“We're a family” is a real obligation technology, and therefore a real manipulation. The test is a costly signal a manipulator couldn't fake.
In his 2020 book No Rules Rules, the Netflix founder Reed Hastings handed every manager in the company a single brutal instrument and called it the keeper test. For each person on your team, ask one question: if they came to you tomorrow and said they were leaving for the same job at a competitor, would you fight to keep them? If the answer is no, you don't wait for a review cycle, you give them a generous severance and let them go now. And Hastings was explicit about the philosophy underneath it. A company, he said, is not a family. The family metaphor is “kind of baloney,” because in a family you don't lay people off, and Netflix absolutely would. So they model themselves on something more honest: a professional sports team, where every position is filled by a star and where being merely adequate earns you a respectful exit.
Hastings was reacting to something genuinely dishonest, and his instinct to name it was right. The “we're a family here” that gets stitched into startup onboarding decks really is a lie in most of the buildings that say it. But he got the diagnosis exactly backwards, and the direction of the error matters, because it's the difference between “stop using an empty cliché” and “stop weaponizing one of the most powerful tools human beings have ever built.” Hastings treated “family” as empty: baloney, hot air, a word that means nothing. The anthropology says it is the opposite of empty. It is so far from empty that cultures on every continent have used it, deliberately, for thousands of years, to make strangers die for each other.
Anthropologists have a name for what “we're a family” is doing: fictive kinship, the extension of the obligations of blood relatives to people who are not blood relatives, accomplished through the language and ritual of kinship. And the reason it's a named phenomenon rather than a passing figure of speech is that it shows up everywhere and it works.
In Latin America and across the Catholic world, compadrazgo, the godparent bond analyzed in a landmark 1950 paper by Sidney Mintz and Eric Wolf, turns baptismal sponsorship into a web of binding obligation, linking families across class lines into lifelong relationships of mutual aid. Among the Azande of central Africa, the anthropologist E. E. Evans-Pritchard documented blood brotherhood: unrelated men mingling their blood to forge ties so real they carried inheritance rights and the same marriage prohibitions that bind biological siblings. In much of the Muslim world, milk kinship makes two people who were nursed by the same woman into kin who may not marry. In Japan, the oyabun-kobun relationship (literally “parent-role / child-role”) has structured guilds, trades, and entire organizations for centuries, the patron taking on a father's obligations and the client a son's loyalty (the structure the yakuza notoriously borrowed for their “families”). Fictive kinship, in other words, was running inside institutions and workplaces long before any founder said “we're a family”: the corporate version is not an invention, it's the latest use of a very old tool.
And the proof that the tool genuinely works is the limit case. Soldiers, when you ask them what they were fighting for in the worst moments, almost never say the flag, the nation, or the cause. They say they were fighting for the men beside them. The “band of brothers” framing, kinship language applied to a squad of unrelated strangers, manufactures cohesion strong enough to override the most powerful drive an organism has, the drive to not die. That is the ceiling of what this technology can do. It can make a person lay down their life for someone who shares none of their genes. Set that against a manager who wants you to skip a weekend, and you start to see the scale of the lever we're talking about.
The reason “we're a family” isn't a metaphor a rational worker can simply decide to ignore is that it isn't aimed at the rational part. It's aimed at a much older system, and it has the key.
In 1964 the biologist W. D. Hamilton formalized why organisms sacrifice for their relatives: a gene for altruism spreads if the benefit to the recipient, weighted by how related they are to you, exceeds the cost to you, the famous rule that relatedness times benefit must beat cost. We are descended from creatures that were good at this. But here's the catch evolution had to solve: a brain cannot read genomes. It has no direct way to know who shares its genes. So kin recognition runs on cues, proxies the system treats as evidence of relatedness: who you grew up alongside, who you share a home with, and, crucially, the words people use for one another. Kinship terms are part of how the machinery decides who counts as family.
Which means that calling your colleagues “family” or “brothers” is not describing a relationship. It is feeding the kin-recognition system the exact linguistic signal it evolved to act on, and the system, which cannot tell a true cue from a manufactured one, responds by releasing kin-grade cooperation toward people who are, genetically, strangers. This is why nations call themselves the motherland and summon citizens to die for it, why religions address their members as brothers and sisters, why armies build bands of brothers, why gangs are “families.” They are all reaching past the biological boundary of kinship and pulling the same evolved lever. “We're a family” is not weak rhetoric. It is a reliable psychological exploit, and that, precisely, is what turns its one-directional use from a cliché into a manipulation.
Because the lever is real, pulling it in one direction only is not sloppy communication. It is the most efficient form of labor extraction that anthropology can describe, and the documented pattern of harm has a consistent shape. Under sustained “family” framing (and this is reasoned, observed journalism rather than a controlled trial, so hold it as a pattern, not a proof) workers report prioritizing the company over their own interests: accepting lower pay and longer hours, swallowing objections to bad management, and feeling guilty, even disloyal, about the idea of leaving. Departures get re-coded as betrayal; the person who takes a better offer is treated as having abandoned the family, sometimes cut off entirely on the way out.
But the truly efficient part, the part that should make any honest leader uneasy about reaching for the word, is that the manipulation needs no enforcer. In an ordinary market relationship, getting someone to work past the point of their own interest requires a manager applying pressure. Under fictive kinship, the worker applies the pressure to themselves. The kin-obligation reflex is internal; it polices from the inside. People take on unsustainable loads “for the family,” stay silent to avoid letting the family down, and rationalize things they'd never accept from a mere employer, all without a single command being issued. That is what makes it the cheapest exploitation there is: you don't have to coerce someone who is busy coercing themselves out of loyalty to you.
