Tuesday, April 13, 2010

Choosing a Graduate School

Besides being tax day, Thursday is the deadline to decide where to attend graduate school. How should you choose? I've blogged on this topic before but a few recent new wrinkles to talk about.

Typically Ph.D. programs in computer science offer funding (via fellowships, research assistants or teaching assistants) that cover your tuition and a small stipend. But a few programs at some financially-strapped universities are offering admission to the graduate program without such support. Should you join such a program if you can afford it? Maybe you'll get lucky and find an outside fellowship or an well-funded advisor but you have to worry about how much commitment the school will give you if they don't have a financial stake in your success. You really need to talk to the faculty involved and make sure you are comfortable with the situation.

As the field of theoretical computer science has gotten quite broad, theory groups at most universities cannot hope to adequately cover all research areas. Some departments have made the conscious decision to build strength in a particular area (like Northwestern in Algorithmic Game Theory). Joining such a group can be exciting if you are interested in that line of work but explore what other options would be available if you were to change your mind. 

Perhaps you are looking at the lousy job market for tenure-track faculty and thinking about not attending graduate school at all. Don't worry. As undergraduate enrollment is on an upswing, the economy recovers and the first wave of computer science faculty starts to retire the market should get much better by the time you get your doctorate. (And if I'm wrong this post may mysteriously disappear).

27 comments:

  1. I disagree with Lance's last piece of advice.

    If someone goes to grad school with the expectation that a tenure track faculty job will be relatively easy to get(especially in theory) they could be in for a very rude awakening.

    Better to go in with low expectations and have a backup plan and maybe things will work out for you.

    Actually my impression is that even in good time, only a minority (less than 50% certainly) of theory grad students end up in tenure track faculty positions. But I could be wrong about that.

    ReplyDelete
  2. (less than 50% certainly)
    You meant 5%, right????

    ReplyDelete
  3. As an undergrad student that tries to choose a graduate school for next year, I find the task to be surprisingly random.
    At least here, in Israel, the scholarships are pretty much the same in all of the decent universities.
    So other then browsing the web sites of the prominent researchers, searching for buzzwords that I find interesting - it's pretty much a coin flip.
    I'd actually love to hear any advice in the subject. I hate to make such an uneducated decision about something that could be career-altering.

    ReplyDelete
  4. "Joining such a group [a group strong in mostly a particular area, like Northwestern's AGT Group] can be exciting if you are interested in that line of work but explore what other options would be available if you were to change your mind."

    I'm actually on the fence right now with this problem, deciding between a school with broad strengths, and a school with basically a single strength. I loved the latter school, so I've almost decided on there, but what *are* the other options if I end up changing my mind? Is it not *too* hard to switch schools in that case?

    ReplyDelete
  5. I would advise any graduate students that faculty looking for grad students to exploit always give the line that finding jobs will become much easier once the next big wave of retirements hits. This line has been used since the early 90s.

    More realistic advice is that after five years of graduate school and four years of postdocs you will have a 10% chance of getting a tenure-track job. Don't trust any Panglossian faculty who lie to you about the job market.

    This is more important in the choice of an advisor than in the choice of a school. Do you want an advisor who is insouciant about your future and dishonest about your prospects? Look at where an advisor's past students have ended up. Track them down, since student lists on webpages can be incomplete and inaccurate.

    ReplyDelete
  6. I'd actually love to hear any advice in the subject.

    It's been said before but I'll say it again: go to the highest-ranked school you get accepted at. For academic positions, the ranking of the school where you did your PhD has a tremendous impact.

    ReplyDelete
  7. if I end up changing my mind? Is it not *too* hard to switch schools in that case?

    I think that Lance meant "within the same university".

    Different schools often have incomparable course or exam requirements so it is important to know that your work in one place will actually transfer to another. Transfer is usually not so hard if it is clear that you are doing well in research but students with graduate school experience usually face a higher bar for admission or transfer than new grad students.

    ReplyDelete
  8. Are there any compelling reasons to turn down an offer from a top-4 school for one outside of the top 10?

    ReplyDelete
  9. The number of North American students getting PhD's in physics has been roughly flat for the last 35 years ... yet most senior physics faculty in the 1970s and 1980s thought the academic job-crunch of those decades would be transitory.

    In recent years, a similar flattening has become apparent in biomedicine ... hmmm ... so is the academic job-crunch in CS going to be a permanent condition?

    For sure, it has happened before ... to academic disciplines in which pretty much no-one saw it coming.

    ReplyDelete
  10. "Are there any compelling reasons to turn down an offer from a top-4 school for one outside of the top 10?"

    Obviously, it depends on which school "outside the top 10" you are talking about. If it is Princeton, then yes.

    ReplyDelete
  11. Are there any compelling reasons to turn down an offer from a top-4 school for one outside of the top 10?

    90% of the time the answer is no. The other 10% of the time the answer is yes, because of two-body problems, visa problems, funding problems, etc.

