Daniel Lemire's blog

, 2 min read

The hard truth about research grants

You must do many silly things to get a large research grant:

  • You must know precisely what you will do__ for the next five years____.__ Yet, in my experience, good researchers only have a vague idea of where they will be in 5 years. If you know the promising research directions you will encounter, you are an astrologer, not a scientist.
  • Your plan must be infallible. Yet, in my experience, research ideas that never fail are not research ideas at all! They are mere applications of known principles. There is no research without risk. The more ambitious the research, the riskier it is.
  • Your work must revolutionize your field. Yet, as stated above, it must also be infallible.
  • Your work must have crucial applications. Make it up if you must! When Einstein came up with E=mc2, he should have pointed out the possibility of dropping nasty bombs on Japan. Again, we expect scientists to know the future.
  • You must work in large teams, packing up a broad range of researchers. Yet, there is no clear evidence that correlation exists between the resort to extramural collaboration and overall research performance. The goal—once more—is to minimize risks. It is unclear whether it actually reduces the risk at all. And the cost is added bureaucracy.
  • You must promise to train many new Ph.D.s. Somehow, flooding the job market so that there are hundreds of qualified applicants for every research job is highly rewarded.

Thankfully, Peter A. Lawrence has many good ideas on how to improve the system:

  • Grant applications should be short. Quick to prepare. Quick to process. Currently, it is not uncommon for scientists to spend months writing grant applications, and weeks reviewing them. Wasteful!
  • You should be able to apply for grants based on your record alone. The requirement to____ submit a project or research program is silly.
  • We should be critical of large groups. For some large projects, such as building a multi-billion-dollar machine, you need a group. To investigate the properties of some algebraic structure, you don’t.
  • You should only offer a handful of research papers for review. Pick your top 3 papers and don’t mention the rest. In Canada, NSERC applied this idea, but they did not go far enough. They ask us for our top contributions, and then they ask for the full list. By reviewing researchers on only their best work, we would lift the incentive to produce many weak papers.

I conclude from a quote by Michael Nielsen:

Some years back I constructed a list of papers I especially admired, and was surprised to discover that with only a few exceptions they were produced from unfunded research. This was sobering, since it suggest that receiving research grants was (…) anticorrelated with doing work of the highest quality. Grants seem to be good at sustaining an established area, but not very good at all at producing the conceptual innovations that start new subfields. (Michael Nielsen)

(And before anyone jumps at this… Michael did not write that having a large grant was incompatible with highly original research.)