Society for Interactive Literature West

Game Design

The Prisoners' Dilemma

Copyright (c) 1992 Stephen R. Balzac, Aimee Yermish
Society for Interactive Literature
Originally printed in Metagame, the Journal of the SIL

A popular version of the classic prisoners' dilemma is this: two men are arrested for a crime they jointly committed. They are put into separate rooms, and told, "If you both keep silent, you'll each get two years. If your friend talks, he'll go free, but you'll be shot. But if you both talk, you each get ten years." The two men must now decide what to do: does each trust the other to keep quiet, does one talk, or does suspicion reign and they both talk? There are a good many different variations on this problem, but the basic idea is simply this: cooperation vs. conflict. Do players risk cooperating with each other, and share the prize, do they compete and risk losing everything, or do they only pretend to cooperate, and try to be the first with the backstab? While the choice is apparently totally under players' control, in fact it is the structure of the game that frequently exerts the strongest influence. At every step of the way, the potential rewards and punishments set up by the GMs for each course of action determine what choices the players will see as favorable.
On one extreme, there is the old-fashioned killer game, where the last person or team alive wins. Cooperation in such a setting is basically pointless, since everyone knows that there can only be one winner. No matter who you ally with, sooner or later one of you must kill the other. On the other extreme, there is the basic table-top roleplaying game. Sure, there's frequently a certain amount of conflict in the party, sometimes players argue over the treasure, sometimes personality conflicts flare, but the vast majority of the conflict is between the party and the external threats set up by the GM. The external dangers, which form an incentive towards cooperation, far overshadow the internal disagreements which threaten to break up the party.
To take this into the realm of live roleplaying, almost every GM has seen, or heard about, a game becoming a total bloodbath. In such games, it almost inevitably turns out that players have very little need of each other's knowledge or special abilities -- the non-transferable aspects of the character. You're not likely to accidentally kill the only person who can build the desperately-needed Framajam, and the more people you kill, the more items you get to sort through to find the ones you want or can sell. Furthermore, violence is simple, usually effective, and generally untraceable within the structure of the game. You're not likely to be overpowered, caught, or punished, or at least not before Sunday afternoon, so there's no fear involved. Conflict, instead of cooperation, has become the most successful strategy, and the corpses pile up.
The other end of the spectrum in a live game is the classic "peace breaks out all over" scenario. In this game, the players quickly realize that they can accomplish their goals easily and efficiently by cooperation. The first cause of this is a lack of real contradiction between player goals. While they may find plenty of pregame conflicts set up in their character sheets, these rarely have any reality within the game. Two characters may be told they hate each other, but if their goals are compatible, they will generally put aside their differences, whether by tossing their characters out the window or by rapidly coming to a new understanding of each others' positions.
The second cause is a greater danger which threatens everyone (or almost everyone) in the game. This becomes a goal upon which all characters can agree and forms a strong basis for cooperation. If there is a monster about to destroy Tokyo, which computer company's stock goes up more is of little importance until after the monster is dealt with; since this sort of plot is usually resolved late in the game, often as part of the endgame, the inter-player conflicts never get a chance to get resolved. Even if they do have time to be resolved, often the warm feelings characters have developed for each other through working together to banish the external threat encourage players to make compromises they never would have favored before. Thus you get the annoying endgames where people find themselves dancing ring-around-the-rosy and singing "Happy Birthday" to banish a demon (this really happened once!), while the whole web of player-player conflict is forgotten.
The third cause of peace breaking out is when there exist strong in-game punishments for conflict. Sometimes these are enforced by the GMs, such as the magical surveillance cameras which enable murderers to be caught and punished rapidly and without fail, regardless of precautions. Other punishments are set up by the GMs to be enforced by the players. If you are playing a warmonger out to invade the neighboring kingdom, but the potential in-game reward for doing it is small (honor and glory are all very well and good, but they frequently have little in-game value), and the other characters will likely team up to repulse your invasion and stomp all over your country to prevent you repeating it, you're likely to think long and hard before going through with the plans the GMs gave your character. Do you have the strength and the motivation to take on the whole rest of the game? Probably not. Setups like this one can be extremely frustrating to the warmonger. All of these mistakes in game design ensure that the potential benefits of competition simply do not outweigh the potential risks, and you find yourself with one big happy family.
