Monday, January 6, 2014

Again on PhD numbers

Two interesting people have written two interesting posts about reducing admissions to grad schools. GMP has described graduate students who are smart, but not motivated, and so obviously have no future in science. Which poses a question: is it morally acceptable to keep them in grad school? And is it ultimately good for science? Maybe it is better to give them some kind of Advanced Masters, and let them transfer to industry (where they were apparently heading all along), instead of forcing them to go through this whole PhD experience. Why would you teach somebody to be an independent scientist if they don't plan to be an independent scientist? And why wouldn't you use the money that is currently spent on training for funding permanent scientists instead?

Another great post by Prof-Like-Substance is about how certain specialists, even within one sub-discipline, may have much better chances of employment outside of academia than other specialists. The question here is pretty similar: is it morally acceptable to let people invest their lives in potential dead-ends? Especially where there is, apparently, a viable alternative? Or should students be discouraged from entering certain fields, and be by force redirected onto more promising (or safe) tracks?

My general attitude towards this whole "surplus of PhDs" problem did not change: I don't quite agree that the situation is dire, and I definitely don't think that admissions to graduate schools should be halted. As I have stated before, while the existing "end-road bottleneck" of low job prospects for senior postdocs makes people anxious, unhappy and desperate, bottlenecks placed at the very beginning (hard admissions to grad schools) would discriminate against vulnerable categories of people; against all those who either start low, or think low of themselves, such as ethnic minorities, lower-income students, career-changers, women, foreigners etc., which is both unfair, and inefficient. People should be given chances to try things out. Late bottlenecks are not particularly bad, but rather all bottlenecks are bad in general. Ideally, scientific job market should allow promising scientists to gradually converge onto permanent positions. Lots of people should be allowed to try; lots of people should be allowed to fail, and while the judgment shouldn't be brought too early (lest it be arbitrary), it also should not be postponed for years (it is cruel).

Below I tried to illustrate some scenarios of this kind as "people sifts", or reversed pyramids. If everybody are admitted to the program and kept in it, but after 22 years of learning 95% of people get sacked (left shape), the remaining 5% will probably be very worthy of the jobs they got. But it would be an extremely inefficient, and, at the same time, a very cruel scheme, as everybody participating in this rat race will be extremely anxious and unhappy all the time. Moreover, many women, for example, would probably opt out of this race altogether, as they will reason at year 3: "Either I have a baby now, or never. But if I have a baby, I'll be at least a year behind all those males around me. I guess it's better to withdraw altogether, and find another job". Which would be double-bad, as it is, again, both unfair, and inefficient.

If, on the other hand, everybody are selected early on, and then the employment is practically guaranteed (right shape), all scientists (those who have made it) will probably be happy and friendly, but they will either all share rich parents and good undergraduate institutions, or will be chosen at random. Because when a person is 20 years old, and they studied engineering for 4 years, it is impossible to tell if they will become a good biologist or not. You will have to either use criteria that are predictive, but intrinsically unfair (such as their GPA and pedegree), or just throw a dice.


The only workable solution, in my mind's eye, is the shape in the middle, in which many are called, and at every step many are chosen, but lots of small decisions gradually make this groud converge onto a successful group of highly efficient scientists (unlike in the right scenario). Some people will, obviously, be disappointed, and some people will still feel betrayed at the very end, but most of them will escape rather early (unlike in the left scenario). I think a sift like that could work.

Now, if we draw modern-day neuroscience world in a similar way, how will it look like? According to my estimations, it will look roughly like that:


As you can see, in essence, it is pretty bad at both edges. It is not that easy to get into grad school, but at the same time the pyramid has a dead-end-style bottleneck at the very end. Postdocs slam at it daily, weeping in sadness and sorrow, and it is this sound that shapes the emotional background of the field. Which is not at all healthy.

What can be done? Let's smooth out the corners!


Accept more people into graduate programs, but, indeed, make Masters a requirement. Make PhD more competitive by design, not by personal failure of individuals, but at the same time, don't kill candidates early: let them try things out. Then shorten the postdoc, and compensate this loss in manpower by creating permanent (secure) research positions. Make the salary in these positions somewhat higher than that of a 3d year postdoc (because a person with 10 years of experience is usually more productive than a newcomer), but more importantly: make them secure in some way or another. Maybe let them be employed by universities, and then be distributed between groups in a grant-dependent manner. Maybe make them employed by NIH directly, and let PIs compete for specialists, and not for money. There should be ways. But overall, I think, these 2 changes: widening the base (aka "Masters", or "Research Assistant"), and the pinnacle (aka "permanent positions") would totally do the trick, and make people on average happier, while at the same time making the system more efficient.

Interestingly, recent changes in K99 rules not only don't push towards this scheme, but in a way push away from it, by creating an "alternative track" with early decisions and higher competition. Early-career decisions are inevitably biased by irrelevant factors, such as the parent's income (manifested by the choice of undergraduate institution), country of origin, or just pure luck (noise). Essentially, it is the "Right scenario" from my first figure, superimposed onto the "Left scenario". Not "combined", but exactly "superimposed", so that you have bad effects from both early and late bottlenecks, and good consequences from neither of them. Decisions are noisy, people are unhappy, efficiency is low, and fairness is really hard to achieve.

Fight for the changes! Join the Pyramid-Smoothing Party!