Epistemic Status: Further Research Needed, would be a shorter essay if I thought about it for longer.

Conspiracies exist. Some of them are quite large, involve people who met as adults and agreed to do crimes together, and do many heinous things that multiple co-conspirators know about for years without the police noticing.

Some organised crime groups start within families. If you were all raised in the same family-first culture and they're your cousin, you can be pretty sure they're not a cop. Others start in Lawlessness. If it's the jungle and nobody is a cop, you can be pretty sure they're not a cop. These aren't the situations I want to draw attention to here.

Many groups maintain control through massive threats of violence, bribing people more than the cops could ever equal, holding loved ones as hostages, reliably killing snitches, etc. These are closer to sovereign states than shadowy conspiracies, and also not our subject today.

Today, our topic is Conspiring as an Information Theory game. I think it is a very weird game.

Theory

Let me first describe a math problem:

You are a standard economist's perfectly rational logician, in a community of other perfect logicians. You are (secretly) a Conspirator, and would like to start a conspiracy. X others in the community are also (secretly) Conspirators, and want to start the same conspiracy, but you don't know who. If the Conspirators can obtain mutual knowledge of their identities, then they win.

To stop you are Y Secret Police, whom you also do not know. They win if they can obtain knowledge of who any of the Conspirators are, or else forever stop the Conspirators from winning. If you wish, you may instead imagine these as snitches, fools, or other non-contributors who must be kept out of the group project at all costs. There are also a much larger number of Normal People, who will report you if you admit to being a Conspirator but aren't actively out to get you.

You have the full variety of communication options: You can directly message anyone under your real name, a pseudonym, or anonymously. You can start group chats. You can put up public announcements on the town notice board or anonymously graffiti information on some wall where everyone will see it. You can use powerful encryption. The Secret Police can also do all these things.

You haven't agreed on a plan with the other Conspirators in advance (how could you, you don't even know who they are yet), but you know everyone is a perfect logician, so if there's a strategy, you'll all deduce the same strategy and do it in perfect synchronization. You're also all very risk-averse, so you'd like to win with at least near certainty, and not rely on lots of random luck.

Do you have a winning strategy?

I'll give you a minute to try to think of the answer.

...

Done yet? Great. There isn't one. It's a simple Strategy Stealing Argument: Whatever your scheme is, the Secret Police figure out the same scheme and do the same moves. At some point, your scheme must tell the other person your real identity, but you can't know they're not a cop because the cops behave identically to your allies. No matter how clever your plan is, at the end of it, you'll be in an anonymous chatroom with X+Y people, every Conspirator and every Secret Police agent, and no one will be able to leave anonymity and meet up in real life with confidence, on account of the pure luck shot in the dark you'd have to take. Since you're risk-averse and clever enough to deduce this all from the get-go, you don't even try in the first place. 

Practice

Real life is more forgiving environment than this math problem. You can tell, because people have solutions to this problem. There are many examples we could use, but the one I want to pick out is Jeffrey Epstein's alleged child sex trafficking operation.

To run an alleged child sex trafficking operation, you need to be able to find a continuous stream of rich people who are inclined to do sex crimes. You need to let them know that you offer sex crimes on your private island, if they're interested. The overwhelming majority of people (I hope), if someone just told them that out of the blue, would go to the police immediately. You need to be so good at picking them that you only tell people in the tiny minority who'd say yes, or who at least wouldn't turn you in about it.

The rich person you're trying to sell to has to know you're serious. The police do attempt sting operations about this kind of thing, but they mostly get poor idiots. The kind of people Jeffrey is looking for don't fall for that kind of thing, but they (apparently) do take up the offer when it's the real deal. One way to achieve this is by owning a private island: Police sting operations knowably never involve owning a private island, because it's not in their budget. It may or may not be in the budget of something like the CIA, but if that's your threat model, this whole game gets ten levels harder again. Another threat is the sort of person who changes his mind later and turns you in, or being kicked out of every social space because the normal people think you're disgusting. For the operation to work, you have to avoid both of those, too.

Clearly, someone at some point is solving our mysterious secret recognition problem: The villains can identify each other, without tipping their identity in either direction. Meanwhile, the normal people can't identify the villains, even while they are receiving whatever strange secret handshake has been invented by all these aspiring co-conspirators who have never met before and have had no opportunity to coordinate on a secret handshake.

I'd like you to think for a minute about how incredibly hard this problem is, and how bizarre it is that so many people successfully solve it.

If the police could (reliably) spot the bad guys, they'd arrest all the bad guys. If the bad guys couldn't reliably spot the bad guys, they wouldn't be able to form (these types of) conspiracies. But somehow the bad guys (limited resources, no prior coordination, everyone else against them) have such a natural advantage at self-recognition that they can beat the odds, some of them regularly for years. We only know about them at all because they get caught for some completely unrelated mistake. How are they doing it?

