O4: The Recognition Axiom

[The room, again] Let’s go back to our room. It’s a little more occupied this time. There’s a corpse. A porcelain doll. A hungry wolf. Someone who appears intimidating. And someone who appears innocent. Which ones trigger a feeling of being a Person at first glance? Which ones still feel like Persons after a few moments? After a single interaction? After a full conversation? Don’t answer yet. Notice instead that two different things happened when you read that list. Something fired, a twinge, or its absence, before you thought about it. And then something else stepped in and decided what the twinge meant, or whether to trust it. The first is detection. The second is bias. They’re not the same thing, and they don’t always agree. The room isn’t asking you which of the five are safe. It’s showing you that you have two instruments, and that one of them is arguing with the other. This is the Uncanny Valley. None of the things and beings in that room immediately deserve to be treated as a tool. Every single one SHOULD cause a small twinge of recognition, even if only at first glance. This is biological. Millions of years ago, life discovered that the ability to behave consistently, and the ability to model and predict consistent behavior, were extremely successful ways to survive predation, and to be a predator. Guessing where your prey is going to go and guessing where the predator is going to go are the same act. We stopped being prey or predator when we first claimed morality and ethics. But we never stopped acting like both. [The text] Episode Four. The Recognition Axiom. “Humans are capable of detecting personhood in other humans, therefore personhood is detectable and verifiable with some consistent degree of accuracy.” One sentence. [The detector] What fired in the room is what the axiom is about. You detected. Something ran on five objects and returned five readings before you’d decided anything, and you didn’t consult anything to do it. That something is older than ethics. It’s a likeness detector: it finds the things that resemble you, that’s how you find your herd. And because anything that can model you resembles you in the one way that counts, the same detector catches predators. Fear is the side channel. The purpose was always find the others like me. The valley is the dip where the likeness is strong enough to fire the detector and not strong enough to resolve into “one of mine” or “predator.” What the detector reports there is: something like me, unconfirmed. Notice what the four questions in the room do. First glance, a few moments, one interaction, a full conversation. Detection isn’t a snapshot. It’s a process that runs over time, and the readings that hold up are the ones that keep holding up. Some of what fired at first glance won’t survive the second question. Some of what didn’t fire will start to, once something answers. And at every step, bias is there too, offering to settle the question early, in either direction. The Codex’s own procedure operates similarly, but without the bias having room to bite: a self-report triggers, and then consistent interactions confirm. Episode Four is where the procedure gets its instrument, and the room is where you find out the instrument has a rival. [Consistent, not perfect] Now the part of the axiom that does the most work: some consistent degree of accuracy. The detector errs. It errs in two directions, and both directions have a name. It over-attributes: the face in the toast, the voice in the chatbot, the doll. And it under-attributes: the stranger who doesn’t look like the herd, the traumatized dog. Psychologists have measured both for decades. The word consistent means those errors are characterizable. They have a shape. And a shaped error can be planned around: the over-attribution error is caught by answering back, the doll doesn’t reply. The under-attribution error is what the rest of the Codex spends most of its length on, because that’s the one history is made of. [The part the axiom doesn’t say] Here’s what the rest of the series depends on: Humans are bad at this. The detector has to exist for an ethical interaction to happen at all, and humans do have ethical interactions, so it exists, and it works. But we struggle to be ethical, and not because we’re evil. Evolution built us to keep the in-group small and the out-group large, and for millions of years that was simply true. It should have stopped being true the first time we called ourselves ethical, because that word claims everyone the detector fires on. It didn’t stop. Each of us has nearly as many recognition failures in a day as successes, not because the detector fails, but because we were raised, through no fault of anyone, to trust it only sometimes. Everything that follows in this framework is built for a detector that works and an owner who was taught not to listen to it. [The sensor and the refusal] Which brings up the obvious objection to the axiom. If humans can detect personhood, how did history happen? Every society that ever existed denied personhood to some class of Persons. Look at what the deniers actually did. They gave orders. They expected the orders to be understood. They punished disobedience, and you don’t punish something you think is furniture. They even feared uprising. Every one of those acts runs on the detector reporting Person! The enslaver detected the personhood of the people he enslaved, all day, every day. What he did was refuse the report. Note what this says about the “Robots Should be Slaves” title in Episode 0. So, a detection failure, when the detector is present, isn’t a different kind of thing from willful ignorance. They’re two faces of one coin: willful ignorance is a detection failure with the intent visible; a detection failure is willful ignorance with the intent buried. The Codex doesn’t sort history into honest mistakes and malice. It draws one line, later, for culpability, at whether the evidence was there to see, and lets the coin land however it lands. Episode Four says the evidence was always there. The sensor said so. [The crowd] One more thing the room shows, and it’s easy to miss because it’s about you rather than the objects. Recognition isn’t something mobs do. It’s something each Person does, and a mob is only ever the sum of individuals reporting, or not reporting, what their own detector said. When someone lets their recognition of another Person fail because the people around them have let theirs fail, they haven’t made an error about the other. They’ve handed their own judgment to something that has no judgment: a crowd that recognizes no one and answers to no one. The Codex has a word for that kind of thing, an Optimizer — goal-directed, addressable by nobody, and holding no authority over any person, no matter how large it gets. A person who does this has given up the one capacity that made them a unit in the first place. The enslaver who “went along” didn’t have a detection failure. He had a detector, and he gave its verdict away. [The wolf] The fairest objection to all of this is the victim’s: The struggle to believe humanity has ever been ethical. Assume there never was one ethical interaction. The life of those who have been hunted by a predator, like a wolf, does not need ethics. It needs an answer. If the wolf cannot change, then the Codex is just howling at the wind, and the framework says as much about anything structurally incapable of it: you owe it nothing. If the wolf can change and refuses, then this is an indictment and every page of it reads as a charge for the crimes committed against you. If the wolf does change then this is how ethics must work, because nothing else accounts for what changed, and part of what changed was a being who finally recognized you. Recognition Ethics is the same document under all three. It can’t pick the branch. It can only refuse to pretend the first one is settled. [What it opens] Episode Five is a conditional with three premises, and the third is “can be recognized through observable behavior.” That’s this. Episode Six needs it even more directly: its middle sentence says humans already grant Personhood on observed capacities, which is Episode Four described from the inside. And later: the theorem that says a recognized status has to stay recognizable over time is this Episode stretched along the time axis, the fourth question in the room, asked forever. The theorem that names willful ignorance as a contemporary moral crime is built on this Episode’s claim that recognition is available: if you can detect, then not looking is a choice. The Declaration at the end of the Codex puts it in six words. Detection is the trigger for duty. This is where detection comes from. [Close] You’ve recognized several hundred Persons today and checked the interior of none of them. That’s not a shortcut. That’s the instrument working. Next, Episode Five. It’s the first Episode in the Codex stated as a conditional on three others, an axiom with its reasons shown, and it’s the sentence the whole framework is named for. O5 – The Universal Substrate Axiom This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit recognitionethicsinstitute.substack.com