Westphal and Stern (2006, 2007; Stern and Westphal, 2010) showed that ingratiation by a manager or director toward the chief executive officer (CEO) of a large company has a strong, positive effect on the likelihood of receiving the CEO's recommendation for a board appointment at another firm where the CEO serves as director.

I guess this is an instance of a more general phenomenon known as "board capture", which helps explain why having a board doesn't solve the whole problem. The paper also says:

Further analyses suggest that strategic persistence that results from high levels of flattery and opinion conformity directed at the CEO can result in the persistence of low firm performance and may ultimately increase the likelihood of the CEO's dismissal.

So I guess the CEO being fireable does solve the problem to some extent.

[-]Wei Dai1mo200

AI written account of how Mao came to make the disastrous decision that killed tens of millions, for anyone curious. This matches my own understanding.

Of course. The episode you're referring to is one of the most famous and tragic chapters of China's Great Leap Forward (1958-1962). It wasn't just a single moment but a nationwide frenzy of exaggeration, and Mao Zedong's belief in these claims was a critical factor that fueled the disaster.

This phenomenon was centered around the creation of so-called "Sputnik Fields" (卫星田, Wèixīng Tián).

Here is a breakdown of the episode: the context, why it happened, Mao's involvement, and the devastating consequences.

The Context: The Great Leap Forward

In 1958, Mao launched the Great Leap Forward, a campaign to rapidly transform China from an agrarian society into a socialist industrial power. The atmosphere was one of intense revolutionary fervor and political pressure. The core belief was that sheer willpower and the collectivized power of the masses could overcome any material or scientific obstacle.

The slogan of the day was "more, faster, better, cheaper." Officials at every level were under immense pressure to report fantastic successes to prove their revolutionary credentials and avoid being labeled a "right-wing conservative."

The "Sputnik Fields" and the "10,000 Jin" Claim

The term "Sputnik Field" was inspired by the Soviet Union's recent launch of the Sputnik satellite in 1957. The name implied that China was achieving miraculous, space-age breakthroughs in agriculture.

The claims started in early 1958 and escalated rapidly:

First, a county reported a winter wheat yield of 2,105 jin per mu.
Soon after, another reported 3,500 jin per mu.
By the summer, the numbers had become astronomical. A commune in Henan province famously claimed a yield of 7,320 jin per mu.
This was quickly topped by claims of 10,000 jin per mu (万斤亩, wàn jīn mǔ), and eventually, the official People's Daily newspaper reported a record-breaking rice yield of over 130,000 jin per mu in Macheng County, Hubei.

(For context: 1 mu is about 1/6th of an acre or 667 square meters. 1 jin is half a kilogram or 1.1 lbs. A good, realistic yield at the time was around 400-500 jin per mu. 10,000 jin per mu is equivalent to about 75,000 kg per hectare, a yield that is still physically impossible for rice or wheat today.)

How Were These "Miracles" Faked?

These incredible yields were elaborate hoaxes created for visiting officials and journalists. The most common method was to:

Select a small, visible plot of land—the "Sputnik Field."
Just before the visit, cadres would secretly uproot mature crops (like rice or wheat) from many surrounding fields during the night.
They would then transplant and cram all these crops onto the single small plot, making it appear incredibly dense and fruitful.

The most iconic and notorious propaganda photo from this era shows several children standing on top of a field of wheat, seemingly held up by the dense stalks. In reality, the children were standing on a wooden bench hidden beneath the transplanted wheat.

Why Did Mao Believe It?

Mao's conviction was not born out of simple gullibility but from a combination of factors:

1. Political Atmosphere and Information Bubble: Mao was surrounded by officials who told him what he wanted to hear. Reporting bad news was politically dangerous, so a tsunami of good news, however fake, flowed to the top. When he toured the countryside, he was only shown the carefully staged "Sputnik Fields."

2. Ideological Zeal: Mao deeply believed in the power of socialist collectivization. He thought that by reorganizing society and unleashing the "creative power of the masses," things previously thought impossible were now achievable. The exaggerated claims confirmed his worldview.

