So, Nonlinear-affiliated people are here in the comments disagreeing, promising proof that important claims in the post are false. I fully expect that Nonlinear’s response, and much of the discussion, will be predictably shoved down the throat of my attention, so I’m not too worried about missing the rebuttals, if rebuttals are in fact coming.
But there’s a hard-won lesson I’ve learned by digging into conflicts like this one, which I want to highlight, which I think makes this post valuable even if some of the stories turn out to be importantly false:
If a story is false, the fact that the story was told, and who told it, is valuable information. Sometimes it’s significantly more valuable than if the story was true. You can’t untangle a web of lies by trying to prevent anyone from saying things that have falsehoods embedded in them. You can untangle a web of lies by promoting a norm of maximizing the available information, including indirect information like who said what.
Think of the game Werewolf, as an analogy. Some moves are Villager strategies, and some moves are Werewolf strategies, in the sense that, if you notice someone using the strategy, you should make a Bayesian update in the direction of thinking the person using that strategy is a Villager or is a Werewolf.
As I mentioned to you before, I suspect werewolf/mafia/avalon is a pretty bad analogy for how to suss out the trustworthiness of people irl:
in games, the number of werewolves etc is often fixed and known to all players ahead of time; irl a lot of the difficulty is figuring out whether (and how many) terminally bad actors exist, vs honest misunderstandings, vs generically suss people.
random spurious accusations with zero factual backing are usually considered town/vanilla/arthurian moves in werewolf games; irl this breeds chaos and is a classic DARVO tactic.
In games, the set of both disputed and uncontested facts are discrete and often small; this is much less the case irl.
in games, bad guys have a heavy incentive to be uncorrelated (and especially to be seen as being uncorrelated) early on; irl there are very few worlds where regularly agreeing with the now-known-to-be-bad-actors is a positive update on your innocence.
EDIT: Rereading this comment, I think it was unclear. Basically in games, if we know Alice and Bob seem in-sync, (eg they vote similarly, often go on the same missions), if we later learn that Alice is definitely evil, this is not always an update that Bob is evil. (and in some fairly common scenarios, it’s actually a positive update on Bob’s innocence).
This almost never happens in real life.
Similarly, if Alice repeatedly endorses Bob, and we later learn Alice is evil, irl we can often write off Alice’s endorsements of Bob. In games, there are sometimes structural incentives such that Alice’s endorsements of Bob are more trustworthy when Alice is evil (Good guys are often innocent/clueless, bad guys know a lot of information, bad guys usually don’t want to be paired with other bad guys).
In games, the set of actions available to both good and bad actors are well-defined and often known in advance; irl does not have this luxury.
etc
All these points, but especially the second one, means that people should be very hesitant to generalize hard-won lessons about macrolevel social dynamics from social deception games to real life.
random spurious accusations with zero factual backing are usually considered town/vanilla/arthurian moves in werewolf games; irl this breeds chaos and is a classic DARVO tactic.
In my experience this is only true for beginner play (where werewolves are often too shy to say anything), and in advanced play it is a bad guy tactic for the same reasons as IRL. Eg I think in advanced Among Us lobbies it’s an important skill to subtly push an unproductive thread of conversation without making it obvious that you were the one who distracted everybody.
It’s not clear/concrete to me in what ways points 3 and 5 are supposed to invalidate the analogy.
in games, bad guys have a heavy incentive to be uncorrelated (and especially to be seen as being uncorrelated); irl there are very few worlds where regularly agreeing with the now-known-to-be-bad-actors is a positive update on your innocence.
I don’t understand this—it reads to me like you’re saying a similar thing is true for the game and real life? But that goes against your position.
Eg I think in advanced Among Us lobbies it’s an important skill to subtly push an unproductive thread of conversation without making it obvious that you were the one who distracted everybody.
