Sunday, December 25, 2005

(long version of: an engine, not a camera. and not just any old engine, either.)

Consider the game of Go, which works better than chess only because computers are not now better than the best Go players and that point might contaminate intuition when thinking about chess. (You don't have to understand anything about Go for what follows, other than its a game where players take turns and no random element--e.g., dice--is involved.)

Imagine there was a academia-based enterprise called "Go Science." Along with whatever else, Go Science might have two projects, with the respective central questions being:

1. How do people who actually play go decide what moves to make? (descriptive)
2. How should people who play go decide what moves to make? (normative)

You can imagine the debates in Go Science about what real Go Scientists should be spending their time focused on, etc.. All the while, there is also this thing in the world that is the world of competitive Go, with big prizes to the winners.

You can also imagine that for the descriptive projects of Go Science the explanations for why people make the moves they do will fuss much about the distinction between: (1) moves are made because they are the optimal move if we assume people are trying to win the game, (2) moves that are suboptimal. Obviously, one can imagine a sprawling array of studies to understand how people come to make the optimal move in those times they do make it--in other words, optimal move-making is itself a thing to be explained. And, even more obviously, one can imagine the all the work of trying to understand why people, despite that they want to win and thus want to make the optimal move, fail to make the optimal move on the table.

It could be that, from the standpoint of understanding why players do what they do, the best route to understanding is not to make use of the information that the move the player made was the optimal. Like maybe people's styles of Go playing vary dramatically depending on whether they are from Japan or the US. In a particular situation the Japanese player makes the optimal move and the US player makes the suboptimal move, but in other situations the US style leads them to make the optimal move while the Japanese player would make the suboptimal move. Understanding why each made the move they did in some situation is to be explained by this cultural difference in style, an expalanation which works the same for the optimal and suboptimal case, and indeed the fact that the move the Japanese player made was "optimal" can actually be misleading if introduced into the explanation as the cause of the behavior.

Of course, one could only make this mistake of invoking the fact that the Japanese player's move was optimal as a cause of the player's behavior if there was some way of identifying the move as being, as far as could be discerned, optimal. Which is exactly what the normative arm of Go Science is trying to figure out.

There is, of course, a huge catch to all of this. If the normative arm of Go Science succeeds in figuring out the optimal move in a given situation where the US player makes the suboptimal move, then we might expect US players who hire Go Scientists as consultants will pick this up and start making the optimal move instead. And if the Japanese players didn't likewise make use of Go Scientists to figure out and provoke change in cases where they were making the suboptimal move, then we could expect the US players to start beating Japanese players regularly.

In the end, if Go players are capable of playing to the state-of-the-art of normative Go Science, then that is exactly how they will play. For the parts that Go Science has worked out, there is no individual 'style' of play if everyone is playing to win. Go Science, in this scenario, comes to "perform" Go, as the explanation of why a given move is performed in a given situation is to be found in the science, not the player who wins by allowing herself to be performed by the science.

Sure, maybe Go Science is wrong in its judgments of optimality sometimes. The error either goes undetected, or some enterprising Go player figures it out, and wins a big tournament because of it. And thus Go Science would change. Either you follow Go Science strategy, or you make a short-term gain but thereby contribute to Go Science.

The state-of-the-art findings of Go Science would result for some social process and are thus "social constructions," and yet explaining why these findings are state-of-the-art instead of some others would run up the issue that, as far as anyone can discern, they are right. One could imagine critics insinuating that Go Science is really a collective delusion, perhaps even using all kinds of complex prose in offering their arguments, but these critics would always be held in check by the point that they can't actually provide an alternative that wins on the board.

Go Science is made possible by the rules of Go. Indeed, the fact that Go Science is today a very uncertain endeavor and Chess Science is instead a successful project is to be found in the greater complexity of the former than the latter. Tic-Tac-Toe Science is something for which you figure out all its findings for yourself by the age of eleven. Meanwhile, something like Charades Science would certainly be able to provide people with strategy for better clues, etc., but you can't even talk coherently about one clue being best in all situations, etc.. Dating Science would be even harder, even though anyone could identify some things as better to do on a date than others, because of the fundamental divergences that exist outside of fraternity houses about what constitutes a "winning" date.

Economics is like a successful normative Go Science, and it works very well in situations where decisions are like Go. A question about economics is the extent to which its existence and success--or the success of the broader changes in cognitive technology of which advances in economics are part--result in the world reconfiguring itself to be more like a series of games of Go as opposed to being a series of games that range from Tic-Tac-Toe to Dating.

In other words, to what extent does an apparatus for figuring out the right move in well-defined but tricky situations put into play social processes that increase the extent to which the world is reconfigured to present actors with a series of problems that are well-defined but tricky? That is, does it increase the number of situations for which sophisticated levels of rational decision making become a handy tool indeed? One may even be able to imagine increasing kinds of differences in the fates of individuals based on their success in assembling patterns of behavior consistent with what economics would recommend versus those who, for various reasons, are not very good at assembling behavior patterns like this.

Although I realize no one is still reading this post, the last two paragraphs basically characterize the way that I think this century will go. Put together the three key components that protect the rational actor from predation: low preference malleability, high ability to make use of information, high time-consistency of preferences. These things should not be thought of as personality traits, necessarily, as people can change their environments to effectively make themselves better at these things despite themselves. But they basically comprise the three main fronts for the war of the colonization of individuals by various kinds of "consumerism," and, while already important, variation in these effective achievement of these three things will be increasingly vital for understanding why some people's lives turn out better than others.