Depth vs novelty in research: differences between disciplines and across time.

I think that, with some degree of approximation, we can summarize the quality of a piece of research by two variables. The first is the novelty of the research question asked. I call this variable n. The second is how exhaustive the answer to this question is. I call this variable d for depth.

We can think of the importance of a given piece of research (call it V for value) as determined by both n and d

V= α n + d
where α determines the relative importance of novelty vs depth. V in turns determines the standing of a specific piece of research: how well is published, how widely is read, its influence on subsequent works and so on.

I think that α is discipline specific. For example, papers in marketing, strategy, oganizational behavior usually ask super interesting research questions. To my eyes, however, the answers to these questions are often highly incomplete. My interpretation is that these disciplines have a high α. Similarly for psychology: a super interesting research question followed by an experiment with 10 subjects. On the other hand of the spectrum I would put mathematics. Most ground-breaking, super influential math papers provide very detailed answers to well known puzzles. Not only, but mathematicians have the habit of throwing math puzzles at each others (sometimes via blogs), as if the novelty of a research question is not particluarly important to them, but providing the answer is. Using the above framework, therefore, I can say that math has an α close to zero. Economics (my discipline) is somewhat in between: both the novelty of the research question and the depth of the answer matter in how a piece of research is evaluated. As a consequence, if a researcher thinks that he/she has stumbled upon an extremely novel research question, he/she will probably not blast it to the world without first also having produced a research paper (of course, exceptions to this rule exist!). At the same time, research papers often have endless appendixes, that are supposed to prove that the results are actually robust.

Before I say anything else, it is important to clarify one thing: in every discipline there are research papers that are both extremely novel and extremely deep (maybe yours!). Those are the top papers: they have very high V and are extremely influential. But to think about α, you need to think about the papers that are just below a given threshold (for example, a threshold for publication). Then you have to ask: is it more likely that this paper crosses the thresholds if it improves on the n dimension or on the d dimension? The point I'm making is that the answer to this question depends on the discipline we are considering.

To some extent, the specific tools employed by each discipline are actually a function of α. Taking this logic to its extreme, we can say that mathematicians are a group of people with a very low α; and as a consequence they employ math. Economists have, on average, an intermediate α. As a consequences, economists use math and statistics in a somewhat rigorous way, but are willing to cut some corners (relative to pure mathematicians) in order to provide an answer to a question they think is interesting. Other disciplines have an even higher α and therefore are happy to use case studies or work with very few observations to answer their questions, provided that those questions have a high n.

Finally, I think that α is also time specific, that is, there are subtle shifts in α over time. These shifts determine subtle changes in the type of research that is read/published in a given discipline, and in the methods used. If I had to take a wild guess on where we are heading with α, I would say that it is increasing over time: novelty will become more important. I say this because we live in an era in which information (including scientific research) is almost completely freely available. Hence, the limiting factor in the consumption of information is not the availability of information itself, but rather the availability of complementary inputs such as attention and time. Obviously, attention has more to do with n than with d: I'm more likely to read past the title of a paper if I think that the research question is interesting.

Does all this matter? Well, it matters if you are a researcher, especially a young one. You should know what the α of your discipline (or your subdiscipline) is and where it is heading, and write your papers accordingly. Second, it matters for the general direction of research. If α is indeed increasing, then we may be heading to a world in which a lot of interesting questions are being asked, but not very many deep answers are given. How does such a world look like? Well, this is definitely a very interesting research question!

p.s. Of course, the assumption that we can describe all research in all fields by simply 2 variables is quite heroic. In particular, depth may mean different things in different disciplines (number of equations, number of observations, length of the questionnaire, ...). So not only alpha changes with discipline/time, but also how we measure d. But, hey, this is a blog post and therefore mostly about n than d!