Last month I solved the problem that is American healthcare, if only in my mind. This act of mental self-pleasuring produced a feeling of exhilaration, let’s call it the economist’s high, that was the complete opposite of my usual oncologist’s dread. Chasing after another high, I found another area of need: science funding.
I won’t dwell on the many ways in which US science funding is broken, it has been covered time and again. This particular tweet is damning in the common sense mediocrity it reveals. And if you have any remaining doubts that funding mechanisms need change, this gut-wrenching thread will surely convince you that we are in trouble. But, even minor attempts to remedy it fail spectacularly under pressure from the old guard; and as ossified as the system is, there are just enough scientists trickling up from the early and mid-career pool into the multiple-R01 club to keep it going.
There is a way to make fund allocation more equitable for younger generations of scientists without messing it up too much for those who are already established — though after decades of same old, some amount of messing up is in order. It takes into account two factors which vary from early, mid, to late career: 1) how much time can the applicant afford to spend on the proposal, and 2) how certain can the grant committee be in the proposal’s impact. In the current setup, the presumed answer to 1) is loads, and to 2) completely and utterly. Of course, both assumptions are wrong.
Early-career scientists do have some time, not yet having fully formed labs or many ongoing clinical trials. In late career you have even more (but never enough: because there’s never enough time for begging for money). It is in the middle of the career that there’s no time to spare, in the midst of multiple projects, committee memberships, teaching, and, for physician scientists, clinical obligations (read this thread again for a good reminder). So in regards to available time: early career — some, mid career — none, late career — just enough.
The second factor, insight into the impact, is easier to plot because it’s linear: the more of a track record someone has, the better we are at predicting potential impact of their research, since most scientists incrementally build on their previous work (Linus Pauling notwithstanding, but look where that got him). So insight into the impact of an early-career scientist is low, mid-career it’s medium, and late-career — high.
One component is constant across the spectrum, and that is the number of grant reviewers: not enough, and the time they spend on reviewing grants: too much. Whatever the proposal, it should optimize for more reviews in less time.
Early career: The Lottery
The best predictor of future success is past success, say a million hiring guides, and the best way to get your project funded is to show that it’s already produced results. But you can’t follow this maxim with religious zeal and then wonder why innovation has slowed down.
It is, in fact, hard to impossible to predict which early-career scientists will do well, and their perceived past success has more to do with which institution they come from and who their mentor is than anything else. This is how well-intentioned meritocracy turns into a scientific bubble.
So why not have a mechanism that consists of a quick and dirty feasibility screen followed by a lottery. This isn’t a new idea, but since it hasn’t caught on it’s worth repeating: let’s stop pretending we know which young scientist with a new idea will do well, and let’s randomize their funding.
This is the feasibility screen: 50+ reviewers speed-read a brief (1-2 pages) writeup and answer a single question: Are there any red flags? (Yes_No_Undecided). Anything that gets more than 5% Yes is screened out. Everything else is fundable and gets into the lottery.
Funded applications get a second, more thorough screen. This could be just a deeper read of the existing applications, or submission of additional documentation, or even on-site visits. If any or all of those reveal inadequate mentorship and/or infrastructure, funding is withdrawn and additional penalties applied: e.g. the institution can’t have its candidates apply for the next year, the mentor can’t mentor anyone for the next 2-3, and the applicant can’t reapply for any grants for the next five. Queue self-policing and different institutional policies competing with each other.
Mid-career: The Anonymous Benefactor
This is where it gets busy: the lab is up and running, physician-scientists have a full panel of patients, there’s teaching and mentoring and sitting on committees, and if you haven’t had much of a family life but want one, this is for most people the last chance to start working on it.
So, plenty of time for making up stories of your future successes, right?
Fortunately this is a solved problem, and it was solved by Mr. and Mrs. MacArthur of the MacArthur Fellowship fame. Most people think that the common name for the Foundation‘s Fellows Program — the Genius Grant — came from the recipients’ superior intellect. I like to think it’s because of the ingenuity of the award itself: there are no applications, no study sections, no funding lines. Instead, a mysterious benefactor nominates you, an anonymous committee evaluates your work, and one day you get an email saying that Congratulations, you’ve just won more than half a million in funding.
Two questions arise: who nominates, and who approves? Whatever the answer, and it will take some trial and error to get to the right one, they should all come from a limited we’ll-defined pool, serve a limited term, and be anonymous to the awardees. The first instinct is to leave it to the existing R01 recipients and study section members, but that would be a missed opportunity: how about patients, industry, non-profits…? More variety won’t hurt, and with so many NIH Institutes there’s plenty of room for experimentation.
Late career: Status Quo
The current system is obviously working for some, or else they wouldn’t work so hard on resisting change. There should always be a safe haven for well-established researches to get money for incremental research. Yes, one could argue that they’re competing against the idealized versions of research going on in their colleagues’ heads; but at this point in their career they are the thought leaders in their fields anyway, so their colleagues better be thinking what they want them to think. Or so the thinking goes, anyway.
As the incremental system is best improved incrementally, I offer these suggestions: 1) lower the applicant pool by providing other funding mechanisms (as above), 2) lower the grant amount to discourage grand castle-in-the-sky (made of straw) projects, maybe by having the funding line set at 20 (30? 50?!) percent and adjusting the award size accordingly, 3) experiment with having post-funding site visits in the line of the NIH intramural programs instead of the grand prospective narratives that are half-done at submission time anyway, and with actual impactful work not even being a part of the grant.
Post scriptum: alternative takes
I wrote this mostly on the phone over a half-dozen subway rides and a few visits to the playground. For a better-researched and less gloomy take on life sciences see Alexey Guzey’s blog post on the subject. In fact, read the whole bog, it is outstanding.
I’ll mention this Twitter thread from Howard Crawford for the third time because even if the situation is better than people think, as Guzey suggests, they are still pretty terrible for individual actors, and far from optimal to boot.
EconTalk podcast has many episodes that touch the subject, but one with Patrick Collison gets right to the center. Collison’s article in The Atlantic he co-wrote with Tyler Cowen is a good companion to the podcast. It is on scientific progress in general, but funding is inseparable from progress and gets plenty of mentions.