The Game of Thrones board game is similar to Diplomacy (so I hear: I've never actually played Diplomacy). You often need to make alliances to survive, but these alliances are weak. It is both expected and required that you will eventually break your alliances, otherwise you will lose. My first time playing this game, I made an alliance with a neighboring House which turned out to be unwise, and severely limited my options. To me, breaking an alliance to win a game (even if it was socially acceptable) didn’t feel right/wasn’t worth the negative feelings, and so I ended up stuck on my island for the whole of the game. 

Instead of adapting by learning to be ok breaking alliances, which I considered to be a sub-optimal solution, I fixed the problem by targeting my alliance terms. Now, instead of a general alliance, my offers were along the lines of “I won’t attack you across this border for the next four rounds, if you agree to the same.”

This had the effect of actually strengthening my alliances. Limiting the terms of the alliance to something that could easily be complied with, meant that defecting was no longer expected. The cost goes down, and the benefits go up. In the previous game, both I and my ally would’ve had to leave our mutual border semi-defended, because we knew the alliance wouldn’t hold against a strong enough temptation of conquest. This was facilitated by the fact that the alliance was expected to be eventually broken. In the targeted alliance, I and my ally can leave our mutual border undefended, since we can expect the alliance to hold. This is facilitated by the fact that there would be social sanctions against breaking the alliance. (e.g. I wouldn’t form alliances with that person in future games, because I knew they would break them.)

This lead me to the thought that prohibitions seem to focus on three axes:

  • Specific/Vague Wording: When you ask for a favor, say “please” v. Be polite
  • Targeted/Broad Expectations: Don’t do this one thing v. Don’t do this whole class of things
  • Social expectation to comply/fudge: Don’t have sex with another (wo)man v “Don’t even look at another (wo)man!" 

 

General Examples:

  • Speed limits- specific, targeted, but expected to fudge
  • Terms of use agreements- specific, broad, expected to fudge (no one even reads them)
  • Bribing officials in corrupt societies (there are laws against it, but it is expected as the way to get anything done, or even considered a perk of the office)- vague (giving “gifts” is appropriate?)
  • "Thou shall not kill"- specific, targeted, expected to comply
  • Paper shields- being asked to sign something that is vague and broad. The idea is that it is broad enough to cover anything and everything, and so is expected to be broken. But as long as you don't do anything egregious, they don't enforce it.

 

It seems to me that making an injunction specific and targeted increases the expectation of compliance. This is important to me, because I seem to dislike injunctions that I am expected to fudge. 

Another example of how this plays out in my life: Being enmeshed in a poly network, there is a lot of talking about people (not necessarily in a bad way-- if you ask me about my day though, my answer is going to involve other people). To get around worrying about if I am ever breaking confidence, I specifically tell people I am close to that I don’t consider any information to be private unless it is specifically stated as so (this goes two ways). This way, I get to "gossip" but also people know they can strongly trust me with any information that is prefaced with "This is not for public consumption..." In this example, like the board game, I am turning a broad, vague injuction that isn't strongly expected to be followed ("don't ever talk about other people") into an injuction that can be trusted to be followed by making it specific and targeted. 

Relevant to previous discussions on: ask v guess cultures, and the idea that if it's expected that everyone breaks a specific law then the government can arrest anyone they want to

New to LessWrong?

New Comment
11 comments, sorted by Click to highlight new comments since: Today at 11:10 AM

Paper shields- being asked to sign something that is vague and broad. The idea is that it is broad enough to cover anything and everything, and so is expected to be broken. But as long as you don't do anything egregious, they don't enforce it.

It's not a shield, it's a whip. Something to threaten to beat you with if you get uppity. Or if they just feel like it. It's not whether you do something egregious, it's the threat of beating you generally to control you, and the added power to beat you when they find it in their interest.

We could imagine an actual shield, designed so they can get away with anything. Like, "you agree to not ever sue us".

However, such shields have spikes. If you can't sue them, you lose leverage, and they can get away with hitting you with the shield…

Interesting. Reducing the scope of an agreement in this way also makes defection easier to detect, so it would be more feasible (in a gaming environment with the same group each time) to play the "violate agreements with me and I precommit to attacking you with everything I have" strategy. In which case you're still vulnerable if you lack second strike capabilities, but it still could be useful.

