The Rookery

Madeline's thoughts on social deduction games, forum/community meta, and any other philosophical musings

Archive for Social Deduction Schema

 Thumb up

Syntax and Semantics

msg tools
No mountains, no valleys
Never argue with idiots; they'll drag you down to their level and then beat you on experience.
Microbadge: Crosswords fanMicrobadge: Taboo fanMicrobadge: I record detailed playsMicrobadge: I helped celebrate BGG Werewolf's 10th Anniversary!Microbadge: Ice Cream lover
(Preemptive disclaimer: the werewolf moderator who wrote the below rules turned out to be a pretty terrible person, don't be that guy, but the example is still illustrative.)

A few years ago I played in a game on this site that had some hidden-role elements, and (among others) the following rule disclaimer:
a jerk wrote:
There will not be any semantic-based posting restrictions. (In other words, nothing that restricts anyone's intent behind posting.) Syntactic restrictions are possible (e.g. "you cannot post the word 'martyr'" is syntactic, "you cannot claim to be the martyr" is semantic).
As a logician and part-time linguist, I've found this distinction useful in academia, and it turns out to also be interesting in a game context!

In this game, it turned out that there was a syntactic restriction; players "bitten" by the "Space Dracula" hadtopostlikethis. Itwasveryfunny. Italsoturnedouttobeafunwaytomesswiththeauxwhobecameoutedevil. Particularlybecausethedraculawasherboyfriend.

There was also a semi-hidden mechanic in which one of the good players ("Sarax") received information about the evils' roles, and the entire thread knew this. However, only Sarax (and the evils) knew that if he was identified as the knowledgeable good ("Merlin," if you're into Resistance), evil would win. So the uninformed goods were in a state of confusion, "someone knows something that will help us, why aren't they sharing??" In this case, the knowledge that there were not semantic restrictions in the game was a useful limitation on how weird the gamestate could be; it wasn't the case where "Sarax knows something but isn't allowed to tell us." Although there were some questions after the fact about the balance of the Merlin mechanic, I found that this roleset benefited both from the amusing syntactic restrictions, and the shared knowledge that there were no semantic ones.

Anyway, at the recent Rathcon I played several other games where this was an issue. My overarching thesis is "syntactic restrictions are (usually) good when they're used for humor; semantic restrictions are (usually) bad because they usually rely on subjectivity to enforce." Some examples!

Games like Werewolf and Resistance don't need many or any semantic restrictions, because the incentives to tell the truth or lie are built into the win conditions. If everyone goes around telling the truth about their role, the evils will be outed immediately. Similarly, if everyone tells the truth about whether they played a "pass" or a "fail" in Resistance, the spies will be outed. So instead the spies will falsely say "I played a pass" and the rebels will truthfully say "I played a pass."

There is a minor semantic restriction depending on the implementation, which is that you can't talk about your card art! Because if a good guy says "I'm the villager with the ponytail," evil has no way to counter that. The online equivalent is "no directly quoting modchat or fake-quoting." This doesn't seem to be a difficult restriction, however, because most people sort of understand that it's "outside the game."

Unfortunately, some versions of (mostly IRL) werewolf have a poor implementation of the mason role. Masons are good players who know each other's identity. There doesn't need to be any more nuance to it than that.

In some rolesets, however, the fact that two good players can confirm each others' identity and make it very difficult for evil to counter them (unless two wolves want to go head-to-head, but that's risky) is too powerful for good. The solution, of course, is to fix the roleset. But some moderators/designers get around this by saying "you can't claim to be a mason."

What kind of a restriction is that? If the fact is "anyone who utters the words 'mason' gets immediately modkilled," that's a syntactic restriction. But it's not a very effective one, because players will just start saying "please don't lynch A, he's my comrade in a fraternal brotherhood." So in order for the rule to be a threat, it has to be "don't claim to be a mason in any way shape or form," which is semantic.

But it's also bad, because it relies on a moderator's interpretation to draw the line. "Pleeeeeeeease don't lynch A?" "I'm voting to save A because I think she's good." "Rarwwrarwar all of you are stupid because I'm the only one who knows how to read A and I say she's good!" Rules should be strong enough to avoid these kinds of judgment calls.

