Post Docs in Computing

The CRA has released their report on the status of post docs in Computing. It should not come as a surprise that they are worried about the increases in the number of post docs, and the potential expectation that one is required prior to taking a full time job. They offer some nice advice on when a post doc is most optimal.

But the report avoids asking some of the questions that will have to be asked and answered if we are, as the report urges, to avoid the situation that has occurred in some other fields where post docs are required.

Most importantly the question that I think they have to ask is why has the number of post-docs gone up? I suspect the answer is because there has been a decrease in the number of permanent positions in academia, industry and government. The post doc is actually a unique situation in the labor market. It’s the place where the difference between the numbers of people admitted into graduate school and the number of employment opportunities post graduate school come together. And I think as the gap between those two numbers has increased, so the number of post docs has increased.

Thinking about this makes me think about incentives and rewards. Where in the system can we change the incentive and reward structure in such a way that we reduce the difference between those two numbers.

Another observation I’ll make is their remark about post docs being useful but not optimal for dual bodies. I think it would be instructive to collect data about whether a choice to optimize for location is actually simultaneously a choice to not optimize for career. What are the long term consequences for those who chose to do this in the short term. The reason I ask is because as much as it sucks to be apart, if it turned out in the long run it was more advantageous for securing two faculty positions say, then people might think about that. I say this as someone who did spend two years apart. I know how much it sucks. I also know that I do like that we have great balance in our careers, and I think some of that came from the time we spent apart when we were both pursuing careers as industrial researchers.

CRA Postdocs Report

The CRA recently released a report about post-docs for discussion among the community.

Why now?

In the last few years, the number of post-docs has risen dramatically within the field of Computer Science and simultaneously the number of tenure-track Faculty positions has declined. One concern that the authors have is that of avoiding the situation that the Life Sciences found itself in, having a large number of post-docs who had to hold repeated post-docs before finding permanent employment. They also suggest that Universities are looking for faculty candidates who increasingly have a C.V. commensurate with the experience that a post-doc would provide. Another interesting context they highlight is that by the time tuition is factored into the cost of a student, post-docs and graduate students become almost equal in cost to the faculty advisor sponsoring them.

There’s far more in the report, and I would encourage Computing researchers to read it, but I want to take up an issue that I think is highly related to the report but could use some examination. It’s the way that we, faculty, organize our disciplinary business and how that is rewarded. We have established practices that turn on the creation of groups of students, with whom we individually and collectively produce the research that leads to publications. Additionally, I think our culture of evaluation turns, partially, on where those students go, particularly if they choose to become faculty members.

But, what if the marketplace is changing? What if many departments are not growing much anymore? What if there is only one faculty position that a faculty member can place a student into, their own when they retire? That’s the extreme of course, but it is an interesting counter-position to what we have been used to, a culture of growth in Computing. Is it time to set expectations that alter the value of that placement, or at least calibrate it (particularly for newer faculty in the field)? Is it time to reconsider the size of a research group (i.e., bring the denominator down)?

I think there are other reasons to consider the way that we in Computing do disciplinary business. Recently, the NSF released it’s 2012 budget request and one of the things I was struck by in the report was the dependence of Computer Science on NSF funding. Eighty two percent of all Computer Science research is funded by the NSF. I am excited by the request for investment, I wouldn’t be in Computing if I wasn’t. But I keep wondering about the interaction between this discretionary investment and the American people’s desire to see the Government reduce if not eliminate the deficit.

Another is the continued declining enrollments in Computer Science education. As they decline, will the question be asked, how many people do we need to teach those who remain interested in Computer Science? I’m excited about the continued efforts to support the growth of interest in Computing related education. I think that there’s a really important set of futures in Computing industry, but we have work to do to convince people to take up those career paths.

One final reason of particular interest to me is that I suspect the labor market has changed in another way. The report mentions one reason that people post-doc, because they are part of a dual body opportunity and the post-doc helps the pair align their careers. I wonder to what extent dual body opportunities are becoming the new normal in hiring, and what implications that has in the short and long term. I don’t think I’ve ever seen data that tracks this, and yet, as a dual body myself, I know its implications for finding employment and the trade-off between career and relationship as the possibilities of long-distance are weighed against future aspirations.

