I've been reading an Matt Ridley's enlightening The Origins of Virtue, in which there's a long discussion on the logic of cooperation. For years now, I've thought that the Tit-for-Tat strategy from Robert Axelrod's tournament represented the best strategy. Essentially Tit-for-Tat is: you begin by cooperating but are quick to retaliate if the other person defects, and then you revert back to cooperating.
As Axelrod says, "What accounts for Tit-for-tat's robust success is its combination of being nice, retaliatory, forgiving and clear. Its niceness prevents it from getting into unnecessary trouble. Its retaliation discourages the other side from persisting whenever defection is tried. Its forgiveness helps restore mutual cooperation. And its clarity makes it intelligible to the other player, thereby eliciting long-term cooperation."
In a world with a mix of strategies and at least a few people playing Tit-for-Tat, Tit-for-Tat will always beat the consistently evil. It is also worthy to note that Tit-for-Tat only makes sense in a series of games, and that in a single game, it fails against "nasty" strategies. Given that we happen to live in a world with a time dimension, though, this seems to be a somewhat academic point.
The dark side of Tit-for-Tat, as Ridley notes in his book, is that if two players operating under Tit-for-Tat meet each other, and one of them accidentally defects, then a continuous series of mutual recriminations begins from which there is no escape. This is how blood feuds start and why they never end. So Tit-for-Tat was discovered to be not evolutionarily stable.
Martin Nowak designed a new tournament where nothing was certain and everything was statistically driven, including 'random' mistakes and tactics. The system could 'learn,' evolving over time. In this world, a new strategy prevailed called Generous-Tit-for-Tat. Generous will randomly forgive single mistakes one-third of the time. To forgive all single defections (called Tit-for-Two-Tats) is to invite exploitation, but to do it randomly broke cycles of recrimination. So Generous could beat Tit-for-Tat - but Novak then discovered that Generous could be beat by nicer strategies, but then these strategies could be beat by the perfectly cooperative strategy, which in turn could be beat by the perfectly nasty strategy. So this was not evolutionarily stable either.
Eventually, Karl Sigmund and Martin Nowak came upon the Pavlov strategy, which essentially meant that when the player wins, he continues behaving the same; when the player loses, he shifts tactics. It's also called the Win-Stay, Lose-Shift. Essentially, you don't change until something is wrong. Pavlov builds cooperation, tends to punish when partners defect, and then forgives. But it also screws naive suckers by defecting (and winning) consistently. So 'it creates a cooperative world, but doesn't allow that world to decay into a too-trusting Utopia where free-riders can flourish.' In a learning world, Pavlov is evolutionarily stable.
It makes me wonder about what lessons there are to be learned here about the stability of different political systems. Communism's principles are close to the 'perfectly cooperative' strategy. But it is beat easily by an invasion of the 'perfectly nasty' strategy of tyranny. If you agree that capitalism is like the Tit-for-Tat strategy, then it can be highly effective but is dominated by more cooperative strategies. Ridley implies, however, that Tit-for-Tat, by doing a effective job of punishing, demolishes nasty strategies and clears the way for nicer (but not naive) strategies to dominate. So basically, you should be as nice as you can possibly be without inhibiting feedback and learning.
As Axelrod says, "What accounts for Tit-for-tat's robust success is its combination of being nice, retaliatory, forgiving and clear. Its niceness prevents it from getting into unnecessary trouble. Its retaliation discourages the other side from persisting whenever defection is tried. Its forgiveness helps restore mutual cooperation. And its clarity makes it intelligible to the other player, thereby eliciting long-term cooperation."
In a world with a mix of strategies and at least a few people playing Tit-for-Tat, Tit-for-Tat will always beat the consistently evil. It is also worthy to note that Tit-for-Tat only makes sense in a series of games, and that in a single game, it fails against "nasty" strategies. Given that we happen to live in a world with a time dimension, though, this seems to be a somewhat academic point.
The dark side of Tit-for-Tat, as Ridley notes in his book, is that if two players operating under Tit-for-Tat meet each other, and one of them accidentally defects, then a continuous series of mutual recriminations begins from which there is no escape. This is how blood feuds start and why they never end. So Tit-for-Tat was discovered to be not evolutionarily stable.
Martin Nowak designed a new tournament where nothing was certain and everything was statistically driven, including 'random' mistakes and tactics. The system could 'learn,' evolving over time. In this world, a new strategy prevailed called Generous-Tit-for-Tat. Generous will randomly forgive single mistakes one-third of the time. To forgive all single defections (called Tit-for-Two-Tats) is to invite exploitation, but to do it randomly broke cycles of recrimination. So Generous could beat Tit-for-Tat - but Novak then discovered that Generous could be beat by nicer strategies, but then these strategies could be beat by the perfectly cooperative strategy, which in turn could be beat by the perfectly nasty strategy. So this was not evolutionarily stable either.
Eventually, Karl Sigmund and Martin Nowak came upon the Pavlov strategy, which essentially meant that when the player wins, he continues behaving the same; when the player loses, he shifts tactics. It's also called the Win-Stay, Lose-Shift. Essentially, you don't change until something is wrong. Pavlov builds cooperation, tends to punish when partners defect, and then forgives. But it also screws naive suckers by defecting (and winning) consistently. So 'it creates a cooperative world, but doesn't allow that world to decay into a too-trusting Utopia where free-riders can flourish.' In a learning world, Pavlov is evolutionarily stable.
It makes me wonder about what lessons there are to be learned here about the stability of different political systems. Communism's principles are close to the 'perfectly cooperative' strategy. But it is beat easily by an invasion of the 'perfectly nasty' strategy of tyranny. If you agree that capitalism is like the Tit-for-Tat strategy, then it can be highly effective but is dominated by more cooperative strategies. Ridley implies, however, that Tit-for-Tat, by doing a effective job of punishing, demolishes nasty strategies and clears the way for nicer (but not naive) strategies to dominate. So basically, you should be as nice as you can possibly be without inhibiting feedback and learning.
RSS Feed