Now, what can be done to achieve a middle ground between these two poles? Well, one simple, obvious, solution that has been tried more than once to avoid the conflict extreme is to simply disallow conflict. The rules of the game simply don't permit effective violence, either explicitly (there exists no mechanism to do it), or implicitly (the victim always makes a miraculous recovery). Other forms of conflict, such as the theft of information (eavesdropping or truth serum) are also disallowed or made so weak as to be useless. Unfortunately, this "solution" has a rather serious problem: removing conflict as a means to changing the game frequently results in a game of "haves" and "have-nots." The power in the game (political, physical, economic, information, or whatever confers power in your scenario) is locked up in the "haves." They have no need of the "have-nots" to accomplish their goals, and the "have-nots" lack any means to dislodge them. The "have-nots" start off unable to affect the game, and the structure of the game makes it impossible for them to change that. They can cooperate with each other all they want, and the "haves" might even allow them along as mascots, but they are completely powerless when it comes to their own destiny or that of the game as a whole. If no one needs them, no one will pay any attention to them, and they have no way of enforcing their desires, no matter how loudly they yelp. Most players feel powerless enough in real life -- they don't need to pay money to experience it in a game. Boredom and frustration aren't entertainment.
So what's the middle ground? Optimally, conflict shouldn't be all that difficult, so as to prevent stagnation. However, choosing the wrong person to knock off or permanently alienate, or getting caught and punished, should be likely enough and carry high enough penalties to form an effective deterrent. Players must be put in situations where they will think twice before pulling a gun and shooting, or screwing someone else out of a political plum. If there is no risk, people playing cooperation-heavy characters who try to hold to their characters will be punished for their good roleplaying, while everyone else will tramp all over the game, doing as much damage as they can on the chance that some of it will pay off. Random, senseless conflict as a means to one's goals should be a losing strategy, while well-thought-out and carefully directed conflict should be highly effective. Punishment should not be guaranteed, because that removes all conflict as a potential solution. For example, Master Thief Wagner plans to steal the Super-Duper-Shield generator from the space station to weaken its defenses against his allies' planned attack. Unfortunately, the GMs have placed a secret and undetectable homing beacon on the generator. No matter how careful Wagner is, no matter what precautions he takes, he will be caught and the generator replaced. To make matters worse, the GM-run police and judiciary are swift, brutal, efficient, and unbribable, so Wagner's best and most in-character efforts to wriggle out of his situation are doomed to failure. The player needs to feel that his thought and actions will make a difference to whether or not conflict is an effective strategy in his situation. The point here is that the mechanisms for conflict should be easy, but the choices of who and when to double-cross should be difficult, such that a shotgun approach is ineffective.
Now, to jump to the other end of the spectrum, how do you reduce the incidence of cooperation, without eliminating it entirely? In other words, how do you add a certain sense of uncertainty to those deals the players are making, while still making it worthwhile to make deals?
The first step is to put much -- if not all -- of the nastiness inside your game. Rather than having the GMs provide the threats, let the players threaten each other. In other words, let various player factions be out to destroy or take over the world, instead of having it be all external. Make it less obvious who is doing what, rather than hanging signs around the necks of the bad guys. Make it impossible for everyone to accomplish their goals simultaneously, and make compromises inherently unsatisfactory. For instance, several groups are competing to get their candidate elected President, a powerful and influential position within the game necessary for each group to accomplish its other goals. Clearly, only one group can possibly succeed.
How do you keep conflict-enforcing plots like this from eliminating cooperation entirely? The second step is to set up a balance. Give players a choice: they can work together toward a greater goal, at the risk of losing out on their individual goals; or they can refuse to cooperate, try to go it alone and claim all the glory for themselves, and risk not only getting nothing, but having everything blow up in their faces. To extend the example from the previous paragraph, the groups do have some hope of compromise: one group can promise to support another's bid for the presidency in exchange for some reward later on or in another part of the game. Of course, once the first group has had its candidate elected, it can choose to renege. Taking this a step further, suppose that one candidate is actually a villain whom nobody wants elected. However, the villain (with typical villainous forethought) is in disguise, so no one knows which candidate he is. Now the groups can never be entirely sure whether to cooperate or not.
To put it another way, if you put enough contradictions between different groups' goals, players can't just assume that they can work out a deal whereby they all win. The best a faction might be able to do is work out a deal whereby all the members are guaranteed not to lose; but such alliances, based on fear of an in-game threat, tend to last only as long as the threat does. Once the villains are defeated, the alliance may find itself breaking into small factions, each vying for power. The question then becomes, when to backstab? Too early and the external threat you allied against might not be defeated; too late and someone else might beat you to the punch.
What we see is here is that cooperation can be reduced by strengthening the reasons for conflict and by making the reasons for cooperation transient, just as earlier we reduce conflict by forcing players to have a need to cooperate. The trick is to set up your game so that from moment to moment it is never clear which is better: conflict or cooperation. Each strategy has its moments of strength and weakness, and they switch back and forth frequently during the course of the game. As conflict increases, the "weaker" players band together to protect themselves; conversely, as people make themselves vulnerable, the temptation to backstab increases. So long as neither strategy has immediate, overwhelming benefits (or punishments), this system of dynamic feedback will work quite well. The players will need to constantly adjust their thinking and strategy; at no time can they be certain they have adopted the correct strategy, because it may no longer be correct. And that, of course, is what a prisoners' dilemma is all about.