In nineteen seventy-eight, a researcher at Johns Hopkins named Mary Ainsworth published a procedure that lasted about twenty minutes. A mother and a baby in a room. A stranger comes in. The mother leaves. The mother comes back. That's it. And from twenty minutes of watching how a one-year-old greets a returning parent, she built a taxonomy of expectations about whether the world is safe.
The Strange Situation.
And the thing that jumped out at me is what those twenty minutes predict. Not just how the kid acts in the room. How they read a stranger's face at four. How they handle a friend's ambiguity at eleven. Whether they assume the person who didn't call back forgot or decided they weren't worth calling.
The internal working model.
Which is a clinical way of saying: you walk around with a running guess about whether people are on your side, and you almost never check it against anything.
And that's what Daniel wrote in about.
That's what Daniel wrote in about. He's continuing the thread we've been pulling on for a while, the difficult people, the narcissistic ones, the ones whose presence reorganizes a household. But he's pointing at a consequence that doesn't get its own conversation. His words were, roughly, the exhausting part wasn't simply their chronic anger. It was the anger and negativity they held toward every human being they encountered.
Every human being.
Kind family members, relentlessly attacked. Somebody does this person a favor, and they find the one thing that wasn't done well and use it as the attack. Professionals are incompetent. Friends have hidden motives. And if you listen long enough, and take it at face value, the whole world is conspiring against one person.
And it's not the listening that does the damage.
Right. That's the part he wants to sit with. The harmful part wasn't hearing the negativity. It was internalizing, through osmosis, the belief that the way this person talked about others was how everybody behaves when the doors are closed. You start to believe people can't be trusted. That they might seem perfectly nice on the surface while saying the cruelest things behind your back. And here's the strange asymmetry he flagged. You don't use your own decent behavior as counter-evidence. You don't say, well, I don't do that, so the sample is flawed. Instead you become reserved. Scared to open up. Because you've learned that other people cannot be trusted.
And he named the distortion.
He named it himself. Faulty extrapolation. Taking a data set of one pathological person and extending it to every human being in the abstract. Including people you've never met.
That's the whole episode, right there.
That's the whole episode. Two halves. How the distortion gets installed. And how somebody gets enough faith back in basic decency to form open relationships again. So let's start with the mechanism. What's actually happening in someone's head when one person's worldview becomes their own.
The frame I'd put on it is this. Most of the conversation about difficult people is about the direct damage. Your self-esteem. Your nervous system. Your sleep. This is about the indirect damage. The target isn't how you feel about yourself. It's your epistemology of human nature. What you think people are.
Epistemology of human nature. That's a big phrase for a small thing.
It's a small thing that runs everything. An internal working model is a generalized expectation about whether other people are safe, reliable, well-intentioned. Bowlby's term, Ainsworth built the empirical scaffolding. And the key property is that it operates outside awareness. You don't consult it. You don't see it. It's the thing doing the consulting.
So when somebody cuts you off in traffic and you think, that guy's just careless, versus, that guy thinks he owns the road. The model picks the interpretation before you've noticed there was a choice.
Ambiguous social cues get resolved by the model, not by the evidence. And the model is normally updated by experience. That's the design. You meet enough reliable people, the model tilts toward reliable. You meet enough betrayals, it tilts the other way.
But under prolonged exposure to a hostile person, the updating mechanism gets captured.
Captured is the right word. The hostile person becomes the primary source of data about what people are really like. Not one source among many. The dominant one. And they present themselves as the only one who sees clearly.
The clear-eyed realist.
The clear-eyed realist who sees through everyone's facade. And they're relentless about it. Every day, a new installment. So let's dig into the mechanism itself. What's actually happening when you start believing that everyone is two-faced.
I want to take the first piece, because I think it's the least intuitive. The hostile person is a constant stream of social data. That's the frame. When you're around someone who badmouths everyone, you're not just hearing complaints. You're being fed evidence. Evidence about human nature. Presented with total confidence, by someone who knows these people personally, whose read on them is supposedly firsthand.
And the brain treats it as a sample.
The brain treats it as a sample. That's the crux. You don't experience it as one person's opinion. You experience it as reporting. Here's what the cousin did. Here's what the colleague said. Here's what the neighbor is really like. And the brain, doing what brains do, generalizes from available data. One person, especially an intimate or trusted figure, consistently describing others as cruel and selfish and two-faced. The brain treats that as representative.
And that's the faulty extrapolation. Treating n equals one as if it were n equals all.
The thing I want to be careful about is that this isn't stupidity. It's not gullibility. It's a normal cognitive shortcut applied to a poisoned data set. The shortcut is correct. Generalize from your sample. The sample is the problem.
