Tuesday, June 13, 2006

Sushi and Tit for Tat

Since I’m ostensibly thinking about evolution a lot of the time, I started thinking about how I decide where to shop and eat in terms of game theory. I think that, when I’m in the process of deciding which restaurants to go to, or which airline to fly, or any decision like that, I play a pretty standard strategy of Tit-for-Tat.

A lot of evolutionary game theory is based on trying to answer questions relating to the evolution of cooperation. Cooperation at first seems counterintuitive from an evolutionary perspective: selection acts on organisms, and organisms that are selfish will generally end up with more resources than organisms that are unselfish, and so end up having more and better offspring. Selection should then generally favour selfish individuals over altruistic ones.

But altruism, and cooperation between organisms are quite commonly observed. How does this evolve? One classic, and tractable way of approaching the problem is the Prisoner’s Dilemma. Briefly, in an iterated game of Prisoner’s Dilemma, two parties are subjected to repeated interactions with each other. Each round, they are able to select a strategy of cooperation or defection. The highest possible reward is for defecting when your partner cooperates, followed by both individuals cooperating, then both individuals defecting, and finally the worst possible outcome, cooperating while your partner defects. The object of the game is to select a strategy that maximizes your individual fitness over time.

The story of the original Prisoner’s Dilemma competition is really quite interesting, but I won’t repeat it here (read about it here if you’re interested). The relevant question is, what is the optimal strategy? If you know that your opponent will always cooperate, the best strategy is to always defect. Similarly, if you know your opponent will always defect, you should always defect too. But what if you don’t have any advance knowledge? What if your opponent’s strategy changes in response to yours? It turns out that a remarkably good strategy is a very simple one: Tit-for-Tat. Simply, a Tit-for-Tat strategy involves cooperating on the first turn, then repeating your opponents last move. If your opponent betrays you, betray them back next time. Tit-for-Tat is an evolutionarily stable strategy (or ESS), implying that a whole population playing this strategy cannot be invaded by individuals playing any other strategy.

Back to me and shopping. I was thinking that I’m just very boring and stuck in a rut of only going to a few places. I’ll give new places a chance, and if I enjoy them, I’ll return. As long as I keep having good experiences, I keep returning. But, if I’m unsatisfied, I don’t get mad, I just never return. So if I’m really playing Tit-for-Tat, it would seem that to get me back as a customer, a business would have to do something for me - a free pizza, or something like that. But even if I was offered something, I might just cynically take their offer, then never return anyways. So perhaps my strategy is “Cooperate until defected upon, then always defect, regardless of whether the opponent cooperates or defects”.

I’m not sure the analogy completely holds up - as a consumer, I usually have quite a bit of prior knowledge about a service before I enter the store. So I don’t just go out and try everything once, then only return to the places I like. Also, I have the advantage of knowing how likely I am to interact with that business in the future. In an iterated Prisoner’s Dilemma game, the players don’t know how many times they’ll interact in the future, otherwise defecting on the last turn is always the best strategy.

So maybe I am stuck in a rut of having a few favourite haunts. But at least this lets me feel like I’m just being strategic.

EDIT: There was an actual point to titling this post "Sushi and Tit for Tat", though I forgot about it when I actually wrote the post. The sushi place I've been going to recently always tosses in a few extra cucumber or avocado rolls. Awesome! Last time, however, they didn't. Do I keep going back, or have they betrayed me?

2 comments:

Len said...

As my philosophy prof once said, in reference to why social darwinism is flawed, free will screws up evolution. I think your restaurant darwinism might share a similar fate!

Len said...

Then again, isn't free will a product of evolution? Hmm...