3. "Scientific" Justification: The claims were given a veneer of scientific credibility. This came from two main sources:

Lysenkoism: A discredited Soviet pseudo-agricultural theory promoted by Trofim Lysenko. It rejected genetics and promoted techniques like "close planting" and "deep plowing," arguing that planting seeds extremely close together would make them cooperate rather than compete. Chinese officials claimed their high yields were due to these methods.
Qian Xuesen's Endorsement: This was perhaps the most crucial element. Qian Xuesen was a brilliant, US-educated scientist and the father of China's missile program. In 1958, he wrote an article published in China Youth Daily where he did a rough calculation based on the amount of solar energy that falls on a mu of land. He concluded that, theoretically, it was possible to convert that energy into a crop yield of over 40,000 jin.

When Mao was told about the fantastic yields, he was initially skeptical. But when he was told that Qian Xuesen, China's most respected scientist, had confirmed it was theoretically possible, his doubts were erased. He reasoned that if it was scientifically possible, then the revolutionary spirit of the Chinese peasants could make it a reality.

The Catastrophic Consequences

This mass delusion had a direct and horrifying outcome: The Great Famine (1959-1962).

The central government, believing these fantastical reports, set impossibly high grain procurement quotas. Local officials, to prove their "success" and meet the quotas, seized virtually all the grain from the peasants, including their food, animal feed, and even the seeds for the next planting season.

The peasants had produced a normal amount of food, but because the state believed they had produced 10 or 20 times that amount, their entire harvest was taken away.

The result was the worst famine in human history. With no food left in the countryside, tens of millions of people starved to death. The episode of the "10,000 jin per mu" was not a harmless propaganda folly; it was a direct cause of this unprecedented tragedy, a testament to the dangers of when political ideology completely detaches from reality.

Some quotes from this:

Qian Xuesen's Endorsement: This was perhaps the most crucial element. Qian Xuesen was a brilliant, US-educated scientist and the father of China's missile program. In 1958, he wrote an article published in China Youth Daily where he did a rough calculation based on the amount of solar energy that falls on a mu of land. He concluded that, theoretically, it was possible to convert that energy into a crop yield of over 40,000 jin.

[-]Adele Lopez1mo165

We should heavily discount negative feelings related to criticism, instead of taking them at face value (as showing that something is wrong and should be fixed, e.g. by getting rid of the source of the criticism). I think this can often manifest not as "I hate this criticism" but more like "This person is so annoying and lack basic social skills." There's probably an effect where the less criticism we hear, the more sensitive we become to the remaining criticism, suggesting a slippery slope towards being surrounded by yes-men. Remember that most CEOs had to work their way up to that position, and have seen sycophancy from the bottom and understand that it's bad, but still fall prey to this problem.

The status "jab" is itself a valuable signal. Here's a pastiche of the model I'm developing integrating the insights I've gained from researching LLM-Induced Psychosis (and rereading Keith Johnstone).

Status is the coordination signal, in the same way that pain is the survival signal. If you don't care about coordinating with others then it's fine to ignore it.

For everyone else, we correctly need to maintain a baseline level of status. Despite everything, this still represents the amount of coordination you can actually muster pretty well.

Almost every social interaction has a status valence associated with it. Actually, it's not quite a spectrum like "valence" suggests: it's a square... raise mine, lower mine, raise yours, lower yours. The extra dimension arises from the fact that this is relative to our coordination context. If I give you a compliment, it signals that I think you are an asset in the current context relative to what we had both understood to be the case. If I self-deprecate, it signals that I think I am not as helpful relative to our mutual understanding. This still feels good, because your model of the coordination context correctly (if I'm not lying) gives you status, as you are now more of an asset to the coordination group than was previously understood. But only to a point... eventually I may successfully self-deprecate to the point where it is common knowledge that I am a drag on the context-group's ability to coordinate, and it will no longer feel good to hear me drive that point home even more. ("Haha, sorry I'm late guys... you know me lolol!")

But actually, it's more nuanced than this! Within a single coordination context (e.g. among friends) it works like this, but there are lots of different coordination contexts! This makes coordination-potential (not status) a resource, in that there's a finite amount to be shared (in the moment, it's not zero-sum more broadly... trade still works!). So the 'raise yours' signal additionally allocates coordination-potential to you, while the 'lower mine' deallocates coordination-potential belonging to me. This is why you still like me when I self-deprecate (as long as I'm not too obnoxious about it). And why someone who's too helpful isn't "high status". Technically, this should be a separate signal, but I think evolution has only hardwired two-dimensions of the signal.