I’m not much of an avid Among Us player, but I suspect this only works in Among Us because of the (much) heavier-than usual time pressures. In the other social deception games I’m aware of, the structural incentives continue to point in the other direction, so the main reason for bad guys to make spurious accusations is for anti-inductive reasons (if everybody knows that spurious accusations are a vanilla tactic, then obviously spurious accusation becomes a good “bad guy” play to fake being good).
I don’t understand this—it reads to me like you’re saying a similar thing is true for the game and real life? But that goes against your position.
Sorry that was awkwardly worded. Here’s a simplified rephrase:
In games, bad guys want to act and look not the same. In real life, if you often agree with known bad folks, most think you’re not good.
Put in a different way, because of the structure of games like Avalon (it’s ~impossible for all the bad guys to not be found out, minions know who each other are, all minions just want their “team” to win so having sacrificial lambs make sense, etc), there are often equilibria where in even slightly advanced play, minions (bad guys) want to be seen as disagreeing with other minions earlier on. So if you find someone disagreeing with minions a lot (in voting history etc), especially in non-decision-relevant ways, this is not much evidence one way or another (and in some cases might even be negative evidence on your goodness). Similarly, if Mildred constantly speaks highly of you, and we later realize that Mildred is a minion, this shouldn’t be a negative update on you (and in some cases is a positive), because minions often have structural reasons to praise/bribe good guys. At higher levels obviously people become aware of this dynamic so there’s some anti-inductive play going on, but still. Frequently the structural incentives prevail.
In real life there’s a bit of this dynamic but the level one model (“birds of a feather flock together”) is more accurate, more of the time.
This is very tangential, but: if that’s your experience with e.g. one night ultimate werewolf, then I strongly recommend changing the mix of roles so that the numbers on each side are random and the werewolf side ends up in the majority a nontrivial fraction of the time. Makes the game a lot more fun/interesting IMO, and negates some of the points you list about divergence between such games and real life.
The game theory behind Werewolf goes deeper than that. Werewolf is an iterated game, if you play it at least once on a friday you’re probably playing at least four more times in succession. A good way to pick up whether someone is a Villager or a Baddie is to notice how their behavior during the game correlates with their revealed role at the end of the game.
Alice is a noob player and is always quiet when she’s a Baddie and talkative and open when she’s a Villager. She’s giving off easy tells that an observant player like Bob picks up on. He can then notice these tells while in the middle of a game and exploit them to win more against Alice.
Declan is a more skilled but somewhat new player. He is open and talkative regardless of his role. This makes it very easy for him to play Villager but he struggles to win when a Baddie because his open behavior leads to him often being caught out on provable lies.
Carol is a sophisticated Werewolf player. Each game she is maximizing not just to win that game, but to also win future games against the same players. Carol knows that she is the most sophisticated player in her group. When she’s a Villager she can figure out which other players are Baddies much more often than the other Villagers. Her best plan as Villager then is to convince the other Villagers that her reads and analysis are correct without regard to the truthfulness of her persuasive strategies. Some people notice that she’s not being 100% truthful and call it out as Werewolf behavior, but everyone at the table acknowledges that this is just how Carol plays and sometimes she lies even as a Villager. This serves her well in her next game as a Baddie where she uses the same tactics and doesn’t give away any tells. Carol is no more suspicious or less open about her own info on average as a Baddie than as a Villager.
Errol is a Logical Decision Theorist. Whenever he’s playing a game of Werewolf, he’s trying to not just win that game, but to maximize his probability of winning across all versions of the game, assuming he’s predictable to other players. Errol firmly commits to reporting whether he’s a werewolf whenever he gets handed that role, reasoning that behind the veil of ignorance, he’s much more likely to land as villager than as werewolf, and that villager team always having a known villager greatly increases his overall odds of winning. Errol follows through with his commitments. Errol is not very fun to play with and has since been banned from his gaming group.
Each game she is maximizing not just to win that game, but to also win future games against the same players.
This sounded really wrong to me. Like, what is the analogy in real life? I am a good guy today, but I predict that I may become a criminal tomorrow, so I am already optimizing to make it difficult to figure out?