The sample is curated. That's the piece people miss. It's not a random sample of human behavior. It's a curated exhibit. The hostile person selects the stories that confirm the model and discards the ones that don't, and they do it so consistently that you never see the curation happening. You just see the exhibit.
And the exhibit has a curator with a motive.
The motive is that the model has to be true. If everyone is two-faced, then the hostile person's isolation isn't their fault. If everyone is a user, then the burned bridges aren't bridges they burned. The worldview is load-bearing. It's holding up something.
So the negativity isn't incidental. It's doing work for the person producing it.
And that's why it's so relentless. It's not a mood. It's maintenance.
Okay. Second piece. And this is the one Daniel circled. Why is chronic negativity directed at everyone more harmful than anger directed at you?
Because of where the problem lives.
Say more.
When someone is angry at you, the problem has a location. It's them. It's the relationship. It's the situation. You can point at it. You can leave the room. You can leave the relationship. The problem is bounded. But when someone is angry at everyone, the problem becomes the world. And you can't leave the world.
You can't leave the world. So the hostility stops being a feature of one relationship and becomes a background condition of existence.
A background condition. Which is a completely different psychological object. You can't problem-solve a background condition. You can't negotiate with it. You can only adapt to it. And the adaptation is: assume the worst, because the worst is the weather.
That lands. That's the difference between a bad relationship and a bad reality.
And it explains why the anger aimed at you is almost a relief by comparison. At least it's aimed. At least it has a target. The diffuse version has no target, so it hits everything.
Third piece. The asymmetry of self-evidence. Why don't you use your own decent behavior as counter-evidence?
Two reasons, and the second one is the interesting one. The first is that the hostile person's model already accounts for you. You're the exception. Or you're someone who just hasn't been caught yet. Or you're naive and you'll learn. The model has an answer for every counterexample, which is what makes it unfalsifiable.
And the second reason?
Your own behavior doesn't feel like data. It feels like just you. You don't experience your own decency as a data point about humanity. You experience it as your personality. It's not evidence, it's identity. So it never gets entered into the ledger. Meanwhile the hostile person's claims about others feel like revelations about reality. They arrive with the texture of insight.
So you've got one stream of claims that feels like truth, and one stream of your own behavior that doesn't feel like anything.
And the claims win, because claims are about the world and your behavior is just about you.
Fourth piece. The osmosis. This is the one I find unsettling. Because it's not a conscious belief adoption. Nobody sits down and decides that humanity is a cesspool. It's gradual. It's pre-verbal. And it's usually only visible in retrospect.
You notice you've become reserved. Guarded. You screen your own sentences before you say them. You don't reach out. You don't share. And you attribute it to your personality. You think, I've always been a private person. I'm just introverted. I like my own company.
And the actual cause is sitting three feet away, narrating humanity as a sewer.
For years.
And here's where I want to bring in the example Daniel gave, because it's the cleanest illustration of the whole thing. The kind family member. You know this person is kind. You've watched them be kind. You've been on the receiving end of it. And they get relentlessly attacked. And the hostile person's framing is, you don't see what they're really like. Which plants a seed. Not the seed that the family member is bad. The seed that your own perception is naive. That's the actual injury. You don't start believing the family member is cruel. You start believing you can't tell.
That's the epistemic hit.
That's the whole thing. The hostile person doesn't just claim other people are bad. They claim that your read on people is unreliable. And they claim it by implication, every time they correct your perception. You thought she was kind. You're not seeing it. You thought he was helpful. You're not seeing it.
And the second example is the favor.
The favor. Somebody does this person a genuine favor. Nine things done well. One thing done imperfectly. And the one thing becomes the whole story. It becomes proof of incompetence, or proof of hidden motive. And you watch this pattern repeat across dozens of people, over years.
And the standard gets internalized.
The standard gets internalized. That's the piece I want to land. The hostile person's standard for human behavior is impossibly high. Everyone fails it. And you don't notice that you've adopted the standard. You just notice that you're disappointed in people. Constantly. In a way you can't quite explain.
And the standard is unmeetable by design. That's the function. If the standard were meetable, the model would collapse. So the standard has to be set where no one clears it. Nine good deeds and one flaw. The flaw is the data.
So you end up walking around with a test nobody passes, and you think you're just perceptive.
You think you're a good judge of character. That's the cruelest part. The distortion disguises itself as discernment.
So that's the mechanism. But here's where it gets really troubling. Why doesn't your own experience of being a decent person serve as a corrective? We touched it. I want to push on it.
Push where?
Toward the consequences. Because the mechanism doesn't just change what you believe. It changes what you do. And what you do then feeds the belief.
The behavioral consequence. Once you believe everyone is cruel behind closed doors, vulnerability feels like handing someone a weapon.