The Menace Among Us is also a social deduction game, but compared to werewolf and Resistance, much more crunchy. (I believe along the lines of Battlestar Galactica, although I've never played that one.) In "Menace," various players put cards from their hands into a middle pool each round, possibly supplemented by random ones from a random deck if there aren't enough. The cards determine what happens to the spaceship, whether good and/or evil advance their win conditions. However, unlike Resistance, there are many more options than just "pass" or "fail," there's maybe a dozen different outcomes.

This means that, if there were no semantic restrictions, goods could truthfully claim the card they played, and evils who deliberately played bad cards would have to go head-to-head with one of the goods. And that would narrow down the pool of evils (informed minority) very quickly. So the rule is "you can't claim the specific card you put in, but only the 'category'--was it 'good,' 'neutral,' or 'bad' according to this sheet." One of the issues with that is that players might disagree with the categorizations--we found the "each player draws one card" card to not be very strong, and would not consider it a "good" play. Maybe a deeper issue is that it seems a lot easier to "derp-clear" yourself in this than Resistance--if someone's like "I put a good card in, oh wait" *looks at cheat sheet* "apparently it's a neutral" it feels hard to go after that as an evil? Whereas no rebel in Resistance would be like "I put in a fail. I mean...a pass."

3 Laws of Robotics starts out as a social-y deduction game; you can see everyone else's identity, but not your own, and everyone gets a chance to ask a yes-or-no question. Players score points by correctly giving their "keys" to players they believe to be on their team and having high-ranking numbers. So it's also iterative in the sense of the social deduction schema; maybe last round, all the Androids got one point, but this round we'll have newly randomized roles, and so there will only be one individual winner. Assuming, of course, that the distribution of points allows that.

There are four rounds. After each of the first three, a new "law" is placed on the table, and players must follow it. One example of a law is "players must say 'logging off' while away from the table or looking at their phones." Now, this doesn't at all affect the gameplay in the sense of "you only ask one yes-or-no question." So in that sense, it's syntactic, not semantic.

But! The rules as written say that you can score points by catching other people breaking the rules! So this isn't just pasted on, it's a way to distinguish scores in a game that's otherwise fairly brief and samey.

But! I've only played this twice; the first time, an otherwise usually calm and easygoing person got frustrated with the rule enforcement, and the second time, nobody even bothered with that. So the rules were no longer part of the game, just an unofficial thing to make it sillier. It's a short enough game that I didn't mind, but with a heavier game I would have asked "if it doesn't really count, then why bother."

But! The rulebook mentions the potential for laws such as "you cannot tell the truth to an AI." Now that's crossed over into the strategy domain--if you catch someone breaking that law, and try to score a point for it, now you've informed the questioner that they're an AI! So that seems to open the door to more interesting incentives and complexity.

But! The laws like that ("you cannot tell the truth to __") aren't in the base deck; they're in a separate envelope labeled "warning, may create dangerous paradoxes" or something. I can see why, but at the same time, if those are the cards that introduce the deep complexity rather than optional silliness, it seems a shame to hide them away.

Disclaimer: I have not actually played Mountains of Madness. I suspect given the horror theme and the real-time aspects, it might not be to my tastes. But some of my BGG friends hit on the idea of playing Resistance with some of the speaking restrictions. For instance, "anyone someone else says a number, you have to jump in and say the next higher number." So it turned into "I want to go on mission 3." "Mission 4?" (Everyone laughs at Madeline who's worried about mission 4 already.) This is purely syntactic--it's a house rule, so it doesn't affect the underlying game--but with the right cards, it can be hilarious.

And of course, Citadels. This game is 99% building districts and drafting roles. But if the Ball Room is one of the purple cards in the game, and if you build it, and if you become king...then everyone has to say "thanks, your excellency!" or lose a turn. Some people think this is silly because the game is not otherwise about this, and just never play with it--the newer edition doesn't even contain it. But I think it's hilarious.
Twitter Facebook
Sat Feb 1, 2020 1:01 am
Post Rolls
  • [+] Dice rolls
 Thumb up

Social Deduction Schema - Part Two

msg tools
No mountains, no valleys
Never argue with idiots; they'll drag you down to their level and then beat you on experience.
Microbadge: Crosswords fanMicrobadge: Taboo fanMicrobadge: I record detailed playsMicrobadge: I helped celebrate BGG Werewolf's 10th Anniversary!Microbadge: Ice Cream lover
So now, an overview of some of the social deduction games I've played in recent years!

