Robin Hanson writes:
It seems obvious – in the vast space of interesting topics, academics clump around a few familiar themes, neglecting vast territories between the currently fashionable clumps. This is sure how it seems to outsiders and students, at least for fields like social science or literature, fields which must cover a vast territory. For example, economists have thousands of papers on auctions, and hardly any papers on romance, even though most people think romance far more interesting and important than auctions.
This is an interesting question. I don’t see it as special to academia–I’m not quite sure how to measure trendiness or clumpiness, but by any measure I assume it would be higher in art, literature, journalism, business, and lots of other fields. Back in the 50s and 60s, there were lots of westerns on TV, now not so much. And check out the Museum of Modern Art if you want to see trendiness. Or look at cars over the years.
However, academia is what I know best, and also there is some sense that academia should be held to higher standards, so Robin’s original question seems worth addressing. I have three main thoughts here:
1. The scientific landscape is fractal. When an area is studied in depth, it commonly spawns related research. From this perspective, I think clumpiness is inevitable. I think it’s misleading to think of research topics as being situated on a smooth Euclidean-type space.
2. I think there’s a feedback-and-overcorrection mechanism. Overstudied fields often seem to be things that were recently the hot new thing. For example, for the past 15 years or so, there have been a zillion statistics Ph.D. theses in genetics. Much of this comes from funding, I’m sure, but I think a lot comes because people (students and faculty alike) think of genetics as exciting and new, a lot more exciting than seemingly-boring topics such as sample surveys (which isn’t actually boring at all!).
3. Three words: Division of labor. Economists study auctions, psychologists study romance. Gains can be made by people working outside their fields (psychologists studying auctions, economists studying romance), but, by and large, it makes sense for people to work on their topics of expertise.
Auctions don't compare in complexity and fuzziness to romance. One analyzes what one can.
However, I believe a paper on the economic aspects of romance could be tremendously enlightening…
There seems to be a threshold, where a certain number of papers are published before it becomes a "serious" field (maybe the origin of life, in my evolutionary metaphor). Until then it is a risk to spend time on it, and it is a risk for journals to publish. Then there is a rapid evolution, as every possible idea is tried and it is easy to produce papers incorporating minor adjustments. Then the plateau or decline where various lines of research become extinct. I think this is what is happening with algorithms for gene clustering in bioinformatics.
It will be interesting to see how the economic aspects of everything trend goes. Often the conclusions don't seem to be testable, they are based on too many assumptions and the data isn't very good. Someone previously posted a link to a paper on HIV and risk. A person with a low life expectancy and income may be more likely to adopt riskier sexual practices, but is it some rational evaluation of utility or related to something else entirely? How robust is the model to changes in the assumptions? People acting rationally seems to be a very dubious concept, but still debated in the economics community. I will stop now, and consider the utility of typing this comment.
"Psychologists study romance"? Not really. One of my students complained that our department did not teach any courses on the really useful psychological stuff, such as relationships. He was right.
Robin's example isn't well-chosen. Could economists study auctions much more than romance because their tools are numerical and auctions provide many more numbers than romance?
For example, for the past 15 years or so, there have been a zillion statistics Ph.D. theses in genetics. Much of this comes from funding, I'm sure, but I think a lot comes because people (students and faculty alike) think of genetics as exciting and new,
or because genetics is generating a ton of data, which geneticists have no idea how to deal with. and <a>genetics and statistics have always been strongly associated.
Seth,
That might be a problem with your department. We have several social psychologists here at Columbia.
P-ter,
Good point. There's certainly a need for statisticians in genetics research, just as decades ago there was a need for statisticians in animal breeding and agriculture. Beyond this, though, I think there is some trendiness in what people view as a hot topic.
sure, sure, trendiness absolutely plays a role. but as a statistical geneticist (more geneticist than statistician), I can tell you that, if anything, there's a need for *more* statisticians who really understand genetics.
by the way, the link above was supposed to go to the wikipedia page on RA Fisher. I hear this galton fellow was interested in genetics as well :)