It doesn't just feel like it. In the model, it is. If the model is true, opening up is not brave, it's negligent. You're giving away the thing that protects you.
So you stop sharing. You stop reaching out. You stop letting people in. And here's the loop. The more reserved you become, the fewer positive relational experiences you have. The fewer positive experiences, the more the hostile person's model seems confirmed. The distortion becomes self-reinforcing.
It's a self-fulfilling prophecy with a maintenance crew.
And the maintenance crew is you, at this point. Which is the part that's hard to sit with. The hostile person installed it, but by year three, you're running it.
You're the one enforcing the model. You're the one declining the invitation, screening the message, deciding not to say the true thing. And every time you do, you generate one more piece of evidence that people aren't safe. Except the evidence is manufactured by your own avoidance.
Which means the model never gets disconfirmed. Because disconfirmation requires contact. And contact is what the model prevents.
There's a third consequence I want to name, and it's the one I think is the deepest injury. It's not that you stop trusting others. It's that you stop trusting your own judgment.
The loss of confidence in your own social cognition.
If the hostile person saw through everyone, and you didn't, then your perceptions are unreliable. That's the logic. And it's airtight, given the premise. So you stop trusting your read on people. You second-guess your own warmth. You wonder if you're being naive every time you like someone.
It's an epistemic injury. That's the right term. It's not an emotional wound, it's a wound to your capacity to know things. Your instrument for reading people has been recalibrated by someone whose instrument was broken.
And you can't just decide to trust your instrument again. That's not how it works.
No. And this is where I want to bring in what actually helps, because the episode can't just be a diagnosis. The recovery question. How do you restore faith in basic decency?
And I want to be clear about what that's not. It's not becoming naive again. It's not swinging to the opposite pole and deciding everyone is wonderful. That's just a different distortion.
It's recalibrating the model based on better data. That's the whole project. The model was built from a poisoned data set. The fix is a better data set. Which sounds simple and is not, for reasons we've just spent twenty minutes establishing.
The first step is recognizing the data set was poisoned. That the hostile person's worldview wasn't insight. It was pathology. That's not a small step. That's the hinge the whole recovery turns on.
And it's hard because the worldview felt like realism. It felt like the sober, adult, clear-eyed view of how the world works. Giving that up feels like choosing to be a fool.
Which is exactly why it survives.
Exactly why it survives. So the recognition is: I was not being realistic. I was being colonized. And I mistook the colonization for maturity.
Say more about the data set part. Because I think there's a practical thing in there.
The practical thing is that disconfirming experiences have to be noticed and registered as data. Small, low-stakes interactions where someone is kind without motive. Reliable without drama. Honest without cruelty. Those happen constantly. And a distorted model filters them out. You don't see them. They don't count.
So you have to actively look for them.
You have to actively look for them, and you have to count them. That's the unglamorous work. It's not a revelation. It's bookkeeping. Somebody held the door and meant it. Somebody remembered the thing you mentioned once. Somebody told you a hard truth gently, which is the exact thing the model says doesn't exist.
Each one is a data point.
Each one is a data point. And you're rebuilding a sample. Deliberately. Because the old sample was curated by someone with an agenda, and the new one has to be curated by you.
The second piece is distance. The hostile person's voice needs to stop being the loudest one in your head.
That's not about forgiveness. I want to be careful here. This isn't a forgiveness segment. It's about stopping the input.
Stopping the input. Because the model is maintained by reinforcement. Every conversation, every installment of the exhibit, is a dose. And you can't rebuild a model while someone is actively dosing you against it.
Physical distance helps. Emotional distance helps. Sometimes the only available move is to stop taking the calls, or to keep the contact so structured that the exhibit doesn't get delivered. But the principle is: reduce the reinforcement.
The third piece is meaning-making. And this is where the recovery literature is actually useful.
Posttraumatic growth research. The finding is that people who can construct a coherent narrative of what happened recover better than people who just try to move on. The narrative matters. Not the events. The story you tell about the events.
The story is something like: I was exposed to a distorted worldview, and I internalized it, and it wasn't true.
That's the shape. And notice what that sentence does. It names the mechanism. It assigns the distortion to its source. And it asserts the falsity. Three moves. And those three moves are the cognitive antidote to faulty extrapolation, because they take the one pathological person and put them back where they belong. In a sample of one.
They un-generalize.
That's the whole game. The distortion was a generalization. The recovery is a de-generalization. Same mechanism, run backward.
The fourth piece is the one that's hardest to hear. Selective vulnerability as practice.
You don't restore trust by trusting everyone. That's the naive pole. You restore it by taking small risks with safe people, and having those risks pay off.
Each payoff is a new data point.