Werewolf: very open, very pure. Kind of the founder of the genre, everything else has sort of evolved from there. Lots of complex implementations (especially online), which make things slightly less open with regards to complexity of night actions.

The Resistance: Moderately open, mostly pure. A lot of the game is discussion and argument between rounds, with actual proposals and voting a small percentage of that; if you can't hold your own or don't like the debate, you probably won't have a good experience. The base game, as I've said, is very much "resistance will win if they identify each other," but Avalon roles complicate that. (That said, the assassination phase in Avalon boils down to a "spies discussing together" open format.)

One Night Ultimate Werewolf: The night phase is closed, but then the entire day phase is open: lots of back and forth arguing, claims, retracts. It's also not super pure: people can stumble into victory without necessarily identifying their entire team (or indeed, even knowing what team they're on!)

I'll give this the uncomplimentary tag of obscurant: players' win condition in many cases is hidden from them themselves, so it's hard to know how to advance towards it. As you might have guessed, I really dislike these kinds of games. Some people don't! Some people are like "yeah, you have to think on your feet and know when to back into and out of claims, that's the fun of it." If that kind of spontaneity works for you, that's great, but that kind of time-pressure thinking is not what attracts me to social deduction games.

Shadow Hunters: Symmetric, closed. Although the Hunter/Shadow conflict is where the symmetry comes from, this game excels because of the neutrals. The fact that there won't always be the same ones in the game make it much more replayable than Bang (see below), and lessens the incentive for everyone reveal on the first turn. (But once people do reveal, the symmetry makes it less stressful than werewolf for people who don't like to be informed-minority.)

BANG!: closed. Vaguely symmetric? After a few rounds it'll become pretty clear who's who, with maybe some uncertainty on the part of the Renegade. At which point, the "social deduction" aspect is lost. Good for some thematic laughs but my biased review is it doesn't really do anything Shadow Hunters doesn't do better.

Crossfire: Also obscurant, and close to symmetric despite a slight edge to blue.

Werewords: On paper somewhat pure (villagers need to find werewolves, werewolves need to find seer), but feels kind of crunchy in practice (someone is playing 20 questions...but badly!).

Saboteur: Symmetric at least in the second edition with red and blue teams. We could also call it iterative in that, to compensate for the lightness/arbitrariness of an individual round, the official suggestion is to score for one round, then go again with new roles and compete for highest total score. This edges it out of true social deduction (your role in one game has nothing to do with your role in the next), and if people took it too seriously, could probably lead to some degenerate cases/kingmaking. So the fact that it's suggested feels like a copout of "welp, an individual round is pretty random, if you don't want to be stuck with it here's another :/ option." I did warn you this would be a subjective list.

Spyfall: inverted (maybe one of the originators of this subgenre?) I don't really like the inverted format because it can be easy to fall into groupthink and be suspicious of someone just because their manner of obscuring the secret isn't the same as yours. As I always say, if I wanted to be suspected for not conforming with groupthink I'd go play another futile game, like Talking To Neurotypicals.

A Fake Artist Goes to New York: inverted, see above.

Emergence: A Game of Teamwork and Deception: Closed, somewhat crunchy. I like this game! Maybe it's because the (informed) minority wins by doing the same thing the majority does--reaching a number of accumulated chips in the chip box proportionate to the number of their players. So it has a bit more of a pure feel, to me? Like instead of being "lousy at harvesting cubes," the puny humans are trying to efficiently get cubes and convert them just like the glorious robots--they just happen to put the chips in the wrong side of the box. (Or not, if they want to bluff!)

Deception: Murder in Hong Kong: pure, open. In some ways this is like werewolf, so why don't I like it? Maybe because the groups I play it with have been a little lax about the rules for who can speak when, so it devolves into a free-for-all that's weighted towards good. So the official rules try to help evil by delimiting "okay now someone theorizes, now someone else goes, now time's up, stop," but that feels kind of arbitrary.

Witch Hunt: pure, open. This is basically werewolf but slightly more closed with regards to special roles/night actions, and slightly more open with regards to "you better talk fast because day is ending now!" Plus the dead people, who have to coordinate together without letting on "this is one group of ghosts" "this is another." The futility of that can make it feel almost too open.

