The insanity of research grant proposals

Most people will never have to write a research grant. That is a good thing.

How do you write a successful grant application?

  • Your work should follow established methodologies. It should follow closely prior work. Departures from either your own work or other people’s work will sink your proposal.

    Your results have to be predictable. Years ahead of time.

  • Also… Your work should lead to major breakthroughs.

It does not compute.

I do not care what kind of research you do: a predictable breakthrough is no breakthrough at all.

The good scientists always have speculative ideas. Sometimes these ideas come out of nowhere, in the moment. Most of these ideas are very bad… but a few represent the real breakthroughs. And that is what research is really about. Trial and error on a massive scale. You try things until it sticks. If you knew what you were doing, it would not be research. But that is not what you will find in research grant proposals.

What you find in grant proposals are soviet-like 5-year plans… any scientist that follows such plans is doomed to mediocrity. So, what do good scientists do? They lie about what they will do. To each other. All the time.

Published by

Daniel Lemire

A computer science professor at the University of Quebec (TELUQ).

13 thoughts on “The insanity of research grant proposals”

  1. You should be willing to divert from the plan as soon as something better comes up. That does not mean that it is not a good thing to do some planning.

    I think the granting agencies should put slightly less emphasis on the prediction of the outcome. The scientists need to know what they are studying, but as you point out they should not and cannot know the results before they start.

  2. Hence the time-shifted nature of the work by established researchers. At time t_0, you get your seed money, get some results, and propose to get them in a grant. By induction, propose to deliver results achieved in time t_i at t_{i+1}. Then actually deliver as promised, miraculously every time.

    Failure to launch with your seed money leads pretty surely to tenure denial.

  3. The same happen in large corporations too. If you want to start an innovation, you have to tell exactly how much time and money it will cost, and how much it will provide in return for the corporation.

    Which is obviously totally nonsense, who on Earth has a crystal ball ?

    The whole point of this strategy is to protect the budget approval process. This way, the board in charge of this approbation will safely be able to claim in the future :
    “We are so surprised, the project leader told us it would cost much less and gain much more !!! He lied, so it’s not our fault !”

    And that’s the main objective : shift blame. Yes, they want you to lie. They want it to be protected from their choices.

  4. _It should follow closely prior work. Departures from either your own work or other people’s work will sink your proposal._

    This does not fit my experience at all. If the reviews I’ve received on my own proposals are typical, proposals that include halfbaked ideas that radically depart from my previous work are much _more_ likely to be funded than “safe” proposals describing staid five-year plans with predictable outcomes. (And yes, I’ve had grants based on halfbaked ideas that utterly failed.)

    Your mileage, of couurse, may vary.

  5. @Corporate, this just happened to me last week. Great idea, but I couldn’t quantify both the effort and the results in time for budget submissions. BUT, I got a pat on the back and an ” ‘at a boy!” out of it — so i got that going for me.

    I guess an increase in knowledge is not a good enough ROI for most people with money, because the reason they have money in the first place is that they care about making more money. And, of course, making the kind of money that can fund research requires that you put a ROI value on everything you spend money on. Torus Knot completed.

  6. Absolutely on the mark.

    Now NSF proposals are shot down because you do not provide “sufficient preliminary work” as in NIH- one reviewer wanted details on how I would create the training set (where I have tens of papers that have used Machine Learning)! When you have to give nitty gritty details with specific details on algorithm you will use to solve a problem, how are you going to invent totally new strategies and unproven approaches?

    Capable scientists are giving up or busy submitting their proposal, so some warm bodies show up for reviews. I really liked the ideas Fabio Casati puts in his blog along this direction:

Leave a Reply

Your email address will not be published.

To create code blocks or other preformatted text, indent by four spaces:

    This will be displayed in a monospaced font. The first four 
    spaces will be stripped off, but all other whitespace
    will be preserved.
    Markdown is turned off in code blocks:
     [This is not a link](

To create not a block, but an inline code span, use backticks:

Here is some inline `code`.

For more help see

You may subscribe to this blog by email.