During my time leading the Biological Sciences Directorate at NSF, I learned that the agency spends around a billion dollars a year on graduate education—the training that is required after the undergraduate degree to turn an aspiring scientist into a true discoverer. Of that money, roughly 15% is spent on NSF’s flagship graduate research fellowships—a fantastic program that’s been around since the 1950’s and has played a central role in the early careers of many of the US science superstars. These are folks who have gone on to win Nobel prizes and the like. Winning these fellowships involves an intense competition of ideas and is peer-reviewed by the science community. I’m pleased that NSF tracks the career trajectory of these trainees pretty carefully. There is hard evidence that the graduate research fellowship program works.

Another 5% of the total is spent on trainee grants—the current version of these are the National Research Traineeships. These are training awards that go to universities which then award the support to graduate students that they select. I was trained under such a program (although it was NIH funded) when I was at the University of Michigan training in neuroscience. These are excellent funding programs and once again those folks who are supported in this way are tracked pretty carefully (I still get contacted regularly by the NIH asking what I’m up to).

But the vast majority, 75%, of what the NSF invests in graduate education is untracked. These embedded in the dollars that go to research grants of scientists at US universities who then hire graduate student research assistants to actually do the work. We don’t know what happens to these trainees. There simply is no easy way at getting at the data.

In strikes me as unwise to make such a large investment without getting feedback on how things are going. In particular, I am concerned that those graduate students are being inadequately mentored in some pretty substantive ways. For example, I fear that they are too often treated as an extra pair of hands rather than a future professional colleague. Time spent teaching these students about career options or how to effectively teach undergraduate students is time away from the laboratory bench.

There are ways of tracking such students. One such mechanism is the orcid id system. There are others. If all students supported by the NSF were required to be registered in such a system, then it would be possible to track their career easily (as long as they stayed in science). But success on that front requires one other thing: that journal publishers and data repository sites require that a person’s id be attached as meta-data to every single piece of scientific data from the results of a single bench top experiment or a field observation all the way to a finished journal article.

This is not impossible. I think it is important to move this direction because it will allow for evidence based decisions about how to optimize NSF’s graduate student support in the future.