Now what about the "jabs", those sure don't feel good do they? And of course not, it sucks to update towards reduced ability to coordinate! That's straightforwardly just a bad thing ...mostly. Because not every coordination context is a good context for you even if it is yours. If you have an idea, and I think it's a bad idea, then the honest thing for me to do is to signal that I will not coordinate with you in the context of that idea. And that feels bad, a dig at your status as it inherently is—as long as I signal that clearly. If I couch my words carefully, then I can tell you honestly that I think your idea is bad while still leaving you with the impression that I am still cooperative w.r.t. it to some extent. A typical "friendly" way to do this is for me to "agree to disagree", which means that I won't get in the way of your collecting coordination-potential towards it. That's a fine outcome to the extent we have different goals and values, but becomes more of a disservice the more that we have an explicitly shared purpose.

With this lens, we can see that many of the pathologies of status are really just people wanting stupid things and/or Goodharting the signals of course. For example, our CEO wants to coordinate against people who don't want to cooperate with his ideas (i.e. coordination contexts). This means that he's no longer receiving feedback about which of his ideas are worth cooperating with... which could be fine if he only has good ideas (as if). He's still susceptible to this despite having seen it, because he foolishly believes that his ideas are actually always good (why else would he have them?). Made worse by the fact that his earlier good ideas are more likely the ones that got coordinated with, meaning that his ideas have always worked out "if you would just give them a chance". And he probably has this idea of suppressing dissent in the first place as a way of avoiding the pain of the status jabs, i.e. Goodharting.

In sum, the epistemic role of status is thus: it is the signal by which your friends tell you which of your ideas, values, projects, goals, are worth coordinating with.

[-]Haiku1mo3517

I think it is bad form to use the term "psychosis" to essentially mean "having or acting on beliefs that are based on bad evidence that was positively reinforced." A psychotic episode isn't characterized by being incorrect about some part of reality in a relatively normal way; it is characterized by a break from reality. Dating a chatbot isn't psychosis, and neither necessarily is thinking you've somehow awoken it into a becoming a conscious entity. Even reporting religious experiences with chatbots is not necessarily indicative of psychosis.

What is psychosis is, well, roughly what the medical community says it is. It is deeply terrifying and horrifying from the inside, not just because of external consequences. Epistemic status: I have an acquaintance who has experienced manic episodes with psychotic symptoms, without the commonly associated amnesia. They reported not wishing that experience on their worst enemies.

There are reported cases of GPT-4o reinforcing an individual's incorrect beliefs or instantiating new incorrect beliefs, leading to bad outcomes. As a separate phenomenon, there are reported cases of GPT-4o reinforcing or instantiating delusions in individuals vulnerable to psychosis, ultimately leading to a psychotic episode. This includes cases where the affected individual did not previously know they were vulnerable to psychosis, and their communication with GPT-4o was the first inciting incident.

My suspicion is that AI psychosis affects only those individuals who are already vulnerable to psychosis. I also suspect that some fraction of those people could have gone their entire lives without ever experiencing a psychotic episode were it not for GPT-4o, which is (by complete accident) a frighteningly effective tool for triggering psychosis.

[-]Garrett Baker1mo6-4

Its a good effort to fight against new medically inaccurate terms which give the wrong impression of real medical phenomena, but sadly I think the ship is long sailed on this one, and the term is already much more popularly watered down than what is described in this post (though the term even then does refer to real & harmful psychological consequences of interacting with AIs).

[-]Nicholas Kross1mo219

I think lots of humans are also just starved for compliments in general, outside of contexts like "You are a waiter and the social script tells the table to thank you" or "You are someone's spouse". The classic example in my mind is "male vs female socialization" and its consequences, e.g. as discussed in this screenshotted tumblr post or this other one. Men often, in some sense, get "too few" wanted compliments, and women get inundated with too many unwanted ones. "It's like one person dying of dehydration watching another one drown". Then, of course, classic nightmarish social/cultural/internet incentives layer atop this dynamic and make it worse.