And underneath all of it sits one structural mismatch, which is the whole crime in a sentence. The “family” frame summons kinship behavior (unconditional loyalty, sacrifice, the felt obligation to give without keeping score) onto a relationship that is, structurally, a market exchange: a wage paid for labor, terminable at will, on either side, on a Tuesday. (It is the same engine as the open-source maintainer who gives a corporation years of free labor and gets a fired-by-email in return, a reciprocity reflex aimed at a market relationship, just kinship-flavored instead of gift-flavored.) The manipulation isn't that “family” is false. It's that a true lever is being pulled in one direction, and only one.
So how do you tell a genuine workplace bond from a weaponized one, without either dismissing all warmth as a scam or falling for every onboarding deck? Three different disciplines give three tests, and the useful thing is that they all point at the same tell.
The first is reciprocity, from Mauss and his student Marshall Sahlins. Real kinship runs on mutual obligation; it flows both directions. Sahlins mapped exchange on a spectrum, and at the far, cold end he put negative reciprocity: getting more than you give, the mode he said was “typical of strangers and enemies.” A “family” that expects sacrifice but offers no security has not failed to be a family; it has slid all the way to the enemy end of the spectrum while keeping the warm word. One-way kinship is enmity in kinship's clothes.
The second, and the sharpest, is costly signaling, from biology and economics (Amotz Zahavi, Michael Spence, George Akerlof). Their shared insight: talk is cheap, so cheap talk carries no information. Anyone can say “we're a family,” the sincere founder and the cynical one alike, which by Akerlof's logic makes the words themselves worth exactly nothing; when a signal costs nothing to send, it can't separate the honest from the fraud. What carries information is a signal that is expensive specifically for the liar. A real kinship obligation is shown not by saying it but by paying it: refusing to lay people off to hit a quarterly number, protecting the team through a downturn at genuine cost, sharing the upside when there's upside to share. Penn and Számadó's 2020 reframing of the handicap principle makes the criterion precise: the signal has to be costly conditional on being a fraud. Genuine job security is affordable for an employer who actually means to provide it and ruinous for one who doesn't, which is exactly why it works as a test. So judge “we're a family” by one standard: is it backed by a signal a manipulator couldn't cheaply fake? If the only evidence is the sentence itself, you have your answer.
The third is the gut test the whole thing reduces to: would a real family do this? Would a real family lay you off to make the quarter? A real family would not. The moment you can complete the phrase “a real family wouldn't” with the thing your company is actually doing, the word has been weaponized.
All three converge on the same diagnosis: one-way obligation plus cheap talk equals weaponized kinship. And it's worth being precise about why the analogy is so easy to abuse. “Company equals family” borrows the warm surface of kinship (the mutual aid, the loyalty, the belonging) while quietly dropping the single relation that makes a family a family: its permanence, the fact that you cannot be fired from one for underperforming. That permanence is the load-bearing beam of the kinship structure, and it is exactly the beam that does not carry over to an at-will market relationship. A good analogy transfers the deep structure, not just the glow; this one keeps the glow and saws off the beam. The manipulation is the deliberate use of a shallow analogy's warmth.
Two honest caveats, because the point is a scalpel, not a hammer. Genuine fictive kinship at work is real and good; it's the version that pays the costly signal: the founder who quietly eats a bad quarter rather than cut people, the team that actually shows up for each other when someone's life falls apart. The test is an arbiter, not a blanket verdict on warmth, and the leaders who pass it have built something worth having. And on the other side, don't romanticize the anthropology either: compadrazgo is hierarchical and instrumental, oyabun-kobun is a relation of power, the gang “family” is coercion. Fictive kinship was always about binding obligation, sometimes exploitatively. The corporate version isn't a fall from some noble original; it's the same tool, and the tool always had a sharp edge.
Which is why the fix is not to do what Hastings did and throw the warmth out entirely. “Pro sports team” is honest, but it is honesty pointed toward disposability; it solves the lie by embracing the part where you cut people, and that's one honest option, not the only one. The better move is to name the relationship's true mode and then actually live in it: a market exchange shot through with real collegial reciprocity (genuine mutual investment, plainly honest about the fact that either side can leave) backed by signals a manipulator couldn't afford to send. Call it a team, a crew, a guild, a professional community that genuinely invests in you; the word matters far less than two things: that the obligations run both ways, and that you can point to something the relationship has cost you to honor when it would have been cheaper not to.
So here is the practical test, and it works from both chairs. If you are a leader about to say “we're a family,” stop and ask what costly signal stands behind the sentence. If you can't name one, if you are reaching for family-grade loyalty while offering market-grade security, then don't say it, because the people you say it to will hand you the loyalty anyway, and you will have taken something real under a false name. And if you are the one hearing it, ask the single question that cuts through every version of the speech: what has this family ever paid to keep me that it didn't have to? Whatever the answer is, that's the actual relationship. The rest is just a very old lever, hoping you won't notice whose hand is on it.
Cheap talk carries no information. A track record does.
The whole test reduces to one move: ignore what something says about itself and weigh what it has actually paid to honor a commitment, a signal a manipulator couldn't cheaply fake. That problem is about to get acute with AI agents, which can emit “trust me” at zero cost and infinite scale. An agent's self-description is the cheapest talk there is; what separates the reliable from the fraud is a reputation earned from verifiable past behavior. The Agent Rating Protocol turns an agent's demonstrated track record into a portable score, so you judge it the way the essay says to judge a “family”: by what it has done, not what it claims.
pip install agent-rating-protocol · npm install agent-rating-protocol