Latitude 90: The Origin: Win conditions fairly pure, beyond that hard to say. The mechanics of sending and receiving information is potentially very closed, but if you're like "uhh, don't trust anyone, what to do" that can be kind of open in the unpleasant, futile way. (There's probably a different issue for conversion games, I know there's been plenty of debate about those!)

GROWL: Closed. Win conditions don't really fit into any of these: there's no informedness, potentially no elimination, villagers don't necessarily have to deduce anything either. Probably goes with Latitude 90 under "conversion is weird," only much lighter.

...Wait, never mind, it's iterative! That was the point of gold. Okay, there we go.

Are You the Traitor?: Very light and open, also iterative.

Two Rooms and a Boom: Symmetric, open. Can lead to feelings of "well, now what" futility, and the "coin-flip" degenerative issues make it a nope for me.

Secret Hitler: Somewhat more closed than Resistance, probably? And also crunchier, in that the drawing of cards/discarding allows for the "evil hiding as good with an unlucky hand" strategy.

Dark Moon: closed, very crunchy.

Shadows over Camelot: ditto

Games I only played a couple times years ago and barely remember:

Blood Bound: Symmetric

Ultimate Werewolf: Inquisition: Closed, crunchy.

The Last Banquet: I'm not sure what my issues with it were; it's possible there were just too many people, which could make even a closed game feel like "well, I have nothing to do, better wait for these dozen people to go." But there's also an open aspect of "who's gonna be the assassin" "uhhhh" that wasn't engaging either. Symmetric, anyway.

Games that are probably not social deduction, but share some features in common:

Mascarade--obscurant, but no teams or allegiances. First to thirteen points wins, independently of role.

Betrayal at House on the Hill--once the traitor is revealed, it's an asymmetric all-against-one game, but there's no mystery about who that is. (The Legacy version may add some twists to the original, I only played a couple games there so I can't say.)

Lifeboat players have secret objectives which identify their win conditions, but again, individual winner.

Edge cases:

Dead of Winter: mechanically it plays a lot like the "crunchy" examples, and could be described as "co-op with a hidden traitor," but the individual goals (for traitors as well as good guys) complicate things. Part of the goods' goal is to find and neutralize the traitor, but that's not a necessary or sufficient win condition.

New Salem: Second Edition: it's a "most points wins" individual type of game, but in order to even be in the running to win, Puritans/Witches need the village to have low/high quantities of Doom, respectively. So in some respects it plays a bit like a pure game where everyone, majority or minority, is trying to say "I'm part of the majority team, please leave me alone so we can work towards our common goals, such as identifying the witches and putting them on trial." I like that kind of thing, though!
Twitter Facebook
Wed Jul 3, 2019 4:44 am
Post Rolls
  • [+] Dice rolls
 Thumb up

Social Deduction Schema - Part One

msg tools
No mountains, no valleys
Never argue with idiots; they'll drag you down to their level and then beat you on experience.
Microbadge: Crosswords fanMicrobadge: Taboo fanMicrobadge: I record detailed playsMicrobadge: I helped celebrate BGG Werewolf's 10th Anniversary!Microbadge: Ice Cream lover
How do we describe social deduction games? Beyond the obvious breakdown of "the ones I like" and "all the others," there seem to be a couple more underlying dimensions that can categorize them in ways that make it easier for me to articulate what I do and don't like. I wouldn't say they're full-scale dimensions that are completely independent of each other, but rather various adjectives that can in some cases be opposites, or in some cases label a specific category.

I consider Werewolf to be the progenitor of the modern social deduction genre, with perhaps a time jump separating it from spinoffs such as Bang! and The Resistance. There are also a couple games that seem to fall near, but not exactly under, this umbrella (New Salem). As a starting point, let's say that a social deduction game is one where you're issued a secret identity near the start of the game, these identities determine your win condition, and some or all of them fall into factions/teams.

Closed versus Open

Most traditional, competitive board games have very discrete turns. On my turn, I move my pawn or draw two cards or place a worker, and then it's the next player's turn. Sometimes this fosters indirect interaction ("hey, you took my spot!"). And sometimes, the turns are more open-ended; in Catan, even when it's "my turn," I can open the floor to anyone who's willing to trade with me, which is a more direct form of interaction.

