{
  "$type": "site.standard.document",
  "canonicalUrl": "https://serpentsquiggles.neocities.org//posts/posts/newcomb-hypnosis",
  "path": "/posts/posts/newcomb-hypnosis",
  "publishedAt": "2026-06-14T00:00:00.000Z",
  "site": "at://did:plc:ivoe7cntxuy6at7uzmxzs2ft/site.standard.publication/3mfk6cpprzt2t",
  "tags": [
    "ramble"
  ],
  "textContent": "; related\n: - [](basilisk.html)\n  - Functional Decision Theory\n\nIntroduction\n\nFirst, let me reintroduce Newcomb's Problem.\n\nFor this thought experiment, imagine there is a predictor capable of\nobserving human behavior and correctly deducing how one will behave in\nthe future. Maybe this predictor has psychic powers, or is really good\nat reading body language. I often see it described as a\nsuperintelligent AI system. More on that later.\n\nThis predictor already knows enough about you and has already\npredicted what answer you'll give the following problem. Having made\nits judgment, it places down two cardboard boxes. You get to chose\nwhich ones to open, and you keep whatever is inside.\n\nThe box on the right always contains a thousand dollars. The box on\nthe left contains a million dollars, but only if it predicted you'd\nchose only that box.\n\nOnce it explains the situation, it walks away. It can no longer affect\nwhats inside the box, the choice is yours.\n\nAll that matters is whether you open the one box on the left. (If you\nopen the right box, you might as well open the left.) Thus, the two\npossible answers you can give are \"one box\" or \"two box.\"\n\nThis is a very contentious decision problem. Philosophers couldn't\ndecide on the right answer (and still can't). It's been decades, and a\nlot of different theories have been put forth. The mainstream falls\ninto roughly two camps:\n\nOne camp argues: the predictor has already left, so the left box\neither has the money or doesn't. If it has the money, you'd be better\noff picking both. If it doesn't have the money, you'd be better off\npicking both. Therefore one should pick both.\n\nUnfortunately, this sort of person only gets a thousand dollars, by\npremise.\n\nWhich sucks, but what can you do? It isn't your fault for being\nrational! We have created a question that punishes reasoning correctly,\nand reasoning correctly gives the wrong answer, surprised pikachu face.\nPerfect predictors don't actually exist in real life, who cares.\n\nThe other camp argues: everyone who picks one box gets a million\ndollars, and everyone who picks two boxes gets a thousand dollars.\nPicking one box is evidence of a better outcome, so I pick one box.\n\nAnd it has a million dollars!\n\nThe first camp sees this, and says \"you idiot! if you had picked both\nboxes, you'd be a thousand dollars richer right now!\"\n\nAnd they reply: \"if i had picked both boxes, I wouldn't have the\nmillion,\" and they proceed to laugh all the way to the bank.\n\nBut it gets weirder than this. Consider a modified scenario: instead\nof cardboard, the boxes are made out of transparent plastic. You can\nplainly see whether the other box has a million or not.\n\nThis doesn't change the first camp's answer, though they'll probably\nnot bother opening the empty plastic bin (which is all they'll ever\nsee.)\n\nBut I'm pretty sure the second camp will continue to only open one\nbox, leaving the thousand dollars sitting there (no one ever got both)\nbut I'm not actually 100% sure. TBH, I don't think about the second\ncamp often.\n\nBut why not? you might think those girls are on to something, if they\ncan get the right answer here.\n\nBut consider a modified scenario. The predictor is in fact a fraud.\nSure, they're right most of the time, but after thousands of\nexperiments, a researcher figured out the trick being used to predict\npeople. You see, the predictor first collects a stray hair and\nsequences your DNA. It turns out there's a strong correlation between\ncertain genes and philosophical inclinations, so this procedure is\npretty effective at predicting people's answers.\n\nOf course, this explanation is all well and good, but you don't know\nif you have the two-boxing gene or not. How would knowing the mechanism\nchange your decision?\n\nThe first camp continue to pick two boxes, of course. Whether the\nsecond camp picks one depends on how tight the correlation is. If it's\ngreater than around 99.9%, then they know one boxers get more money on\naverage, so it's better evidence of winning big than knowing they pick\ntwo.\n\n(If you're really into the weeds of these arguments, you may\nrecognize the above scenario as one that's usually presented in a more\nconfusing way, by imagining a world where smoking doesn't cause\ncancer.)\n\nAnyway I think there's an important difference between these two\nproblems. The best way to highlight the difference is to ask what you\nwould do if you had heard some predictor was going around springing\nthis problem on people before the predictor learned anything about\nyou.\n\nThe first camp, in the first problem only, would reason as like\nthis: if there's someone who can predict my behavior, it's best for me\nif they predict I'll pick one box.