1) Different Behavioural Rules

In real life, neither Conspirators nor Secret Police are identical perfect logicians, which makes it a lot easier, because in theory almost any material difference offers an easy solution. The simplest way involves something Conspirators can do that Secret Police can't, such as commit murder or share original child sexual abuse material. The police can make asking someone to do murder a crime too, but you can get around that by merely hinting at it (in the first case), or by meeting first under mutual pseudonymity (as in the second case).

Physical Proof of Villainy under initial Mutual Pseudonymity is the clear gold standard here. If that's possible, the bad guys just automatically win and you can't stop them unless you're willing to spend significant amounts of effort faking proof of whatever false claims they pick. Still, most of the time, this isn't the solution used. It's just too high a level of effort, involves working with total strangers (which humans aren't built for), and needs everyone to buy in completely before they can even get started.

A secret society of serial killers could demand you send proof of you committing a bloody murder to their encrypted mail server on the dark web. Epstein couldn't do this. If he demanded Proof of Villainy up front from clients he wouldn't get it (for fear he would immediately blackmail them), and if they demanded it of him they wouldn't get it (for the same reason). For this strategy to work, you need to meet under a pseudonym, which they weren't high-tech enough to be doing.

2) Different Evidentiary Rules

The next way to distinguish cops from conspirators is what they consider convincing evidence. If they have different evidentiary rules, you can invent the Selective Proof of Villainy. The gang convinces the aspiring member that it's a real gang and not a hundred cops pretending to be a gang. The aspiring member (if they are a cop), still hasn't seen anything that constitutes legal proof of a crime, but (if they are a gangster) has seen plenty that constitutes (to a gangster) proof they are a gang. A hundred cops pretending to be a gang wouldn't use the right lingo, wouldn't meet in an old warehouse they took over, wouldn't visibly look like they're on drugs (though at this point, the aspiring member can't have seen any actual drugs, for fear they're a cop). It wouldn't hold up in court, but it works in real life.

Secret Police believe in "Proof Beyond Reasonable Doubt that is Admissible in a Court and is Persuasive to a Jury of Random Normal People" or at least "Articulable Probable Cause sufficient for a Warrant". Fellow Conspirators believe in "Actual Bayesian Evidence as Best They Can Approximate It". The margin between these two is free space to communicate, but if you have that kind of free space to communicate, you can send any message you want and win on day one.

This shouldn't work on a rationalist, because a rationalist also believes in Actual Bayesian Evidence as Best They Can Approximate It, so there's no gap big enough to sneak a message through. Despite this, I'm pretty sure Epstein wasn't immediately detected by every honest thinker he ever met at a party. If normal people think Jeffrey is a monster, they'll stop inviting him to parties and introducing him to potential clients. Jeffrey needs to maintain a public reputation here.

If an answer to this exists, it means there's something bad guys can do that other bad guys consider to be compelling evidence that someone is a bad guy (and they're right!), that you dear reader (as, I hope, the normal person you are) wouldn't identify as evidence of anything in particular (and you'd be wrong!).

Doesn't that spook you? This guy at the party you're attending is doing the "I'm secretly a child sex trafficker" dance, right in front of you, and the one other sex offender in the party is updating correctly from it while you're not even noticing it's happening.

Don't you at least aspire to do well enough to stop inviting that guy to parties?

3) Sliding Scales

A normal person would turn you in to some authority if you said you were a sex trafficker, but not if you told them you were just a little bit edgy. First, you reveal (by your behaviour) that you're a little bit edgy, and watch closely how they react. If they think it was gross, you apologise, return to acting normal, and now you know they're not very fun. If they laugh, you act a little bit more edgy.

So long as there's a continuous spectrum from normal guy to sex trafficker, so long as the other guy could plausibly be at any point along that spectrum, and so long as no one would take huge offense at someone merely 1 step worse than themselves (but would tip off that they're not edgier than that by looking upset), then you can walk down the path and see where the other guy stops. Sometimes it stops at dirty jokes. Sometimes it stops at wishing violent deaths upon the outgroup. Sometimes it just keeps going.

This won't defeat a serious agent of the Secret Police, who will play along at being a fellow villain all the way till you do something he can arrest you for, but it'll defeat all the normal people, and you can exclude the Secret Police in other ways.

How can you respond to this?

• The person who is making the edgiest joke he could get away with, and stopping only when others are offended, might be merely 10% more edgy than the group, or he might be a thousand times worse. You don't know either way. He only stopped because of negative feedback. If you really want to know how bad they are, you'd have to play along no matter what they suggest and see how far they take it. Consider not tipping people off as quickly. Don't stop when you think it's questionable; keep laughing at their jokes until you've made up your mind about them.
• A person actively saying things they can barely get away with, to as many people as possible, seeing who responds well or not, should be seen as being incredibly suspicious. You're not going to get specific proof of anything, but you might want to uninvite them from future parties anyway.

4) Shared Knowledge (The Reason I Wrote This)

Qualia are a convoluted topic in philosophy that nobody has a useful way to talk about. This is a practical problem for practical people, and it would be weird for something as esoteric as the nature of qualia to matter, but I think it's the real secret here.