So, I think this report is just about post-docs. It’s about the organization and culture of Computing research. And like any merit system, the easiest way to do that is look at what drives behaviour, the system of rewards. Without examining the culture of evaluation, and the consequences that it has for the way in which we do business, I think it’ll be hard to make change if that’s what is needed.

Publications and the System of Publications

Recently the Chronicle of Higher Education carried a piece entitled “We Must Stop the Avalanche of Low-Quality Research“.

Everybody agrees that scientific research is indispensable to the nation’s health, prosperity, and security. In the many discussions of the value of research, however, one rarely hears any mention of how much publication of the results is best. Indeed, for all the regrets one hears in these hard times of research suffering from financing problems, we shouldn’t forget the fact that the last few decades have seen astounding growth in the sheer output of research findings and conclusions.

Just consider the raw increase in the number of journals. Using Ulrich’s Periodicals Directory, Michael Mabe shows that the number of “refereed academic/scholarly” publications grows at a rate of 3.26 percent per year (i.e., doubles about every 20 years). The main cause: the growth in the number of researchers.

The article then largely goes on to deal with the papers. More, especially low quality papers lead to increased reviewing loads, as well as more variable quality knowledge that students have to manage in their own learning process. It makes the recommendation that we limit the number of papers that can be evaluated at stages (hiring, tenure, promotion). They also suggest publishing in high impact journals, based on impact factors, and changing (i.e. lowering the page count—which I think makes the suggestion that research can be reported in uniform length, but I suspect that different fields need different amount of space to accomplish their community communication goals). I saw a similar discussion at Snowbird also.

For me though, I was struck by the last sentence of that quote. I thought it was worth more attention than it received in the article.

The main cause: the growth in the number of researchers.

And recently I read a piece that talked about this, the number of researchers, in more detail. While I found the avalanche piece interesting, it was this second piece that I found very thought provoking. (Side Bar: I also sent it to my colleague Mark Guzdial who blogged about it).

The hypothesis put forward by Beryl Lieff Benderly is that the science gap (between the number of people who enter science fields and the number of jobs in science) is actually a product of the types of jobs that come at the end of the Ph.D. experience. The long, underpaid time spent in school is increasingly less likely to result in the very job that inspires people to go into science, i.e., an academic faculty position or equivalent.

I found the piece fascinating. I did not know much about Bush’s role in creating the post-war University expansion in the United States.

But the system had a basic flaw that was revealed only gradually, as the expansion of academe slowed in the early 1970s: The system’s central feature — the “self-replicating” professor who produces a steady stream of new Ph.D.s as a byproduct of grant research — had no control over the job prospects for those graduates.

and

Today, only a handful of young scientists — the few lucky or gifted enough to win famous fellowships or score outstanding publications that identify them early on as “stars” — can look forward to such a future. For the great majority, becoming a scientist now entails a penurious decade or more of graduate school and postdoc positions before joining the multitude vainly vying for the few available faculty-level openings. Earning a doctorate now consumes an average of about seven years. In many fields, up to five more years as a postdoc now constitute, in the words of Trevor Penning, who formerly headed postdoctoral programs at the University of Pennsylvania, the “terminal de facto credential” required for faculty-level posts.

This article left me with the following questions.

Is this beginning to happen in Computing? If the self-replicating professor model is accurate, then does it depend on a marketplace of constant growth (even if there is some retirement, and even if that retirement frees up resources that can be used to hire more than one new faculty member given the differential between start and end salaries)? What does an academic marketplace that is designed on stasis or very limited growth for a substantial period of time look like?

On the other hand I think it positions postdocs in a particular way. The first job I had, at Bell Labs, was initially a postdoc (that’s how they described the two year offer they made me, it was subsequently extended and converted to full-time employment). I knew when I left graduate school that I was not ready to take a faculty position. I felt like that four years later when I declined the first offer that Georgia Tech would make me. I still felt I had some things I wanted to and could best learn from Industrial research. I am glad for all my experiences prior to becoming a faculty member, even though it came at the cost of waiting 15 years to get tenure.

But I still think my questions are interesting ones to ask and answer. Especially for faculty members.