\n\nOne extreme way to accomplish this by giving some guy a gun and paying\nhim to shoot you if you don't pick one box. Now when you meet the\npredictor, it will predict that you will be afraid of getting shoot\nand chose one box, but since it only cares what you chose, it'll stash\nthe million and walk away. Yet now if you try to two box, you get\nshot, so you one box still.\n\nIn general, this is something called a precomittment. It's like the\ncaptain who ties himself to the mast so he can hear the sirens without\nbeing tempted to climb overboard. The details of how you ensure you're\ncommitted aren't that important, as long as they ensure you'll make a\ncertain decision even when you later want to chose differently.\n\nBut consider what you'd do in the second scenario. If the predictor\nis scanning your DNA for a certain gene, then you could scan your dna\nfirst and see if you have the gene. Armed with this knowledge, you can\nchose two boxes even if you have the one-box gene, because you know\nit's not a perfect predictor (the gene influences but doesn't\ndetermine your choice), and any lucky outlier who has the gene and two\nboxed anyway can get both piles of money!\n\nImportantly, in the second scenario, there's no sense in\n\"precommiting\" to take one-box, not if the prediction is already\ndetermined by your genes.\n\nSubjunctive Dependence\n\nSo there seems to be a meaningful difference between \"true prediction\"\nand statistical correlations that might be wrong. But it's not enough\nthat the predictor is sometimes wrong, because consider what happens\nif we modified the scenario so that the predictor remains\n(technically) perfect, but to make things slightly interesting, it\nalso rolls dice so that 0.1% of the time, it ignores its own\nprediction and lets the dice decide whether to put the million in or\nnot. Thus, we get'd the same statistics as the genetic version: strong\nbut imperfect correlation.\n\nAnd this doesn't change anything! The first camp would still take the\ncommitment to one-box because an almost-certain chance of a million is\nbetter than one thousand and an almost-negligible chance of a million.\n\nIf a tree falls in a forest and no one is around to hear it, does it\nstill make a sound? The truth is that this question doesn't make\nsense. The tree creates physics of sound, but not perception. It's\npure semantics: you have a gut feeling that one or the other is the\ndefinition of sound, and the resulting dispute is a merely verbal\ndisagreement.\n\nA similar semantics problem lurks in Newcomb's Problem.\n\nConsider the following explanation of its prediction capabilities: you\nare currently inside the matrix. This world is just a computer\nprogram, a simulation that can be rewinded or restarted. You have no\nidea if you're in the matrix or the real world. Maybe there isn't a\nreal world.\n\nOne simulation of you is presented the two boxes and asked to decide.\nThe simulation ends once you've made your choice. Based on this data,\nthe predictor poses the same problem to another version of you. Maybe\nit's the real you, maybe it's another simulation but one which won't\nend prematurely.\n\nThe point is, with this framing, I think the first camp would treat\nthe scenario differently. Because it makes clear what's happening:\nyou're deciding how the simulation acts, and you're deciding how the\nreal you acts, and you can't tell the difference, by premise.\n\nNow imagine right before you make your decision, the predictive AI\nsays, \"Oh, you're the real version BTW.\" The simulation was not told\nthis, so now the AI's data can't actually determine how this might\nchanges your behavior.\n\nOut of curiosity, the first camp might stop and try to guess how the\nsimulated copy might've acted, but it doesn't matter, what's done is\ndone, and now they can reap all the rewards.\n\n(Another way to get the same results: imagine you are a demon who\ntakes possession of the test subject right before they make their\ndecision, and the predictor can't predict demons.)\n\nPoint is, this is the actual question the first camp thinks they were\nanswering all along.\n\nIt only makes sense to reason as if the prediction \"already happened\"\nif you have ironclad certainty you're not in the matrix right now.\n\nBut if you don't like the idea of simulated copies and the identity\nquestions that raises, there's a bunch of others\nroutes to the same insight. Imagine the predictor is actually a time\ntraveler, and whenever it \"mispredicts\", it goes back in time to try a\ndifferent guess. Imagine the predictor is actually a hypnotist, able\nto put you into a trance where you can forget any memory. Once you\nchose, your memory gets erased, and you chose again.\n\nThe point is, in these scenarios, it's clear that it hasn't already\nhappened, and your choice might actually change whether the \"real you\"\ngets a favorable or unfavorable situation. So you pick the choice\nthat's favorable in both scenarios, which is one boxing.\n\nIf the predictor is unreliable, then you still aren't certain you\naren't in the matrix. If it's 75% accurate, then it's\nindistinguishable from having been in in the matrix 37.5% of the time.\n(or 25% of the time, the time travel or hypnotism fails, etc.)\n\nRemember how I said there were two camps? There's three actually.\n\nA third group of weirdos, who aren't taken very seriously by academic\nphilosoph",
  "title": "Nuances of Newcomb's Problem"
}