Every aspiring Conspirator has something in common. They know what it's like, how it feels on the inside, to be an aspiring Conspirator. There might be no good way to write this feeling down, but it's still knowledge, common knowledge prior to any coordination held as a communal secret on which to base a strategy.

Pick an arbitrary detail of how it feels to be a drug dealer, something only drug dealers know. For example, maybe there are a few different types of drug addicts, as to how they interact with their dealers. Maybe they follow a few patterns of behavioural norms. Encode this information in something non-suspicious, such as a joke. "Have you seen all those addicts on the street? <humorous impression>". If the other party is not a drug dealer, they will think you made a funny joke. If the other party is a drug dealer, they will realise your impression of a drug addict contains information about how drug addicts behave when interacting with dealers, information you'd only have if you were yourself a dealer. Nobody else can tell if it was accurate or not, so nobody else thinks they've learnt anything. Selective Proof of Villainy.

A single event won't be enough, because complicated things like that can happen by chance, and maybe the other party will miss it, but it's a little bit of Selective Proof of Villainy, and it scales well. Another aspiring conspirator might notice but be uncertain, and so respond with the same protocol back. After enough repeats they become increasingly sure, and increasingly transparent about it, until they're buying drugs off each other.

Although this is hard to do in real life, I think of it as the strongest option, because in the pure mathematics game it causes the conspirators to win with ease. Someone anonymously sprays graffiti describing a mapping function from qualia to encryption keys, where everyone can see it. You map the qualia of being an aspiring conspirator to the corresponding encryption key and use it to encrypt graffiti of your real name. On the morning of day 3, the conspirators have a list of all their names and win the game.

It also works for defending yourself against posers, idiots, snitches, and unreliable flakes. Just think of what it's like to be a sane, careful, honest con, the kind who really wouldn't screw up or betray his buddies (you are an honest con yourself, aren't you?), and make that your filter.

The part that makes it hard in real life is that you're not a bunch of perfect logicians, a mapping function from qualia to encryption keys hasn't been invented yet, and probably ordinary humans couldn't use it properly even if it had, but those feel like surmountable implementation details to me. In practice, it'd look like a normal first date, watching each others faces closely for specific reactions that only make sense if they know something not directly stated but entangled with your words, and the thing it is they're knowing is what it's like on the inside to be the kind of person you're looking for.

How can you respond to this?

• Treat people saying weird things that don't make sense to you as suspicious.
• Whenever people say anything, try to think about ways it could hypothetically encode knowledge common among villains. Check if they're closely watching how you react.
• When people seem like they're trying to show you a bit of what's in their heart, believe that they're the kind of person who'd have that in their heart.
• Spend actual time and actual effort trying to imagine what it'd feel like on the inside to be other people making the choices they could face, more effort than the other people are expecting you to, and have some of that private information be less private than they thought it was.
• Play the same game for Team Good in reverse: Reflect on the qualia of what it's like to not be evil. Talk to people about their life and their experience of the world while secretly encoding that qualia in your questions. This still won't get you specific evidence of particular crimes, but it's probably reason to stop inviting people to parties if they reflect an inability to describe what it's like to have a conscience.

Figuring out how to implement these in practice is left as an exercise for the reader.

Further Applications

Most things that follow conspiracy dynamics are not criminal conspiracies. Romantic love is a conspiracy between two people to do right by each other against the interests of everyone else who'd rather they do right by everyone else, and distinguishing a good romantic co-conspirator who really loves you and doesn't just want your money or your soul is approximately the same game. Ordinary friendship, office politics, startup co-founders who really believe in the mission and aren't planning to make you do all the work, fellow philanthropists who aren't running scams. Same Trait Recognition is an extremely valuable skill to have in almost any context. If Epstein is better at this than you, you should wonder what he knew that you don't.

Figuring out how to do that is also left as an exercise for the reader.

Challenge Question

Beliefs are supposed to pay rent, and the contents of this post are supposed to pay rent primarily in new ways to hypothesize around why people say or do things, so let's give an example. It is, annoyingly, a political example, but it's the best one I've got.

Way back in 2002, before any of this was public, Donald Trump said in a magazine:

I’ve known Jeff for fifteen years. Terrific guy. He’s a lot of fun to be with. It is even said that he likes beautiful women as much as I do, and many of them are on the younger side. No doubt about it — Jeffrey enjoys his social life.

What belief state was Donald Trump in when he said this? What did he expect to achieve by saying it? If it was out of force of habit, why would he be in the habit of saying things like this? If he didn't mean the literal content of this statement to a general audience (which seems a crazy thing for him to be doing), who was he trying to speak to, and what was he trying to tell them?

I have my own theory here that I included in an earlier draft, but decided not to put them here as authoritative. I will say I don't think he could have known nothing, because "on the younger side" is a specific fact that someone who didn't know anything couldn't have guessed, and that it's probably intended that nothing in it constitutes admissible evidence of any particular crime.

I think that, whatever the true answer to these sorts of problems is, it should explain why people say things like this. What thought process produced the speech act, what goal they're trying to achieve, what game they think they're winning.

And I want to know what that game is, because I'd really like to be able to tell when people are playing it.