Stardate 20020219.2041

(On Screen): In game theory, one of the most important discoveries of the last fifty years was the Prisoner's Dilemma. It represents a demonstration of how it might be that several agents, each operating alone and attempting to maximize their own gain, would collectively arrive at a non-optimal result.

It's often presented in terms of crime figures interacting because it assumes that there does not exist a global agency responsible for enforcing good faith. In its original form, it involves two crooks who are arrested and interviewed separately. Each is told the following: If both of you confess, you both get 3 years in prison. If neither of you confess, you both get 1 year in prison. If you confess and he doesn't, then you go free and he gets 10 years. Likewise, if he confesses and you don't, you get 10 years and he goes free.

In terms of optimizing your own result, it turns out that if he confesses you do better to confess (3 years instead of 10), and if he doesn't confess you still do better if you confess (freedom instead of 1 year).

But when both apply that strategy the result is 3 years for both instead of 1 year for both or anyone going free.

There are a bunch of ways in which this manifests, but one interesting aspect of it is when it's a repeating game instead of a single session. It's usually portrayed in terms of criminal acts because part of the assumption is that there is no overriding agency which enforces good faith and good behavior. So the classic example is a sequence of drug buys.

You want to buy kilos of grass; he wants to sell them. He wants your money. You can "cooperate" by bringing money and paying, you can "cheat" by bringing a gun and taking the drugs without paying. He can "cooperate" by bringing drugs, or "cheat" by bringing something that looks like drugs but isn't. Each of you remembers the next time what happened this time. (The point being that if either guy cheats, the other guy can't complain to the cops, which is necessary for this analysis.)

Both sides benefit most in a long series if they cooperate; money is exchanged for drugs. Of course, if the relationship is about to end and both sides know it, there's a strong incentive to cheat on the last buy. Right?

There's been a lot of analysis of this, and it turns out that honesty isn't the best policy. One guy decided to run a computer tournament; people were permitted to create algorithms in a synthetic language which would have the ability to keep track of previous exchanges and make a decision on each new exchange whether to be honest or to cheat. He challenged them to see who could come up with the one which did the best in a long series of matches against various opponents. It turned out that the best anyone could find, and the best anyone has ever found, was known as "Tit-for-tat".

On the first round, it plays fair. On each successive round, it does to the other guy what he did the last time.

When Tit-for-tat plays against itself, it plays fair for the entire game and maximizes output. When it plays against anyone who tosses in some cheating, it punishes it by cheating back and reduces the other guys unfair winnings.

No-one has ever found a way of defeating it.

Now let's analyze two different and even more simplistic approaches; we'll call them "saint" and "sinner". The saint plays fair every single round, irrespective of what the other guy does. The sinner always cheats.

When a saint plays against another saint, or against tit-for-tat, the result is optimum but more important is that everyone gets the same result. When a sinner plays against another sinner, or against tit-for-tat, everyone cheats and the result is still even, though less than optimal.

But when a sinner plays against a saint, the sinner wins and the saint loses.

Which brings me back to the point of all this: Is there anything I would rule out in war? Nothing I'd care to admit to my enemies, because ruling out anything is a "saint" tactic. The Tit-for-tat tactic is to be prepared to do anything, but not to do so spontaneously. In other words, if the other guy threatens to use poison gas, you make sure you have some of your own and let him know that you'll retaliate with it. That means that he has nothing to win by using it, and he won't. (A war is a sequence game and not a single transaction because each day is a new exchange. If you gassed my guys yesterday, I can gas yours today.)

I chose that example carefully because that's actually what happened in WWII in Europe. After the horror of poison gas in WWI, the world agreed to ban its use in one of the Geneva Conventions, and in fact no-one did use poison gas in Europe in WWII. Not even Hitler, who apparently knew no bounds at all, was willing to. Because he was following international law? Hell no. It was because the Americans and British maintained stocks of poison gas in Europe and were ready to retaliate in kind. (In fact, an American ship carrying 100 tons of mustard gas was sunk by a German air attack in the Italian harbor of Bari in December of 1943.)

And so it is here. No, I cannot promise ahead of time that my nation will not bomb innocents, or use terrorism, or torture, or poison gas, or bio weapons, or nuke, nor in fact can I exclude anything. If I do, then I am adopting a "saint" strategy and leave myself wide open for the use of such tactics against my own side by an opposing sinner. Only by being willing to do those kinds of things myself can I deter their use against me.

I believe that my nation must adopt tit-for-tat instead of using saint tactics, because it is much better. But for that to work, I have to be willing to be as dirty as he is, if he forces me to be.

This is the theoretical basis for such aphorisms as "To get peace, you must prepare for war." That means that your nation is prepared to use tit-for-tat. The pacifist idea of publicly pledging to never go to war, or to never use a particular tactic in that war, is instead a saint strategy, and it results in disaster.

The Geneva Convention is deliberately constructed to be tit-for-tat. It says explicitly that a nation is obligated to follow the convention only if the other nation is also a signatory and is also following it. If the Geneva Convention was binding on signatory nations even against non-signatory nations, it would be a "saint" tactic. But since you follow the convention with others who also do, and don't against those who don't, that makes it "tit for tat".

Tit-for-tat says that you're civilized to those who are civilized to you, and you're a vile son-of-a-bitch to those who want to be that way.

By the way, the nuclear deterrent was another example of tit-for-tat. "Bomb me, and I'll bomb you back. Leave me alone, and I'll leave you alone."