Thursday, 31 October 2013

Rules and the Problem of Social Cost

There is discussion in the ether about, yet again, the importance and value of having lots of rules versus having fewer.

Let's ask ourselves: What Would Ronald Coase Do?

In "The Problem of Social Cost", to simplify grossly, Coase observed that if the world was perfectly free of transaction costs, people would be able to negotiate more-or-less perfectly efficient outcomes with each other in most situations. In any sort of dispute people can just arrange with each other a mutually acceptable outcome that will be at least as good as any solution an independent arbitrator could come up with. (The solution might not be perfect, but it would be at least as good as a decision that a judge or tribunal would provide.)

However, we don't live in a world perfectly free of transaction costs, so in practice this is not always what happens. In actual fact, there are often costs to negotiation. Chief among them is time - negotiation is costly in that regard - but there may also be costs to do with discovering information, enforcement, and so forth. (It may not even be immediately clear what the real nature of the dispute is, or who it is with.)

Sometimes the transaction costs are nugatory or small, and not enough to outweigh the efficiency and desirability of the result. But on some occasions the transaction costs are so high that they outweigh the efficiency of the result. It is in these situations that a court, in Coase's view, should and does become involved in coming up with a solution.

He used the example of a real-life case about a sweet-maker and a doctor. The sweet-maker causes lots of noise which disturbs the doctor. They could negotiate a solution between themselves: whether they decide after negotiation that the doctor will move, or the sweet-maker will reduce his noise, or the doctor will just put up with the noise, or the sweet-maker will pay some form of compensation to the doctor, the solution they negotiate will be at least as good as what a court would decide.

But the transaction costs to negotiation may be high. What if the doctor holds out for excessive compensation and negotiation takes a long time? What if it is not just the doctor who has a problem with the noise, but other neighbours? What if some of those neighbours have genuine complaints but others make frivolous, vexatious or speculative claims? It will take a long time for the sweet-maker to negotiate with all those neighbours and sort out which ones have genuine complaints.

This is when a court needs to become involved, in Coase's view - when transaction costs become so high that a negotiated solution is unsatisfactory or nigh-impossible. And his opinion was that, when a court comes up with a solution, it needs to do so in the most efficient way possible - meaning, the solution it comes up with must be as close as possible to what the parties would have negotiated between themselves if there were no transaction costs. That should be the guiding principle of a court in any private law dispute - it should pursue the outcome which would have arisen if the parties had been able to freely negotiate an outcome without cost.

So let's postulate a theory - not exactly to transpose or extend Coase's theorem about social costs, but rather to come up with something that is inspired by it only.

Imagine that a "court" in the RPG context is "the game's rules". The general presumption should be that a game should only have a rule to cover situations in which transaction costs for DM fiat will be prohibitively high in comparison to the outcome achieved.

A very good example is combat rules. Deciding who is the winner and loser of a combat without reference to rules comes with an extremely high set of transaction costs - it would take ages to argue it out to a satisfactory conclusion ("Ah, but this character is wearing chain mail!" "Ah, but this character would have ducked!" "Ah, but the floor is wet - this character may have slipped!"), and it may have the more important additional cost of heated debate - PC lives are at stake and the players may disagree vociferously with the decision arrived at. Better to have a court, in the form of rules and dice, generate an outcome - and this may well have the effect of generating an outcome which the parties would have freely negotiated (i.e. being partially random but based on mutually agreed principles worked out in advance).

Similarly the surprise roll: deciding who is actually surprised without rules would take forever; the transaction costs are too high, so we just resort to pre-ordained surprise rules which tell us what to do to get an outcome.

In other words, there are scenarios where we can say that DM fiat or complicated rules would create the same or similar outcome (the PC kills the orc or vice versa), but where the transaction costs associated with DM fiat are prohibitively high - so rules are better.

On the other hand, many social situations can be very quickly resolved by DM fiat to generate a result that is at least as good as a set of complicated social rules - i.e., there are no or very small transaction costs to DM fiat, which do not outweigh its value. Does a PC succeed in persuading the guard to let him pass? Regardless of whether the DM or the rules do the job, the outcome will be either that the PC does succeed or he doesn't. DM fiat (or DM-mandated dice roll), in other words, produces an outcome that is at least as good as social rules. But the DM can do it quickly and easily and the consequences are not life-and-death - there are very low transaction costs. Better for the DM to do it, then.

Similarly, to use a well-worn example, an elf is sneaking up on the party while they sleep. However, the PCs have put little piles of coins around their encampment as a primitive alarm system. There are two outcomes: the elf knocks over a pile of coins or he doesn't. DM fiat (or DM-mandated dice roll) is going to produce an outcome that is at least as good as rules about coins around encampments. But the DM can do it quickly and easily and the consequences are not life-and-death. Better for the DM to do it, then.

