Cooperation is not always an admirable thing. Gang members cooperate with one another to prey on the general public, enhancing their circumstances while reducing the net prosperity of society and inflicting dire pain on some. The winning entries in the 20th-anniversary Iterated Prisoner's Dilemma competition, a team from England's Southampton University, demonstrate this.

Teams could submit multiple strategies, or players, and the Southampton team submitted 60 programs. These, Jennings explained, were all slight variations on a theme and were designed to execute a known series of five to 10 moves by which they could recognize each other. Once two Southampton players recognized each other, they were designed to immediately assume "master and slave" roles -- one would sacrifice itself so the other could win repeatedly.

If the program recognized that another player was not a Southampton entry, it would immediately defect to act as a spoiler for the non-Southampton player. The result is that Southampton had the top three performers -- but also a load of utter failures at the bottom of the table who sacrificed themselves for the good of the team.

Another twist to the game was the addition of noise, which allowed some moves to be deliberately misrepresented. In the original game, the two prisoners could not communicate. But Southampton's design lets the prisoners do the equivalent of signaling to each other their intentions by tapping in Morse code on the prison wall.

Kendall noted that there was nothing in the competition rules to preclude such a strategy, though he admitted that the ability to submit multiple players means it's difficult to tell whether this strategy would really beat Tit for Tat in the original version. But he believes it would be impossible to prevent collusion between entrants.

"Ultimately," he said, "what's more important is the research."

"What's interesting from our point of view," he said, "was to test some ideas we had about teamwork in general agent systems, and this detection of working together as a team is a quite fundamental problem. What was interesting was to see how many colluders you need in a population. It turns out we had far too many -- we would have won with around 20."

Jennings is also interested in testing the strategy on an evolutionary variant of the game in which each player plays only its neighbors on a grid. If your neighbors do better than you do, you adopt their strategy.

"Our initial results tell us that ours is an evolutionarily stable strategy -- if we start off with a reasonable number of our colluders in the system, in the end everyone will be a colluder like ours," he said.

Perhaps not. Cosma observes:

The clever thing the Southampton group did was to engineer a situation that TFT couldn't cope with, namely collusion among the competing players. If, indeed, one agent is willing to be stomped on, forever, to the greater glory of another, without getting anything out of it, then its master will indeed get all the benefits that the dilemma is capable of providing. (At this point, you can add your own allusions to Hegel, or "safe, sane and consensual" jokes, as you prefer.) This does not seem to me at all an evolutionarily stable situation, however, since the slave agents have, by construction, exactly no incentive to participate in the arrangement. In fact, a mutant which used the coding scheme to recognize supposed masters and always defected against them, but played TFT with everyone else, should do better than a slave, and without slaves the master-type agents are not going to do well...

So I strongly doubt that in the wild, e.g., in actual social dilemmas, we will see Southampton-type strategies, which means that TFT should still be robust, and strong reciprocity is saved for another day. --- I should add that I'm pretty sure the Southampton groups recognizes all of this.

(For the really hard core people. If we were trying to model actual biological evolution, I think the prospects for Southampton strategies are even worse. Case 1: slaves are not related to masters. Then slaves obviously go extinct, after which masters are not long for this world. Case 2: slaves are related to masters. Then we've got the usual kin selection case of, e.g., sterile castes in eusocial insects. Since the pay-off function has to be inclusive fitness, we don't really have the Prisoners' Dilemma at all!)

Cosma also mentioned the decent performance of Tit-For-Two-Tat, TF2T, where defectors are given a second chance.

If play is slightly noisy, forgiving-TFT will actually invade a population which just consists of pure TFT, because it's less likely to get accidentally locked into a cycle of retaliation. Even if play isn't noisy, forgiving-TFT does just fine in a population that's mostly TFT. So the mere fact that there's some strategy which does better than TFT is unsurprising. The virtue of TFT lies not in its invincibility but in its robustness: it does pretty well in a huge range of circumstances.

Among those of us with some shit on our boots these are known as Mom's rules and Dad's rules. Mom is softer and gives bad kids a second chance to do right, TF2T, a good thing since all kids are bad sometimes. Dad's rule has more grit and is taught in the sing-song saying "fool me once, shame on you; fool me twice, shame on me", TFT. Some times Mom and Dad change roles in this as in everything else and sometimes the balance is missing and the kids only learn one rule from their parents.

Of course some kids are jerks, sociopaths, and always defect. The smart ones try to become masters and gather a gang of slaves. The dumb ones (or wise ones? Is it possible to be a dumb, wise jerk?) are loners. Sometimes slaves wise up and revolt and often engage in stealthy mini-rebellion against masters, shrugged of as a cost of doing business by the masters with occasional severe punishment to reduce frequency of occurences. In real life all these things happen in society and it looks like this:

Spatial Prisoners' Dilemma can give you very nice pictures, where there are clusters of very forgiving, cooperative agents, all being nice to each other, surrounded by a periphery of unforgiving TFT players, with completely uncooperative, defection-prone hordes beyond. This works because TFT does pretty well against both the nasties and the good citizens; the good citizens prosper because they have the TFT types to ward off those who would take advantage of them.

It takes all kinds since it seems that the protective ring of Dad's rules is what keeps the inner circle of Mom's rules comfy given the predatory wackos on the periphery. Unfortunately, the master types seem to dwell among the inner circle too. We call them politicians.

Also see So Good It Hurts which muses about gene-culture coevolution.