Google Analytics

Thursday, February 07, 2013

Postdocs in Computer Science

Anita Jones is troubled by the growing number of postdocs in computer science, she uses "troubling" twice in the first paragraph of her CACM Viewpoint. But is it really a troubling trend or just a natural outgrowth of a maturing field?

Theoretical computer science leads computer science in having and even embracing a postdoc culture. Nearly every graduating PhD in theoretical computer science that remains in academia takes a postdoc position before taking an tenure-track job. If anything I hear theorists lamenting a drop in theory postdocs this year with the end of the Simons postdocs and CI fellows.

Postdocs give PhDs a chance to focus on research and strengthen them for the future job market. I initially started as a two-year assistant professor at Chicago, basically a teaching postdoc. If I didn't have that opportunity my research career would have died in its tracks.

What would happen if all postdoc funding was stopped. That would lead to more funding for graduate students most of whom would have to take a non-research career especially with no postdoc positions available. Hard to see how anyone wins.

Anita and I both agree that a successful postdoc experience requires strong mentorship and inclusiveness or otherwise the postdoc is just working in a vacuum. The CRA has put out a best practices memo, worth reading for both postdocs and the people who hire them.

14 comments:

  1. I agree that an increase in the number of postdocs is a natural outgrowth of a maturing academic field. However, it raises the question: once a field has matured to this point, should we be training so many PhD students that there are not enough faculty positions for them? (This question only refers to fields where the majority of PhD students believe they want to stay in academia, which I believe includes many of the sciences, as well as mathematics and TCS.)

    How many postdocs (in, say, CS, or even TCS) end up staying in academia? How many dwindle for a long time in what might be called "postdoc purgatory" only to finally give up on academia because of the lack of available faculty positions? (I am asking non-rhetorically - I do not know where to find such data.) Note that for TCS I'm happy to count permanent positions at Microsoft Research, IBM Research, or a national lab essentially as faculty positions.

    ReplyDelete
    Replies
    1. How is being unprepared for a job in academia after getting a Ph.D. a sign of a maturing field? It sounds like a sign of more cheap labor being cultivated by established academics.

      Delete
  2. I don't see how removing the postdoc results in fewer people leaving academia -- there should still be the same number of academic jobs (if anything, perhaps slightly more to make up for the fact that there are no postdocs teaching). The only thing it changes is how long those people who eventually get academic jobs spend struggling in low paying positions.

    ReplyDelete
    Replies
    1. Right to the point. Multiple low-pay short-term post-docs is a problem. Maybe one 2 year post-doc is fine but no more.

      From what I see in some other sciences with high number of post-docs like chemistry post-docs are not the same as post-docs in computer science. They are rather stable mid-term or long-term jobs with reasonable pay, working in conditions similar to industrial research labs as a junior researchers. In computer science that is not the case.

      Delete
  3. It seems she is worried that if post-doc becomes *required* before people are hired for tenure-track positions, it's like saying the PhD is not good enough, it's somehow gotten watered down. But is that true? I'm guessing there are many more good PhD dissertations (and more so-so ones) than in the past. I'm guessing that the demand for PhD-level research has grown faster than the number of teaching positions, which makes post-doc a good place for people to land, initially, after devoting so much of their time to getting that PhD. I'm guessing it's easier for a CS PhD to get a good job than a Physics PhD, even though Physics is core to understanding so many things over all those powers of ten big and small.

    ReplyDelete
  4. Any significant chance that the number of postdoc positions (say the Simon postdocs) will increase next year?

    ReplyDelete
  5. Anonymous #1 (9:16 AM, February 07, 2013) here again: I wasn't suggesting doing away with postdocs. I was suggesting/asking whether the need for postdoc positions should highlight a need to be training fewer grad students, which might eventually lead to a kind of equilibrium where postdoc positions were not needed. I think this is an unlikely outcome, because of the dependency of universities on graduate students for teaching assistants and the dependency of professors on graduate students for an influx of new ideas and extra minds at work (and also, less cynically, because professors actually like mentoring graduate students). But I was just saying...

    ReplyDelete
  6. I share similar concerns, I wonder whether we're going to overproduce in Computing and end up with similar difficulties other fields have faced when students discover that after 6 years in graduate school, and then another 4 years in a series of postdocs that there are still no faculty jobs. I think we have a moral obligation to students. We can't guarantee them jobs, but we shouldn't keep them in a labor market that gives them no retirement plans, little stability for over a decade should we. Is even a decade too long?