Each payoff is a new data point. That overwrites the old model. Slowly. By experience. Which is the only thing that ever updates an internal working model in the first place.
The whole recovery is just running the original mechanism in reverse.
That's exactly what it is. The model was built by experience. It gets rebuilt by experience. There's no shortcut through insight. You can understand the mechanism perfectly and still not have moved an inch, because understanding isn't data.
That's the hard truth. This takes time. The distorted model was built over years. It doesn't come apart in weeks.
It requires something the hostile person never gave you. Evidence that people can be trusted. Which means it requires other people. Which means it requires the exact thing the model says is dangerous.
You have to walk into the thing you're afraid of, repeatedly, and let it be fine.
Notice that it was fine. And count it.
There's a wrinkle I want to get to before we bring Hilbert in, because I think it complicates the picture in an interesting way.
Hilbert: Twelve dollars.
Sorry?
Hilbert: Twelve dollars. That's what the manual cost. A bound copy of the loss prevention procedures for a department store chain. I still have it. The store's gone. The manual's on a shelf.
You worked loss prevention.
Hilbert: Two years. Late eighties. You sit in a room with monitors and you watch people. That's the job. You watch hands. You watch the way somebody's eyes move when they pass the display. You learn the tells. And after about six months, you stop seeing shoppers. You see suspects who haven't done anything yet.
That's the same mechanism.
Hilbert: It's the same mechanism. The job trains you. By exposure. Eight hours a day of watching people, and the ones you're watching for are the ones who are taking something. So your sample is thieves. Your sample is people who are about to be thieves. And the brain does what it does. It generalizes.
The distortion doesn't require a hostile person.
Hilbert: It doesn't require a person at all. It can come from a role. A job. A system that pays you to see everyone as a threat. The input is the same. Somebody is feeding you a curated exhibit of human behavior, and the exhibit is all worst cases.
The difference is you can quit a job.
Hilbert: You can quit a job. That's the difference. I quit. I walked into a grocery store about a month after I left, off shift, no reason to be watching anybody. And I caught myself. Doing the survey. Where are the cameras. Who's watching the aisles. Which guy by the door is carrying something he didn't pay for. I stood there in the produce section doing inventory on strangers.
How long did that last?
Hilbert: Years. It faded. It didn't disappear. I still can't walk into a store without noticing where the cameras are. It's automatic. I don't do anything with it. I just see them. It's like knowing where the exits are. It's just there.
The training leaves a mark.
Hilbert: It leaves a mark. You don't get the room back the way it was. You get to a point where the mark isn't running the show. That's the most you get. And with a family member, you can't quit. That's the part I don't have an answer for. I could hand in the badge. You can't hand in a brother.
That's the asymmetry.
Hilbert: The manual's still on the shelf. I don't know why I kept it. It's not useful. But it's the paperwork from the job that rewired me, so it stays.
What you're describing is that recovery isn't erasure.
Hilbert: No. It's management. You don't forget the training. You stop letting it drive.
The goal isn't to get back to the person you were before. It's to become someone who has the distortion and doesn't obey it.
Hilbert: That's about right. And the guy who trained me, Moretti, he'd have said the same thing. He was on the monitors for thirty years. He retired and moved to a place with no stores. I asked him about it once. He said he just wanted to walk somewhere without counting the cameras.
The distorted model doesn't fully come apart. It becomes something you carry and manage.
Which reframes the whole recovery question. It's not about restoring the old model. It's about building a new relationship with the damaged one.
Hilbert: That's it. You keep the manual. You just stop working the floor.
The closing question, then. If the distortion was built by osmosis, gradual exposure, years of it, can it be undone the same way? Through gradual exposure to a different kind of person? Or does it need something more deliberate than that?
I think it needs both. The osmosis is how it got in, so osmosis is how it comes out, but the osmosis has to be pointed. You have to put yourself in the stream on purpose, because the default is avoidance, and avoidance is the thing maintaining the model.
There's a version of this happening at scale right now that I keep thinking about. The algorithmically curated outrage. The feeds that are optimized to show you people lying, cheating, hiding something. It's the same exhibit. Curated by a system instead of a person, delivered continuously, and the sample is all worst cases.
The hostile person in your life as a microcosm of a cultural pattern. That's the uncomfortable version.
It is. And the antidote at scale is the same as the antidote at home. The willingness to keep collecting data. To let new experiences count, even when the old voice says they don't.
Which is harder than it sounds, because the old voice is very confident.
The old voice is always confident. That's what makes it the old voice.
If this one landed for you, a rating or a review helps other people find the show.
Thanks as always to Hilbert Flumingtop, who produces this and who kept the manual.
This has been My Weird Prompts.
Email us at show at my weird prompts dot com. We'll be back soon.