    So, unless there are truly unique extenuating circumstances just go to the better place.

    ReplyDelete
  12. Small commuter plane: Berkeley
    Fog: University of California administration
    Mountain: CA budget crisis

    Losing altitude fast...

    ReplyDelete
  13. Are there any compelling reasons to turn down an offer from a top-4 school for one outside of the top 10?

    One more reason is your interest; if there is noone working on your topic of interest in the top-4 school, but some big names in the one outside top-10, you may want to go there.

    ReplyDelete
  14. What are examples of topics that no one at a top-4 school works on?

    Is it a good idea to study such things for your PhD?

    ReplyDelete
  15. Thanks to everyone who called Lance out on his job market "forecast". Like many of the other faculty I've spoken to in my years in grad school, Lance seems to be suffering from some serious survivor bias. I really wish professors would take their heads out of the sand instead of stonewalling grad students with bromides like "Oh, don't worry, I'm sure there'll be plenty of jobs by the time you're on the market".

    ReplyDelete
  16. What are examples of topics that no one at a top-4 school works on?

    Is it a good idea to study such things for your PhD?


    Yes and yes. Until recently MIT had no one in databases, Stanford had sparse coverage in theory, quantum wasn't (isn't?) particularly strong in any of the big four, etc.

    ReplyDelete
  17. "For sure, it has happened before ... to academic disciplines in which pretty much no-one saw it coming."

    Everyone sees it coming in CS. Postdocs are exploding. Lance claims not to notice, and gives out the old bromide that everything will be better once the next wave of retirements hits, any day now, but even he must see the difference.

    ReplyDelete
  18. so what's the conclusion/solution? is academia just a big ponzi scheme?

    ReplyDelete
  19. Lest the forest be lost in the anon flame war, theres a few seemingly obvious things I'd like to point out.

    1) You probably shouldn't do grad school unless the idea of spending several years where your primary job is to do good research sounds really exciting!

    2) The discussion re "name brandedness" of school really oversimplifies how folks should best choose a school! Yes, all things being equal, if folks have heard of the university, thats a big help.

    But that won't be what determines your success, its you, the now/future grad student! The quality of your work, your choice in advisor, the net community interest in your research (if you can't ask an interesting question, why should we care about your answer?), and of course your interpersonal skills etc all add together to make an tremendous impact!

    Also, first and foremost, the school you choose should be a place which has a number of faculty that you find interesting, and be an environment that is as conducive as is possible for you to be a productive grad student. But if you don't know that, you bigger issues I suppose.

    3) yes, its hard to get academic jobs, but if you're doing grad school for that reason, you didn't read point 1)! Moreover, in a domain like (theoretical) computer science, theres this curious fall back called industry which curiously can pay quite well for some of those skill you presumably have.

    Yes if you like focusing on research, the academic job would be better, but at the same time most normal jobs leave one with much more free time than the academia job, so if you really do enjoy doing research you have the free time to do so.

    Anyways,
    I'm surely oversimplifying and exaggerating what information I know / have been told by faculty over time, but the point is research is awesome, and personally I'm incredibly excited to be starting grad school next year.

    A soon to be grad student,
    -Carter Schonwald

    ReplyDelete
  20. in Theory students going in for a PhD should be ready for the possibility that there is no academic job in the end, and in fact, that they may end up with a job similar to what they would have gotten to without a PhD. In terms of career, it may delay it by a few years. So, why do it?
    -because research is exciting and the idea of spending a few years focusing on learning how to do research before getting a job is thrilling
    -because in the process you will learn skills that might turn out to shape your future career in unexpected and wonderful ways that I unfortunately can't explain because I don't really know what the world is like outside academia.

    I hope that Lance's last paragraph was tongue-in-cheek. Yes, like him I fully expect that the lousy job market for tenure-track faculty in Theory will get better. But I wouldn't "not worry" - it's going to change from "near impossible" to "extremely competitive".

    ReplyDelete
  21. Are there any compelling reasons to turn down an offer from a top-4 school for one outside of the top 10?

    Certainly! My own decision, made roughly ten years ago today, came down to a top-4 school and a top-25 school. I decided to go to the top-25 school, which has an excellent group in my field even though most people would rank the group at the top-4 school higher. I did this because the group seemed "hungrier" and a better fit with my research interests. The gamble paid off: my first job was a tenure-track position at a highly ranked school.

    To be sure, there are downsides to going to a lower ranked place. For example, the other groups in the department will probably not be as good. So unless you are totally zeroed in on an area and adviser it is risky. But if you think the lower ranked place is where you'll do your best work, it is definitely worth considering. Unless, of course, you have been admitted to my current school and a lower ranked school. Then you should blindly follow the rankings ;-)

    ReplyDelete
  22. if you can't ask an interesting question, why should we care about your answer?