But I suppose, in real life, circumstances also change, so things that are not criminal today may become criminal tomorrow, so you can be a good guy today and also optimize to make yourself safe when the circumstances change, even if your values won’t.
So, Nonlinear-affiliated people are here in the comments disagreeing, promising proof that important claims in the post are false. I fully expect that Nonlinear’s response, and much of the discussion, will be predictably shoved down the throat of my attention, so I’m not too worried about missing the rebuttals, if rebuttals are in fact coming.
But there’s a hard-won lesson I’ve learned by digging into conflicts like this one, which I want to highlight, which I think makes this post valuable even if some of the stories turn out to be importantly false:
If a story is false, the fact that the story was told, and who told it, is valuable information. Sometimes it’s significantly more valuable than if the story was true. You can’t untangle a web of lies by trying to prevent anyone from saying things that have falsehoods embedded in them. You can untangle a web of lies by promoting a norm of maximizing the available information, including indirect information like who said what.
Think of the game Werewolf, as an analogy. Some moves are Villager strategies, and some moves are Werewolf strategies, in the sense that, if you notice someone using the strategy, you should make a Bayesian update in the direction of thinking the person using that strategy is a Villager or is a Werewolf.
As I mentioned to you before, I suspect werewolf/mafia/avalon is a pretty bad analogy for how to suss out the trustworthiness of people irl:
in games, the number of werewolves etc is often fixed and known to all players ahead of time; irl a lot of the difficulty is figuring out whether (and how many) terminally bad actors exist, vs honest misunderstandings, vs generically suss people.
random spurious accusations with zero factual backing are usually considered town/vanilla/arthurian moves in werewolf games; irl this breeds chaos and is a classic DARVO tactic.
In games, the set of both disputed and uncontested facts are discrete and often small; this is much less the case irl.
in games, bad guys have a heavy incentive to be uncorrelated (and especially to be seen as being uncorrelated) early on; irl there are very few worlds where regularly agreeing with the now-known-to-be-bad-actors is a positive update on your innocence.
EDIT: Rereading this comment, I think it was unclear. Basically in games, if we know Alice and Bob seem in-sync, (eg they vote similarly, often go on the same missions), if we later learn that Alice is definitely evil, this is not always an update that Bob is evil. (and in some fairly common scenarios, it’s actually a positive update on Bob’s innocence).
This almost never happens in real life.
Similarly, if Alice repeatedly endorses Bob, and we later learn Alice is evil, irl we can often write off Alice’s endorsements of Bob. In games, there are sometimes structural incentives such that Alice’s endorsements of Bob are more trustworthy when Alice is evil (Good guys are often innocent/clueless, bad guys know a lot of information, bad guys usually don’t want to be paired with other bad guys).
In games, the set of actions available to both good and bad actors are well-defined and often known in advance; irl does not have this luxury.
etc
All these points, but especially the second one, means that people should be very hesitant to generalize hard-won lessons about macrolevel social dynamics from social deception games to real life.
In my experience this is only true for beginner play (where werewolves are often too shy to say anything), and in advanced play it is a bad guy tactic for the same reasons as IRL. Eg I think in advanced Among Us lobbies it’s an important skill to subtly push an unproductive thread of conversation without making it obvious that you were the one who distracted everybody.
It’s not clear/concrete to me in what ways points 3 and 5 are supposed to invalidate the analogy.
I don’t understand this—it reads to me like you’re saying a similar thing is true for the game and real life? But that goes against your position.
I’m not much of an avid Among Us player, but I suspect this only works in Among Us because of the (much) heavier-than usual time pressures. In the other social deception games I’m aware of, the structural incentives continue to point in the other direction, so the main reason for bad guys to make spurious accusations is for anti-inductive reasons (if everybody knows that spurious accusations are a vanilla tactic, then obviously spurious accusation becomes a good “bad guy” play to fake being good).