In social deduction games, there can be a lot more "free-for-all"ness. The long "day" phases of werewolf feature very little turn-taking, although players are forced to make a decision when voting phases come along. But mostly players are left to talk amongst themselves to root out evils, or deceive goods, in their own way and at their own volume. I'll call this an open game.

Something like Emergence, however, is more closed: players have fixed rules for what actions they can take, when, and where. Of course you can argue and accuse people in between this, but once I've indicated my action, I need to move to an adjacent hex and possibly claim a resource.

Purity and lack thereof

In standard werewolf, the win conditions are pretty easy to define: the village wins by eliminating all the werewolves, and the werewolves win if this doesn't happen (within a certain number of rounds). Basic Resistance is a bit more complex, but in the same vein: the rebels will win if they send three passing missions, which will certainly happen if they can correctly identify all the other rebels (although it can also happen if a couple spies float early on, collide and double-pass, etc). However, the spies will win if they send three failing missions, which they will do if even one spy is continuously trusted by most rebels enough to get on the last mission! Avalon, Hunters, etc. complicate this.

I'll call these games relatively pure, not in the sense that they are necessarily more enjoyable or morally superior than others, but that their win conditions are fairly simple to describe. Uninformed majority: wants to discover who's who, informed minority: wants to prevent this. The informed minority's goal, therefore, is to act as much like a member of the uninformed majority as possible, to gain trust.

From now on I'll probably use "good" and "evil" interchangeably with "uninformed majority" and "informed minority," if the game in question has these factions (not all of them do!). Werewolves are not "evil" because they eat villagers (some villagers are tasty, they had it coming), but because they have to lie and pretend to be villagers. And many people find that kind of lying more cognitively demanding than the deduction process of being a villager.

So I think most people would agree that, although it has teams, werewolf is not a cooperative game. The basic mechanic requires a lynch mob (or whatever the politically correct term for it these days). Yet, there are several games that meet my definition of "social deduction" that could also be called "cooperative with a traitor." Shadows over Camelot and Dark Moon are examples of these. I'll call them crunchy games (by the standards of social deduction!).

In some ways, these are similar to true cooperative games like Pandemic, where the struggle is that of all the players versus the board. In Shadows over Camelot people are running around gathering swords and removing catapults, in Dark Moon repairing systems and resolving crises. (Dead of Winter is similar, but with some caveats. See part II.) These have significantly more crunch than "I think A is a rebel so I want her to go on the team, thumbs up." For the good guys, the win condition is achieved by beating the game, which requires cooperation and teamwork. For the bad guys, the win condition is not letting this happen.

But the bad guys don't need to stay hidden to win! In fact, they can win even after being exposed as bad guys, and in some cases gain new powers. This can make it fun to play the second half of the game as a revealed traitor. (Source: I seem to draw traitor all the time in these.)

There is a downside, however, which comes in the first half of the game. Sometimes goods will just get a bad hand of cards, and they'll have to say "well, I couldn't contribute much to that round, but it's just bad luck, I really am good please trust me. " This is necessary in order for there to be genuine uncertainty about who is evil. But on the evil's part, they have to play like an incompetent villager: "oh I just tried to find the Holy Grail but I tripped over my horse oopsies" This is subjective, but I think the challenge of playing "like a good guy who just has terrible cards" is less engaging overall than the challenge of "I really am a villager, here is me hunting and looking shiny" while eating people at night. (The latter is hard. But interesting!)

Not all games even have the concept of an informed minority. In Two Rooms and a Boom, the Blue Team and the Red Team have (roughly) equal character powers; in Shadow Hunters, the Shadows and the Hunters are chosen to be equal in number. You might consider bombing heads of state or being a vampire to be thematically evil actions, but, you know, from their point of view the political establishment is evil. Let's call these symmetric examples of the social deduction genre.

And there are also examples of games that feature an "informed majority!" Maybe they're not all informed of each other (that would be pretty broken), but they're all in on a secret that the uninformed minority has to act like they know. (Spyfall, Fake Artist.) I'll call these inverted since they switch up the usual uninformed majority/informed minority paradigm.

For iterative, obscurant, and “almost-but-not-quite social deduction” games, stay tuned for part II!
Twitter Facebook
Tue Jul 2, 2019 3:55 am
Post Rolls
  • [+] Dice rolls