Posted by: jonkatz | October 29, 2009

Short-Term and Long-Term Academic Utility (guest post)

Another guest post by Yehuda Lindell. My reaction will be in a follow-up post.

Recently I thought about applying game-theoretic principles to our work as academics. What are our utility functions and how are they maximized? I have the following observations (if you don’t detect the cynicism below then this is a problem with written media):

  1. The first observation is that the “best strategy” is to write papers that are just good enough to get into the conference that you want, but no better. This way you minimize your work while maximizing your publications. “Best paper prizes” can somewhat improve this situation, but it’s not clear that going for “best paper” is a good strategy (it involves much more work, decisions about best paper are rather arbitrary so the chances of winning aren’t that great, and it doesn’t make much of a difference for promotion).
  2. The second observation is that you should spend as little time as possible writing. That is, your presentation should be as bad as is possible to have it accepted. Once again, wasting loads of time writing well just reduces your publication count.
  3. If your university doesn’t require that you have journal papers, then not only should you not write such papers, you also shouldn’t bother writing up full versions with full proofs. Specifically, write proof sketches that are “just good enough” to convince the program committee and don’t bother fully verifying. It helps to repeatedly use “the full proof will appear in the full version”, but then make sure that you never actually write such a version.
  4. I’m sure that one can argue that it’s best to research “easy questions” than hard ones and so on, but I’m not going to relate to this.
  5. Finally, you should definitely not waste time writing a book. It seems that a book is considered a nice addition, but 5 papers would probably do more for your promotion.

So, what’s the result of all of the above? You’ll have a nice long CV full of papers that no one will want to read, let alone follow up on. Is it really worth it? After 25 years of research, will you be able to look back and say that you had an impact? What about publishing 1-2 really good papers a year? I personally believe that this will yield much more satisfaction and impact. (I actually did this exercise and looked back over the last 5 years and asked myself how many papers that I published I actually really like and think had a real contribution. I reached the conclusion of 1-2 papers a year, so I’m happy with that. One could always ask me why I bothered doing the others. First, you don’t always know how things will turn out, and there are other reasons. But I don’t want to get too personal with myself 😉 .)

I also argue that the strategy above of doing as bad a job as possible without being detected (which is really what 1-5 lead to) may “make sense” for researchers who are very borderline and may not get tenure. However, for all the rest of the community, it makes no sense whatsoever. I would rather publish much less, but have people be happy to read my papers because they have full proofs and are easy to read. What I would really like to argue is that researchers who don’t need to follow the above strategies (and let’s hope that this is the vast majority) should make sure not to get pulled into such behavior. I strongly believe that it will raise your long-term utility!

Unfortunately, in my humble opinion there are too many people who do. Just for one example, I know of a number of great researchers with excellent papers and results that are ridiculously hard to read (and of course do not have full versions). I guess that this is my loss, because I usually don’t read such papers, but I assume that I’m not the only one…

Advertisements

Responses

  1. It seems that a book is considered a nice addition, but 5 papers would probably do more for your promotion.

    That is actually not true. Writing a good text book can make you way better-known than writing 5 extra papers, and being better known can result in better results, which ultimately will contribute to your promotion/tenure etc. This holds even more strongly if you are part of a large community.

  2. I actually meant better letters, instead of better results. Sorry.

  3. The phenomena you describe happen a lot less in pure mathematics than in CS, probably due to our (CS’s) habit of using conferences as venues of publication.

    (If there’s anyone out there who hasn’t see the big articles and all the resulting blogolalia, take a look at Moshe Vardi’s (“Conferences vs. journals in computing research”) and Lance Fortnow’s (Time for CS to grow up) editorials in the CACM.)

  4. I disagree with your game theoretic analysis…

    (1) I doubt that “counting number of papers in conferences” (which your strategy optimizes) is often used in evaluating researchers. Letters or evaluations (that do reflect quality) and citation counts (that also reflect quality) are more often used (in grants, promotions, hirings, etc). If anywhere, the sin of counting papers is usually limited to journals publications (in higher-up committees).

    (2) You seem to imply a correlation between quality of writing and journal versions. I certainly do not see such positive correlation, maybe even a negative one.

  5. Noam, I certainly disagree with my game-theoretic analysis. However, I also believe that there are too many people who do view that this is the way it works. This is what I was trying to say. (In addition, I believe that it is true that if you really are borderline, then this strategy may fool some committees.)

    Regarding journal versions and quality of writing: writing a full/journal version is certainly better than not. In addition, it is possible for referees to force authors to write a reasonable journal version. Although I agree that this doesn’t always happen.

  6. I think the utility function changes with time in one’s own career: counting papers at top conferences becomes less and less important the more senior you get.

    However, it is hard to argue that this is not important for *most* junior researchers, especially students. Students need good conference publications not only to compete for resources, but also to learn how to write papers, learn the norms and tastes of the community, get better known in the community, etc.

    This has implications: If you have students and you actually care about their future, then as an advisor you do have an obligation to work with them on research projects that are likely to yield good publications.

    This is of course not a defense of any LPU strategy — and we must teach students not to think in this way, by conveying to them how important “important” papers are if they want to compete for the best positions — but I think it explains certain phenomenon.

  7. […] articles on the state of computer-security research and some problems therein. (See here and the follow-up here.) I have my own thoughts on the general issue (which I will post about […]

  8. I really enjoyed reading your post, keep on posting such interesting articles!


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Categories

%d bloggers like this: