Excerpt from Last Interview with Jim March

Peter Ping Lee interviews James March. I provide here what I personally found the most interesting part of the full interview.

Peter: In your MOR 2005 article (March 2005), you referred to the peripheral position of the indigenous research communities relative to the mainstream research community. Do you think the Chinese research community is like that?

James: So far as possible, we would like to maintain some diversity. The ideology of research is international and sharing, but the risk is that you converge too completely and too fast. So how do you keep it from converging too rapidly and too completely? National and other communities being separate is our way to maintain diversity. It is a complicated problem because, from the point of the view of the separate community, that is not advantageous; they will be more advantageous if they converge with the dominant view; from the point of the view of the total community, there is an advantage having diversity. One of the ways to maintain that diversity in our present world situation is being national with a combination of separate cultures, separate languages, and some kind of local enthusiasm. Whether maintaining the optimum diversity is a much more complicated question, but a completely convergent is not optimum, so you need some diversity.

Peter: Is this analogous to the argument that if you don’t maintain exploration, then there is a tendency that the community will convert into exploitation?

James: It can be framed as exploitation-exploration issue. And that issue arises in all places, and in all places that I know, we have no optimum solution to it. We don’t know what the best mix is, but we know that is not the extremes. Diverse totally is not what you really want to be, but how much diversity you want is very complicated. Some parts are very simple. The longer you look ahead, the more you want diversity; the shorter your time perspective, the less you want diversity; that is fought over and over again. And some areas we have theorems that show that is true. The one I know best is the two-armed bandit world.
That is a set of problems that can be characterized as like going to a casino and confronting a whole array of slot machines. You know these slot machines have different payoffs. But you don’t know which one is the best, so you start experimenting. What is your search rule? After a while, you have found one that appears to be the best. And obviously, you will do well by repeating that rule. But when you repeat that one, you don’t search for any other ones. If you search for another one, you do less well in the short run, but you might do better in the long run. We don’t have any real solution to that problem. We do not know how to determine the optimum outcome.