Evolution and Public Policy

In his books The Selfish Gene and The Extended Phenotype, Richard Dawkins (father of the "meme") borrows a term from Games Theory to help explain the seemingly intentional direction of evolution and the emergence of complex behaviors, culminating with intelligence and culture. The term in question names the concept of an Evolutionarily Stable Strategy (ESS). Here's an attempt at a brief explanation:

Strategies are evaluated according to their potential payoff for the players adopting them. For example, in the classic Prisoner's Dilemma (you know, two guys locked in separate rooms, each under pressure to cop a deal by testifying against the other), the optimal result is for both guys to keep silent and thus escape punishment. If both confess, both are equally screwed, but not as badly as if only one confesses. In that case, one guy gets a little time and the other gets lots more than he would have if he'd cracked too.

Prisoner's Dilemma may not seem like much of a strategy game. As a once-off, it's not, which makes it ideal for studying Games Theory, as opposed to say Chess Theory. The strategic aspect of the game emerges over a series of games. Each player can adopt only a single tactic in a single game. But if you're playing a series of games, you can try different combinations of tactics (strategy = combination of tactics), like "confess half the time", so that different combinations of "confess" and "don't confess" yield different payout structures for each player.

If you played a bunch of games like this, you would quickly realize that if either player ever confesses, both players will learn to get the best average score by always confessing. Of course this means a classic game of Prisoner's dilemma has all the dramatic tension of tick-tac-toe.
Classic Games Theory analyzes payout structures in terms of the optimal strategy for an individual player to adopt in a series of games. It evaluates strategies in terms like how long a player can stay in the game by following a particular strategy, or how much the player's final payout will be.

By varying the payout structure, you can change the optimal strategy for an individual to adopt. For instance, say the difference between "both confess" and "one confesses" is slight, and the punishment for the confessor remains relatively harsh. In this case, it will make sense to play "don't talk" more often than not. If the one who talks gets off with a light sentence, because the DA can take credit for a maximum sentence against the other suspect, it will certainly always pay to talk.

That's why cops do interrogations that way. It usually works, because without knowledge of the other player's choice or opportunity for collusion, it's in the rational self-interest of each party to confess. The game is rigged that way. House wins. Except of course the other players have their own way of rigging the game: Death to stoolies! Thus, strategies enter into evolutionary arms-races: Witness protection programs emerge, etc.

Now here's where ESS comes in: ESS looks not at populations of game-players, but populations of strategies. If players can change strategies over a series of games, strategies can be said to have lifespans. Some strategies will be adopted as useful by more and more players (reproduction), while others will be rejected (extinction). Players can also modify existing strategies to create new ones (variation). This means strategies are subject to evolutionary pressure, by dint of the evolutionary pressure on the players (who play longer the better their strategies). Strategies will develop meta-strategies, in order to compete with one another (like, "react to changes in other players' strategies," or "cheat").

Now here's something truly weird: Entropy enters the picture, as the deciding factor in evolutionary stability. To understand how, I have to take on faith something I've read, which is that the equations proving the Second Law of Thermodynamics have been translated from the mathematics of heat transfer to the mathematics of information, so that entropy is now defined as movement toward a condition of lowest overall information in a system. heat/information = 1. Weird.

If that thought didn't break your head (it still makes mine spin), we can apply it to ESS. In any series of elimination games, if there is a great disparity in the final scores, that requires a great deal of information. If everyone gets roughly the same score, that's much less information (for example, it would compress better in a zip-file), thus a lower energy state. So populations of strategies will tend to compete into states of equilibrium, where everyone gets sub-optimal scores, but the overall scores are the best they could be.

So is an Evolutionarily Stable Strategy is one that everyone in a population will eventually adopt? Sometimes. More often, the maximal payout comes with different proportions of the population adopting different strategies. Some babies are boys and some babies are girls. Some people become police and some criminals, some teachers and some lawyers. This fact impacts Evolutionary theory in important ways, for the details of which I'll refer you to Dawkins' books.

The picture of Evolution that arises from Dawkins' analysis of Evolutionarily Stable Strategies reveals that much of what happens in Evolution is not competition between alternative physiologies, but the ongoing evaluation of the comparative advantage between competing strategies. An organism able to rapidly adjust its strategy to circumstances has a huge advantage over one dependent on raw instinct or brute physiology. This fact predestines the development of intelligence, learning and culture as a fatal consequence of the Laws of Thermodynamics. Then Occam appears with his razor to cut God's hand off for stealing the credit.

The implications of competing behavioral strategies should be obvious for genetics, now that Dawkins has spelled it out for us: Genes will develop the ability to replicate strategies not just biologically, but culturally. Genes will evolve memes, as more efficient forms of information-transfer.

The social implications are less obvious. One upshot is that there's almost always a difference between the rational ideal strategy for a population, and the actual policy it implements. This is why the most utopian schemes of Post-Enlightenment Liberalism have failed over and over. It's is just like the fact that, though both players in a game of Prisoner's dilemma choosing "Don't Confess" always leads to the optimal outcome, a calculating player will always choose "Confess" anyhow. There is no more obvious example of this principle than our current Immigration Policy.