Sorry that was awkwardly worded. Here’s a simplified rephrase:
Put in a different way, because of the structure of games like Avalon (it’s ~impossible for all the bad guys to not be found out, minions know who each other are, all minions just want their “team” to win so having sacrificial lambs make sense, etc), there are often equilibria where in even slightly advanced play, minions (bad guys) want to be seen as disagreeing with other minions earlier on. So if you find someone disagreeing with minions a lot (in voting history etc), especially in non-decision-relevant ways, this is not much evidence one way or another (and in some cases might even be negative evidence on your goodness). Similarly, if Mildred constantly speaks highly of you, and we later realize that Mildred is a minion, this shouldn’t be a negative update on you (and in some cases is a positive), because minions often have structural reasons to praise/bribe good guys. At higher levels obviously people become aware of this dynamic so there’s some anti-inductive play going on, but still. Frequently the structural incentives prevail.
In real life there’s a bit of this dynamic but the level one model (“birds of a feather flock together”) is more accurate, more of the time.
This is very tangential, but: if that’s your experience with e.g. one night ultimate werewolf, then I strongly recommend changing the mix of roles so that the numbers on each side are random and the werewolf side ends up in the majority a nontrivial fraction of the time. Makes the game a lot more fun/interesting IMO, and negates some of the points you list about divergence between such games and real life.
Thanks, I haven’t played ONUW much, Avalon is the main game I play, also more classic mafia, werewolf, secret hitler and Quest.
The game theory behind Werewolf goes deeper than that. Werewolf is an iterated game, if you play it at least once on a friday you’re probably playing at least four more times in succession. A good way to pick up whether someone is a Villager or a Baddie is to notice how their behavior during the game correlates with their revealed role at the end of the game.
Alice is a noob player and is always quiet when she’s a Baddie and talkative and open when she’s a Villager. She’s giving off easy tells that an observant player like Bob picks up on. He can then notice these tells while in the middle of a game and exploit them to win more against Alice.
Declan is a more skilled but somewhat new player. He is open and talkative regardless of his role. This makes it very easy for him to play Villager but he struggles to win when a Baddie because his open behavior leads to him often being caught out on provable lies.
Carol is a sophisticated Werewolf player. Each game she is maximizing not just to win that game, but to also win future games against the same players. Carol knows that she is the most sophisticated player in her group. When she’s a Villager she can figure out which other players are Baddies much more often than the other Villagers. Her best plan as Villager then is to convince the other Villagers that her reads and analysis are correct without regard to the truthfulness of her persuasive strategies. Some people notice that she’s not being 100% truthful and call it out as Werewolf behavior, but everyone at the table acknowledges that this is just how Carol plays and sometimes she lies even as a Villager. This serves her well in her next game as a Baddie where she uses the same tactics and doesn’t give away any tells. Carol is no more suspicious or less open about her own info on average as a Baddie than as a Villager.
Errol is a Logical Decision Theorist. Whenever he’s playing a game of Werewolf, he’s trying to not just win that game, but to maximize his probability of winning across all versions of the game, assuming he’s predictable to other players. Errol firmly commits to reporting whether he’s a werewolf whenever he gets handed that role, reasoning that behind the veil of ignorance, he’s much more likely to land as villager than as werewolf, and that villager team always having a known villager greatly increases his overall odds of winning. Errol follows through with his commitments. Errol is not very fun to play with and has since been banned from his gaming group.
This sounded really wrong to me. Like, what is the analogy in real life? I am a good guy today, but I predict that I may become a criminal tomorrow, so I am already optimizing to make it difficult to figure out?
But I suppose, in real life, circumstances also change, so things that are not criminal today may become criminal tomorrow, so you can be a good guy today and also optimize to make yourself safe when the circumstances change, even if your values won’t.
the werewolf vs villager strategy heuristic is brilliant. thank you!
Credit to Benquo’s writing for giving me the idea.