I think the precommitment would be credible, because breaking it would render the strategy useless in the future with the same group.

An alternative that would reduce first-strike vulnerability might be to precommit to attack the defector with everything you have next game, iff he wins this one. This wouldn't be fun, though, because he wouldn't want to play next time when he knew he would lose -- this would probably cause bad human dynamics and, possibly, resulting in everyone teaming up to blow you away next game as a result. In a tournament setting with a large number of games, though, I think it would be a good fix.

In my experience, precommitments are extremely dangerous to use in games in general.

They're seen as somewhat rude - you're doing something that will make you lose "just for spite". Even if you're clearly justified, people still think things like "it's just a game! she's taking it way too seriously!" or "wow, the point is to win, why would you just suicide yourself just to mess things up for someone?" This is especially the case if you're only playing a few games in a row, so the reason for acting vengeful is less salient. In addition, aside from reputational damage, there's also a serious risk of starting a tit-for-tat. Humans can be pretty vengeful, and believe that they're almost always in the right, so if you precommit to attacking someone, they might say something like "well, okay, then I precommit to attacking you back if you execute that attack" - and they might follow through on it too. And then you both lose in a spiral of negativity as everyone else looks on in bemusement.

Well, this is mostly based on online games, though, maybe it's a little better in real life.

A more efficient nuke would be to threaten to ensure the demise of the defector for this game, and the next. To avoid the spiral of hate going supercritical, you can also tell upfront (and in public) that any retaliation will not be met with further vengeance.

Alice: so, we don't attack each other for the next 4 rounds.

Bob: okay

Alice: Oh, and if you break your promise, I swear I will do my best to make you lose this game, and the next

Bob: hey, it's not cool! If you mess up with me on the next game, I will mess up with you on the second next!

Alice: retaliate if you will. I will play normally anyway.

Bob: Wait, you will not seek further revenge?

Alice: Of course not. What would be the point? I don't want to lose every time, and so don't you.

So, what are the expectations here? Let's assume 6 players, with a base expectation of winning of 1/6. If Bob breaks the alliance for certain victory, we have Bob 1, Alice 0, Others 0. Then Alice follows through on her threat. Bob, 1, Alice 0, others 1/4. Then Bob retaliates. Bob 1, Alice 0, others 1/2. And then we play normally. It wouldn't be wise of Bob to break the alliance for anything less than certain victory, because Alice would likely ensure his demise, leading to an eventual score of Bob 0, Alice 0, others 3/4.

Now this could go further, and kill the fun of the game. But I have reasons to believe it may not:

  • Alice's promises makes Bob's retaliation pointless. Not only will it not piss off Alice, but the retaliation will also cost Bob another game. Bob will be aware of this, and may think about it for two seconds.
  • The other players will be aware of the fight. They won't be surprised to see Bob and Alice fight each other, and the game won't be rigged.

Now, making oneself the enemy of the defector for one game and a half may be too much anyway. Just promising to refuse any further alliance for this game and the next may have the desired effect, without some of the nasty consequences.

Wrap it in emotions. People understand emotions.

What do you mean, specifically? "Having fun" aside, being emotional about a game is socially harmful/uncool in the same way a precommitment can be.

I'm not sure how to formalize this, but I think you can be emotional in a playful way that sort of transparently emulates real emotional responses. I haven't played many social board games so I can't speak from experience, it's just that imagining the situation my first response is "transparently act betrayed and insulted, so people understand both that you don't mean it seriously and why you don't help them".

Interesting! Any examples of "vague and broad but expected to comply"? Seems like those would be a bit problematic.

Arguably "don't be racist" could fit the bill, though I suspect it might fall under "expected to fudge". "Don't be a dick" is similar.

Kudos. This approach (improve compliance by narrowing scope) is quite fascinating and probably innovative.

I wonder if there is any literature about it.

Childrearing generally works best when children are trusted with small, short term responsibilities that grow as their capability for self-control improves. "Always do what I've just told you" works far worse than "Clean your room for the next 5 minutes".

Barter and trade are very narrowly scoped alliances. "I promise not to steal, lie, or cheat until the completion of this transaction." I wouldn't trust a shopkeeper who said "Will you join my LLC so that we can trust each other enough to exchange this loaf of bread for some of your coins?"