In other words, there are scenarios where we can say that DM fiat or complicated rules would create the same outcome, and where since the transaction costs associated with DM fiat are low, DM fiat is better.

Of course, nobody would sensibly argue that this metric should be consciously thought through whenever a decision is required. That would be absurd. Rather, it should be used as a guiding principle: there should be a rebuttable presumption that DM fiat (whether outright or in the form of "roll a d6" or whatever) is to be preferred, and the presumption can be rebutted if transaction costs associated with DM fiat would be prohibitive.


  1. Wait, the rules can't be the court, because the rules (in this metaphor) take the place of the law (interpreted by the DM/referee/... judge).

    I thiiiink I agree with your larger point though. Boiled down, you put in rules where you foresee arguments. Yes?

    1. No, the rules in an RPG are more like the court, because they are the neutral arbitrator. Remember, it's extremely unusual for the GM to take the role of an arbitrator in a dispute between players. He is a player and in fact he is one side in almost all disputes.

      But anyway, yes, we agree on the main point!

  2. And if you are an asshole, you will get in more arguments and so need more rules....

    1. ...or at least more arbitratory rules.
      I'm not sure every part of a Rolemaster table, for example, is really an arbitratory rule, more a chance to inject content.

    2. I think there are some 'rules' that aren't really rules. Like random tables. As you say, they're almost like a third category of thing entirely - just producers of content.

    3. They're tied together in an interesting way though. Like a Rolemaster crit IS something that you might theoretically argue over ("would that puncture one lung or both...) but somehow _takes the opportunity that having to look up the arbitratory rule up gives you_ to also give you some content.
      And at a certain point the arbitratory result _is_ content. Like if someone is blinded, that's content. But it's also the result of a situation that needed arbitration.

      It's complicated.

      Sometimes a solution is actually more interesting than the dispute--on purpose.

      Not sure where it ends, but there's something to that...

    4. Damn right, it means that you don't only use those rules when you have arguments, so you don't have the next level up of people going "is this really something you need rules for", "are we really arguing about this" etc.

      The rules are fun in themselves, so no-one even thinks about the possibilty of arguments.

      Also more generally, rules default-solve you problems that are not fun to solve in play, at least not all the time. This can be social/argument problems, content creation problems, finding direction problems etc.

      Play with an underemployed medic in the room, and he'll probably want to solve that problem, you just need something to stop him getting carried away.

  3. Good post; my own rule of thumb, that "the dice should only be invoked if there is disagreement about the outcome" is subsumed.

  4. This has the important corollary of explaining why some groups are better with rules-lite games and others need rules-heavy games. The higher the cost of each transaction created by social conflict, the more important it is to avoid social transactions whenever possible. So you can play a rules-lite game with friends you can trust, but when you're playing with strangers in a high-pressure atmosphere, you need every argument nailed down tight in advance - in the same way, for everyone, every time.

    OD&D is the friendly Saturday game that's customized differently for each local group, AD&D is the nationally standardized convention-tournament game.

    1. Yes, there's definitely something to that.

    2. I agree completely. That's why I tend to start new players with games with rather strict rules.
      When we know each other, we can play a game of Backswords and Bucklers, Fu, or Unknown Armies. Until then, it's time for Reign and Runequest 6.

    3. You know, this has shed some light on the pickle I found myself in, trying to run Backswords and Bucklers for a group of trusted friends who were also new players. Treating them as unknown quantities in need of a potential hard-rules solution to any given transaction might have helped to mitigate some of the awkwardness inherent in the first-time RPG experience. The hand could slacken from the throat somewhat over time as it became clear what did and didn't need the hard rules.

      Curiously, this is the sort of thing that was always very apparent when I was about fifteen and first running RPGs - we almost automatically started off trying to play by the book and worked out what we didn't need as we were going along, without needing to work it all out in essays and Euclid problems beforehand. I wonder what happened?

  5. I'm not sure if I can agree with this. You say "complicated rules" in the penultimate paragraph, but D&D's surprise rules, for instance, are not really that complicated. Actually, it is more like standardised DM-fiat ("let's see, we both roll and whose roll is 1 or 2 is surprised, ok?").

    If a game consists of only suck rules, does that count as rules-light?
    And what if it applies to a hundred different situations listed?
    And what if there are situational modifiers that change the die size accordingly?

    Also, the example about talking to the guardsman may not have serious consequences, nor does a fight between a 9th level fighter and a goblin, although, according to the rules, there is still a few rolls involved. On the other hand, there are no rules for convincing the king not to execute a dear party member, despite being a life-or-death situation.