    I wish you the very best, Carter, and I personally am quite glad I entered graduate school in theoretical computer science. However, the way I read these comments, people are pointing to the unfortunate reality that "mathematically interesting" is not the same as "interesting enough to get paid to do it long term." The reasons for the (growing) imbalance between the two notions of interesting are socioeconomic, and scientists of all sorts, not just TCS, are being caught in it.

    Don't allow your enthusiasm for a field to blind you to the fact that American intellectuals have far fewer privileges than they did a generation ago.

    ReplyDelete
  23. "Yes if you like focusing on research, the academic job would be better, but at the same time most normal jobs leave one with much more free time than the academia job, so if you really do enjoy doing research you have the free time to do so."

    How many people do you know who have done this? Just curious. It sounds like you have been fed some lines by well-meaning professors, but don't know what you are talking about first-hand.

    "1) You probably shouldn't do grad school unless the idea of spending several years where your primary job is to do good research sounds really exciting!"

    I think seven or eight years is more than several. Don't forget that after grad school come the postdocs. It is certainly exciting, but it is also painful.

    "2) The discussion re "name brandedness" of school really oversimplifies how folks should best choose a school! Yes, all things being equal, if folks have heard of the university, thats a big help."

    The truth is that if you want a job, you need all of the above, plus a good helping of luck. Every single factor is hugely important, including the name brand. From missing just one of them your chances are greatly diminished.

    ReplyDelete
  24. How many people do you know who have done (significant research outside of academia)?

    The list is longer than you might think, and includes some *very* prominent names --- Poincare (mining engineer/math), Einstein (patent clerk/physicist), Jane Goodall (conservationist/primatologist) Craig Venter (entrepreneur/genomics), David Shaw (financier/molecular dynamics) ... and even more broadly (and famously) the composer Charles Ives (insurance executive/composer) ... and very many more names might be given.

    It's true that later in life all of these people had access to multiple career options that grew from their earlier research success ... yet it is striking how *few* of them chose the option of full-time academia.

    ReplyDelete
  25. John, those mostly do not seem like good examples to me. Craig Venter did research for a living, not as a side job. David Shaw and Craig Venter both now do research because they are wealthy enough not to need another job. They also have minion researchers to work for them. Charles Ives was a composer. Poincare was in another time, and I don't think his experience is necessarily relevant today.

    ReplyDelete
  26. Anonymous, you make a valid point ... yet on the other hand, it's oxymoronic to assert "I want a non-traditional technical career exactly like the ones they had in the good-old-days!"

    The point here is that---for sure!---the coming generation of non-traditional careers surely will differ from those of previous generations.

    Last century, for example, Feynman classified research into ascent (to fundamental laws) and descent (to practical applications).

    Hmmm ... Feynman's classification seems pretty common-sensical ... and yet nowadays, as engineers seek to simulate quantum systems with fewer-and-fewer computational resources, we find ourselves embracing more-and-more of the tools of string theory.

    Let us imagine that we arrive at the "descendant" Nirvana of completely accurate simulation ... via proof technologies that derive largely from the mathematical tool-set of string theory ...

    Then, mightn't we take our simulation itself for the fundamental theory? ... to which the traditional linear structure of Hilbert-space QM is (in Ashtekar and Schilling's phrase) "only a technical convenience"?

    For me, it has been eye-opening to discover the powerful tools that molecular dynamicists use to simplify simulations by pulling-back onto state-spaces of larger dimensions ... one wonders whether the extravagant dimensionality of Hilbert space arises by this same mechanism?

    This leads to a non-traditional (and even zen-like!) view, that by descending to practical QM applications---as fast and as far as we can---we find ourselves paradoxically ascending to a more fundamental understanding of what QM is all about.

    Clearly, *ours* is a century in which *everyone* is extraordinarily well-situated to travel these paradoxical paths ... thanks to the miraculous and unprecedented confluence of yellow books, cheap computing, the arxiv, and the blog-o-sphere.

    For which, hoorah! :)

    ReplyDelete
  27. back to the original topic of the post - choosing a graduate school and advisor and research topic

    While you can do great research almost anywhere you can find a great advisor, there are some significant intangible benefits to going to larger, better ranked programs.

    1. The likely outcome is that you will end up in industry.

    While there are obvious caveats for working with professors with ties to a certain company, in general a hiring manager and especially an HR person will not know your advisor. They do know about which schools are "tops" and a diploma from a top ranked CS department will open doors that a diploma from a respectable CS department will not.

    2. If you are in a big department with a lot of grad students and postdocs in your area, you will have a much better idea about the job market, where you will stand in it, and what you will have to do than somebody who is all alone at a small school.

    If this does not depress you, it will motivate you or at least steer you in a more productive direction. Moreover, you can see some examples of success and emulate them. It would be frustrating indeed to spend several developing a "big fish in a small pond" attitude, only to be woken up as a postdoc.

    ReplyDelete