Wednesday, October 24, 2007

FOCS VII : Funding? NSF? whats it all about?

(Another Guest post from Nicole!) More from the Business Meeting

Funding status report: I'm not really the right person to blog about funding since I have never yet applied for a grant. From the business meeting, it seems like a lot of people are doing a lot of hard work to funnel more of the NSF budget into our hands. They seem to have had some success. There are two new initiatives -- CDI, Expeditions -- and several old ones which now have more money than before -- ToC, SING. Funding rates seem to be around 25% (33% for CAREER). (An aside, maybe we should also try in an organized way to funnel more of the federal budget into the NSF by, e.g., writing our representatives in congress -- you don't have to be a citizen to do this -- or through PR campaigns.)

But you can find all this info on the NSF website. What you can't find on the NSF website is how it affects us young academics. I am about to start my first faculty job, and funding has suddenly become a very real issue for me. I see these numbers, I hear about these politics, and I am totally in the dark. I have never written a proposal, never even read a proposal. I don't know how to play the game. I know my senior colleagues will help me through this process, but that does not reduce the anxiety. I suppose every career path has rites of passage like these. The thing is some rites of passage are fun. I could be wrong, but I'm not looking forward to this one. ~

12 comments:

  1. I have never written a proposal, never even read a proposal. I don't know how to play the game. I know my senior colleagues will help me through this process, but that does not reduce the anxiety. I suppose every career path has rites of passage like these. The thing is some rites of passage are fun. I could be wrong, but I'm not looking forward to this one.

    It would be better if they just gave you money based on your track record.

    Grant proposals take away your freedom because it is difficult to change direction later if you lose interest in what you have proposed.

    Also, maybe the people reviewing your grant proposal would rather fund research that would likely result in more citations to their research. Or maybe they would like to keep their competition down. Or maybe they just don't care enough to give it serious thought.

    ReplyDelete
  2. Obviously, anon #1 has never served on an NSF panel.

    The grant process stinks, but not for the reasons anon #1 mentioned.

    ReplyDelete
  3. The grant process stinks, but not for the reasons anon #1 mentioned.

    Could you be more specific?

    What's wrong with funding researchers based on their track record?

    ReplyDelete
  4. In 2004, according to NSF figures, the United States graduated about five science PhDs for every one engineering PhD. A pretty strong economic case can be made that this ratio is upside-down.

    ReplyDelete
  5. I've seen some very bad proposals by some very well known people.

    I've also seen some very good proposals by some people I never heard of.

    What's wrong with funding proposals based primarily on the quality of the proposed research, with track record taken into account as a secondary measure?

    ReplyDelete
  6. What's wrong with funding proposals based primarily on the quality of the proposed research, with track record taken into account as a secondary measure?

    Why do you want to take away the freedom of researchers to change their minds as they wish?

    It's hard to predict whether proposed research will work out. Why even bother?

    ReplyDelete
  7. The argument that "proposed research doesn't work out" is sort of bogus: for starters, I am not aware of NSF ever giving anyone trouble for working on whatever they want after receiving a grant (I'm not sure what NSF official policy on this is, though). Second, proposals are usually very general: you don't typically propose to work on one problem; you propose to work in an area. So if one problem doesn't work out, you can move to another one in that area.

    Finally, NSF does not "blame" you if you seriously work on a problem and simply don't get the desired result.

    ReplyDelete
  8. The argument that "proposed research doesn't work out" is sort of bogus: for starters, I am not aware of NSF ever giving anyone trouble for working on whatever they want after receiving a grant (I'm not sure what NSF official policy on this is, though).

    Then why bother with the proposal at all? Why play these games?

    From your publication record, they can decide what sort of area(s) you belong to. That should be sufficient.

    ReplyDelete
  9. I have seen grant proposals. Most are worthy of funding and have enough flexibility for the investigators. A narrow proposal is anyway not likely to win a grant. Make sure your proposals are broad but still clearly makes out the possible gains for the world is the proposal is funded. narrow proposal, like I will resolve P = NP question is not likely to be funded. A broader proposal that I will study the relationships between P and NP among other complexity classes is more likely to be funded.

    When you are asking for money from somebody it is your duty to set the expectations. It is your reputation to meet the expectations you set.

    There is no other way.

    If I review a proposal and do not see the expectations are being set clearly, I am likely to give it a low score on one or more count. I would love if there is a post proposal review, sort of assigning an internal reputation score. That is what expectations are set and what was met. What attempt were done in trying to meet the expectations. Whether money is used judicially. I would held both the investigators and the host university responsible or rewarded if the expectations are missed or exceeded in terms of lower/higher reputation score. That will not only set an incentive structure to do the right thing, but also make sure that the funding goes an extra mile.

    ReplyDelete
  10. What's wrong with funding researchers based on their track record?

    What track record? Most researchers don't have a public track record that can be evaluated in detail by specialists in other areas (and it's impossible to ensure that each NSF decision is made by someone working on exactly the same stuff). There's some information that can be learned from where you work, what prizes and awards you have won, how many papers you published in top conferences, how often your name arises in hallway conversations, etc. However, all this information together really doesn't do a great job of determining who should get grants. Plus jobs and prizes are so heavily quantized and have enough arbitrariness that they can lead to grossly unfair judgements.

    In practice, NSF grant proposals hardly limit the freedom of researchers at all (nobody penalizes you if your research plans change - the only issue is that if you consistently do worse research than you had proposed, you will find it hard to get more grants). The only real drawback is the time they take to prepare.

    The huge benefit is that (well written) proposals articulate the motivation, context, importance, and plausibility of your research plans, so the review panel can judge it. Typically, the panel might include a specialist in your area, who serves as a sanity check to make sure you're not bluffing, but it's up to you to help him/her convince the rest of the panel that you deserve support.

    This has some big advantages compared to a pure track-record-based approach. I've seen occasional grant proposals from people with good but not amazing track records who had recently had wonderful ideas. They would never get grants based purely on their past results, but the proposals make a compelling case that they are on to something great. Conversely, I've seen proposals by famous people who get prizes and tenure at great places many years ago, and who propose to continue doing exactly what got them to this state regardless of how the field has changed. Occasionally, this is brave and appropriate, but more often it just shows that even brilliant people can stagnate and get out of touch. Any system that would award grants to this second type of applicant but not to the first has problems...

    ReplyDelete
  11. So if you would like to do research in an unpopular area (say Wolfram's NKS), then you should write a fake proposal to increase your chances of funding?

    Also, there's an assumption here that the NSF panel knows best. Why do you think this is so?

    ReplyDelete
  12. If you want to work on one single problem, say P=NP then a proposal must establish the following:

    Why your proposal is credible? What new insights you have or recently developed by others? What techniques or methods you would try?

    Panel wants to know these things because in case you fail in your attempt, panel would want to know that science would still be advanced. For an example even if you did not succeed in proving P!=NP or P=NP, you would have established some interesting results on the way. Besides being of independent interests, these results may help future investigators to settle P!=NP question.

    If you just want to write a one line proposal, P!=P then you have to ride a lot on your past reputation.

    Ofcourse panelists are not perfect. Evaluating a proposal is like making a prediction. Nobody could make a perfect prediction. If somebody could, panelists would want investigators to make the prediction. That way you would not need panelists at all or need them to lesser extent. For an example, you may predict that you will prove P!=NP and willing to accept it as a felony if you intentionally made the wrong prediction.

    ReplyDelete