Wednesday, June 22, 2005

Communicating Open Problems

A famous complexity theorist once said "The hardest part of being an advisor is not working on your student's problems." Good open problems are quite rare and one is often torn between the desire to see a problem resolved as quickly as possible versus giving people a fair chance to work on them. So I put together a set of guidelines for distributing problems.
  1. If you ask someone about what problems they are working on you shouldn't start working on those problems or give them to others without permission. When this rule is violated, even students in the same department are sometimes afraid to discuss their own research with each other.
  2. If someone discusses a problem with you shouldn't mention the problem to others without permission. Asking a question like "Do you mind if I tell this problem to my students?" is sufficient.
  3. If you are an advisor and you give a problem to a student you shouldn't work on the problem yourself or give it out to other students without the first student's permission.
  4. Outside the advisor-student relationship the above rule does not apply. You can work on a problem even if you give it to someone else or distribute it as you wish unless you've had a prior agreement.
  5. Once you make a problem public (in a talk, in a paper or on the web) the problem is fair game to all.
I realize I have not always followed all of these rules myself and I apologize. One could argue that one best advances science by making all problems as widely available as possible but following these guidelines will open communication as researchers will have less need to hide what they work on.


  1. Number 5 is set in stone and you will get no sympathy if it happens to you.

    Youngsters eager to talk about their work should decline any invitations to speak about ongoing work where publication is "just around the corner". The exposure of giving one more talk (especially with immature work) is not _that_ valuable, and compared to the possible loss if you get scooped, it can cost you a lot.

    I learned this the hard way: don't do it.

  2. Do you think the same goes for tricks of the trade? For example, tips on getting published passed down from advisor to student, and other political maneuvering that could be worth 10 years of experience? (Typically I'd give an example of what I mean, but I don't want to "give it away.")

  3. Number 5 is tricky though: you can minimize the scoop risk by tech reporting the work and disseminating it as quickly as possible once it's out there, so that the work is identified with you.

    Also, the more important the result, the less this matters, needless to say.

  4. Perhaps the need to hide the problems is because of the current way of evaluating a researcher by her publication list? In my ideal world we should be judged by our own contributions, in whatever form it is, to the community. And in that world, you can tell everyone about every problem because no one will judge you by saying you don't have enough papers.

  5. But the world ain't ideal. So wake up and follow Lance's tenets.

  6. It is not clear that an "ideal"
    world has no judgements or
    evaluation. Incentives drive a
    lot of things and without them
    much work won't get done.