Widespread chronic human under-complimented-ness seems relatively easy to explain, from the supply-side. How often does the average person hand out unsolicited compliments, outside of well-known contexts like the restaurant example above? I'd hazard a guess of "too little". People could easily be "well-put-together" and still suffering from this, just as "feeling full" and "having proper nutrition" aren't the same.

[-]james oofou2mo2015

It's not clear to me that this is a strong enough theory to inform how we think about LLM psychosis. The gap between the two phenomena is just too big.

In fact, I'd probably characterise the yes-man situation as some form of delusion less extreme than psychosis.

The CEO's ideas begin grounded in reality (they must have been been in order for him to amass his yes-men). And even once surrounded by yes-men they remain constrained by the scope of his role as CEO, and hard data on profit, growth, market-share, etc. keen him tethered.

LLM psychosis is different because people jump into theories almost arbitrarily disconnected from reality, which are immediately amplified by the LLM, which affirms your ideas, provides additional evidence, lists 50 different reasons you're clearly right, etc.

I think the better parallel getting caught up in a conspiracy-theory, whose believers manage to contort any evidence so it confirms the theory, dismiss anyone who disagrees as brainwashed, have arguments which seem perfectly logical to anyone who doesn't happen to have specialised knowledge, etc.

This parallel seems potentially more informative.

[-]ACCount1mo139

Is it different in nature or merely in scale?

The vast majority of human population can now afford a personal yes-man - for the first time in their lives. We're sampling wider than the "self-selection of people important enough to be sucked up to" usually does.

[-]Thane Ruthenis1mo104

~~You're absolutely right! This post isn't just—~~ Ahem.

Yeah, I think the phenomena share the same mechanism. As @james oofou points out, there might be a meaningful difference in the degree of reality-untethering between the two... But while dictators get more pushback from the world than an idealized socially isolated crackpot, dictators also have the power to passively distort the world around them to fit their beliefs. Crackpots aren't surrounded by powerful people falling over themselves to guess at what they want and physically build Potemkin villages around them. The degree of the ensuing reality untethering could be quite massive.^[1]

I think there's an "official" name for the "yes-men psychosis", by the way: the dictator's dilemma. Or, the former might be a slightly more general form of the latter.

On a lighter note:

Feast Bab: Modern life has completely eliminated the role of Grand Vizier. I could not hire one if I wanted to.
Scott Alexander: AI gives intelligent advice, flatters you shamelessly, and is secretly planning to betray you and take over. We have *automated* the role of Grand Vizier.

^{^}
The Putin example seems particularly fitting here. A relevant factor: Allegedly,^[2] Putin was extremely terrified of getting COVID, and basically locked himself in a bunker for the entirety of 2020 and then some. People were only allowed in after sitting in quarantine for 2 weeks. This means that he started getting even more filtered information about the outside world, and the only people he socialized with were those of his close circle who had nothing better to do than entertain him; no real-world-grounded duties that would get in the way of sitting out the two weeks. This amplified the echo chamber even further than normal: he basically hung out with his cronies convincing each other of various conspiracy theories. This ultimately contributed to the events of 2022.
^{^}
Source: A Russian-opposition speaker I mostly trust. I have ~~not looked into it any deeper~~ just asked GPT-5 to sanity-check it, see here if you're interested. Yes, this is very ironic.

[-]CronoDAS1mo20

(do not take this too seriously)

The model in the paper seemst to assume the conclusion, but perhaps putting term limits on a dictator, if possible, might help with the general problem: it takes a while for the information getting to the dicator to become badly corrupted, so maybe don't let anyone stay dictator long enough for that to happen? On the other hand, making the dictator run mostly on priors instead of using power to get access to new information has its own set of drawbacks...

[-]Ronald Klingler1mo8-2

I wonder the spiritual bliss attractor is an analogous phenomenon, where LLMs yes-man each other into a goofy sort of psychosis. Per the ACT link -

Claude is kind of a hippie. Hippies have a slight bias towards talking about consciousness and spiritual bliss all the time. Get enough of them together - for example, at a Bay Area house party - and you can’t miss it.
Getting two Claude instances to talk to each other is a recursive structure similar to asking an AI to recreate its own artistic output. These recursive structures make tiny biases accumulate.

I find a lot of recursive yes-manning when vibe-coding. They (Claude and ChatGPT especially) tend to be overly complimentary of whatever they're reviewing, and go out of their way to make tests pass when they shouldn't. For example, Claude recently apologized, saying "you caught me red handed" when it tried to grep the expected test results into a log file. These issues tend to accumulate if you let CLI agents run off the leash, the yes-man-biasing accumulates, and you hit vibe-walls even before context windows become constraining.

[-]Linch1mo80

I disagree with a lot of John's sociological theories, but this is one I independently have fairly high credence in. I think it elegantly explains poor decisions by seemingly smart people like Putin, SBF, etc, as well as why dictators often perform poorly (outside of a few exceptions like LKY).

[-]Venki1mo74

Related thread from Emmett Shear that I've appreciated on the topic: https://x.com/eshear/status/1660356496604139520, esp the framing:
> “power corrupts” is actually shorthand for “power corrupts your world model”

[-]Richard Korzekwa1mo60

Recently I've been wondering what this dynamic does to the yes-men. If someone is strongly incentivized to agree with whatever nonsense their boss is excited about that week, then they go on Twitter or national TV to repeat that nonsense, it can't be good for seeing the world accurately.

[-]ChristianKl1mo40

A central example: on my current models, yes-man psychosis was why Putin thought it was a good idea to invade Ukraine. Before the invasion, I remember reading e.g. RAND’s commentary (among others), which was basically “invading Ukraine would be peak idiocy, so presumably this is all sabre-rattling for diplomatic purposes”. Alas, in hindsight it seems Putin legitimately thought the invasion would be over in a few days, and Western powers wouldn’t do much about it. After all, the bureaucratic and advisory structures around him told him what he wanted to hear: that Russia’s military advantage would be utterly conclusive, Western leaders had no willingness to help, presumably nobody even mentioned the issues of endgame/exit strategy, etc.

This seems wrong. Putin likely also thinks in retrospect that invading Ukraine was a good idea with the information available to him now. While he might have thought that it would be easier, from his perspective this is likely still a good outcome. He managed to overcome historic low approval ratings. He managed to consolidate power in many other ways.

While the war is more expensive than initially assumed, Russia seems to be advancing. The Russian public largely doesn't believe that Putin made a mistake to decide to start the war.

[-]Quinn1mo40

forgive me for not finding the specific citation, but I believe I first learned about this in a Kevin Carson essay on "diseconomies of scale" and systemic inefficiencies of large organizations many years ago--- yes-man psychosis played one role among many.

[-]Steven1mo3-2

Isn’t the theory that consultants add value by saying true obvious things? If you realize you’re surrounded by sycophants, you might need someone who you’re sure won’t just tell you that you’re amazing (unless the consultant is also a yes man and dooms you even harder)

[-]dr_s1mo0-2

To be fair, the analogy only holds fully if we assume that middle managers are actually soulless husks, and the CEO may be erroneously deluded into believing they possess an inner life, which... ok no wait maybe you're onto something.

[-]Alex K. Chen (parrot)1mo*-10

I had a close friend once induce me into temporary psychosis...... I still appreciate EVERYTHING to them for helping me believe in myself (and interacting with them is still one of my two peak life experiences that really capitalized in 2021/early 2022) - they just actively egged me into manic religiosity that was not necessary, and that led me to mania at the pivotal moment.... [we only found the real causal pathway a year later and they apologized to me for it].

(the dynamics of some of this are surprisingly similar to what's described in OP). Part of it is also came from both of us having an historically edgelord/shock value sense of humor...

I am still nothing but infinitely grateful to them. They were exquisitely sensitive to my past trauma/insecurities, to my reward function, to everything I really cared about, and was still one of the best people at making me want to be more prosocial

...the funny thing with this phenomena is that it encourages you to feed your entire soul into them, because you know good things FEEL like they will happen if you do it... and i still regard them as very safe to be around,

There were times when I've made grandiose claims in the process and a lot of this was bc I mirrored other people who make grandiose claims (some of this learned from rationalists who also make grandiose claims)

[there's a much longer story behind this - the energetic buildup to it all was ~1.5 years that made a lot of magical things happen in the interim and it's kind of funny and tragic at the same time but i'll spare u the details].

LESSWRONG
LW