    1. The point with convincing the king not to execute a dear party member is that DM fiat is at least as good as rules to cover that situation, and probably better. As distinct from combat, where DM fiat will be a horrendous and time-consuming mess.

      You're right about 'complicated' - that shouldn't matter.

    2. I'm afraid you're wrong about combat. There are games that run combat on DM fiat, and it works. (Want examples? Amber Diceless, for starters, and among the more contemporary, Stalker RPG is a good one). And it works.
      The king can just as easily kill the PC as a random arrow.
      The rules are for the time where you don't want the group to hold you responsible for looking them in the eyes and saying "not good enough, the PC dies".

    3. Sure, but Amber Diceless requires a level of maturity and also probably friendship to really work properly. It also has plenty of rules and guidelines for how the GM is supposed to adjudicate. It's an exception which proves a rule.

    4. It's possible that this is just semantics over something we largely agree on, but I feel like convincing the king is like a lot of other things that could succeed or fail in a game, which I wouldn't adjudicate solely by fiat.

      I would adjudicate some sort of general action roughly as follows: the player starts by saying what she's doing (crossing the street, jumping across a 10' chasm, jumping across a 200' chasm), next I evaluate whether there's an interesting chance of success or failure (crossing the street succeeds by fiat, jumping 200' fails by fiat, jumping 10' has a chance of success or failure), and then adjudicating the things that might or might not succeed using rules/dice rolls. For me the 10' chasm is a simple dice roll with some sort of difficulty rating based on how good the character is at that sort of thing and any exigent circumstances. I would personally describe the die roll as a rule, just a simple rule.

      I would apply the same basic approach for convincing the king: If the player gives a really great, convincing speech, she just succeeds. If she starts telling at the king about how horrible he is, he gets pissed off and she auto-fails. If she gives kind of a meh speech, it could go either way, and I'd have her roll a die against a difficult rating.

      So then the question becomes: which is more useful to the game: a simple rule, or a more complex rule or set of rules? If it's a game where palace intrigue and influence peddling are core elements of the setting, I might want complex rules so that the players have better guidelines on how to achieve success. But for a more normal campaign, a simple, quick rule (and possibly some background on the king to make it easier to determine what arguments he's likely to find persuasive) is likely to be all that's ever needed.

      Like I said initially, I'm not sure this actually disagrees with anything you posted, but maybe it does?

  6. A very interesting post, which I spoiled for myself towards the end by starting to imagine a short story-game called ELF/SNEAK...

  7. Okay, I'm going to commit a cardinal internet sin. I'm going to disagree with the source material based solely on your summary, with no actual familiarity with it. But I'm admitting it up front, and pro-actively apologizing.

    I think that this argument falls flat due to the classic reason most economics arguments do: It assumes participants who are both perfectly rational and perfectly informed. What if one side insists on a solution that is less than optimal, due to one fallacy or another? What if the arbiter (or, in the RPG, random chance) is able to generate a solution that is actually better than anything the participants could have though of? I'm fairly certain I could construct such a scenario, particularly when it comes to combat results.

    Further, the argument supposes that the rules solely exist as a means to resolve disputes. However, in my experience, that is only one function of RPG rules. They also serve to provide a common framework for expressing the setting (e.g., "Swinging from chandeliers is so cool that it should be the characters' default method of movement."). They serve as inspiration for character roles and actions (this is the primary function of most classes, feats, etc.). They serve as a meta-language to rapidly communicate ideas between the players ("+1 sword" is just short-hand for "a remarkably well-balanced and keen blade that seems to seek out foes with a mind of its own"). They help to maintain the risk/reward balance, suggesting appropriate challenges and results along the way.

    Admittedly, many of those things can be looked at as reducing the transaction cost of free-form storytelling. But I feel that doing so ignores entire dimensions of the interaction between rules, players, and GMs. The rules are not solely meant to be walls, or even bridges. They are intended to be scaffolding that allows us to build something grander than we could on our own.

    1. That's a criticism that's often levelled at Coase, but actually he was arguing the same thing - if there were no transaction costs you would get perfectly efficient negotiated solutions, but transaction costs are ever-present and frequently high. He said in reality his theorem is almost always inapplicable. That's one reason why he argued that firms exist.

      I think your meta-language idea actually builds the argument. There is a high transaction cost to saying "this is a remarkably well-balanced and keen blade that seems to seek out foes with a mind of its own" without a rules context - what exactly does "keen" mean? What kind of bonus should it bestow, if any, and in what situation? Saying "It's a +1 sword" circumvents the social cost.