    ReplyDelete
  7. TCS has had a postdoc culture for a long time. Back in the 1980's when I graduated, there were a small number of places where many of the best people did a postdoc for a year or two prior to taking a faculty position. e.g. Berkeley, MIT, IBM, AT&T/Bell Labs, Toronto, etc. Many people were also hired right after PhD, but these postdocs were considered plums rather than something to do while in a holding pattern. I think that is still true for many, though certainly not all, TCS postdocs today; many of the industrial ones pay better than starting faculty positions do.

    The concern that Anita Jones has fits more with the problems of postdocs in the biological sciences where one needs to do a postdoc as a minion in someone else's lab for years until one can get a faculty offer. (Candidates need to build up sufficient clout that they can convince someone that they fund their own lab.) I know of someone who did a postdoc for 8 years before taking a faculty position, though that was an unusual case. You can imagine that being an issue for big systems projects, but we seem to have avoided that so far.

    She is right that the uncertainty of postdocs places a particular burden for those with families or wanting to start them. The part-year postdocs that are being tried out this year in theory seem particularly problematic in this regard.

    ReplyDelete
    Replies
    1. Paul, the statistics regarding research funding in the biological sciences are more sobering than your anecdote suggests: female biological fertility ceases before NIH funding viability commences (the graph shows the age of the investigator at the time of receiving their first independent funding).

      In the foreseeable future, there are few prospects that this situation will improve  … and many prospects of continued worsening.

      It is fair to say that back in the 1990s, few or no biomedical researchers foresaw this dismal future. That is why arguments of the form "It can't happen to [insert your favorite STEM field]" carry little force.

      As (Intel's) Gordon Moore says: "You never get well on your old products." Perhaps STEM academia needs new products?

      Delete
  8. The question of whether post docs are good for computer science is simple and the answer is yes. Young researchers have more time to hone their skills before being bogged down by the myriad of responsibilities of a professor and universities get a stronger pool of applicants for their tenure track positions.

    The question of whether the post doc culture is good for computer scientists is a more difficult question. It seems frustrating that after spending five or six years getting a phd, there is another couple of years of comparatively low pay and uncertainty in future positions. For people with significant others or families, it seems selfish to ask your family to deal with your reduced ability to provide for them and to interrupt their lives to move with you to your new institution, which you may leave again in a short time, as you pursue your dream of becoming a professor.

    As a field, we should then consider which of these facets we value more or potential middle ground solutions.

    ReplyDelete
    Replies
    1. The question is not if doing one or two post-docs are good for those who find faculty positions, the question if it is good for those who don't find such positions.

      Cheap labor seems good for any business in the short run, and post-docs are cheap labor for sciences. However studies show that productivity and quality drops considerably as a result of working in bad conditions. If a post-doc has to look for finding his next post-doc just as he starts one and at the same time worry about supporting his family then you can guess what would be the product of his research. People are more productive and create higher quality output if they work under conditions where they can focus on their work and not worry too much about other things.

      Delete
  9. Lance is right. If the number of post-docs is cut the funding is likely to be redirected towards getting more graduate students. The root of the problem is accepting too many graduate students with the goal of obtaining an academic job while the number of positions is relatively too small.

    As long as we don't limit the number of graduate students that we accept or significantly increase the number of faculty positions or change the perspective new students regarding obtaining an academic position the problem will continue and worsen. Eventually many bright people who want a life beside research will avoid our field and we will not enjoy a good reputation as we do today. That will be a equilibrium that we are moving towards without any change. An alternative would be limiting the intake of graduate students. Business and medical schools limit the number of their students even though they have funding for many more.

    It should be clear for someone who finishes a PhD with a reasonable certainty that if they continue the post-doc path they will get an academic position or not. Right now, with the exception of exceptional young researchers like Ryan Williams and Mark Braverman the academic job market looks like a lottery with a large jackpot. There are many good suitable researchers and each of them has a small chance of getting an academic position.

    ReplyDelete
  10. The point is, does the production of phd outnumber its need? While estimating the need, apparently post-docs or equivalent temporary jobs should not be taken into account.

    ReplyDelete