Phil Guo’s short online book, The PhD Grind, is the best description of the modern PhD experience in CS that I know of. People working on, or thinking about working on, a PhD in CS should read it. In this post I just want to comment on a few things.
Phil vividly describes the sinking feeling that he was just doing grunt work, not research. This feeling is extremely common, as is the actual fact of doing grunt work. In systemsy areas of CS there’s just no way to avoid building lots of stuff, fixing broken crap that other students (or, worse, professors) have built, running repetitive experiments that are hard to fully automate, etc. But here’s the thing: doing research isn’t like having sex where you might suddenly say to yourself: “Hey, I’m doing it!” Rather, most of the time, the research comes out only in hindsight. You look back over a year or three of hard work and say: What was interesting about what we did there? What did we do and learn that was new? If you can’t find anything then you’re screwed, but usually this only happens when the project was poorly conceived in the first place.
The “but I’m not doing research” problem comes up so often that I have a little canned speech about it that boils down to: “Work hard and trust me.” Really, this is the only way forward, because consciously trying to do research every day just makes no sense when you’re working on a large software system. Students who fail to heed this advice risk getting stuck in a kind of research paralysis where they stop making progress. I’m not saying that “Aha!” moments don’t exist. They do, and they’re great. But their importance is greatly overstated in narratives about research breakthroughs. For one thing, nine out of ten of these moments results in a useless or non-novel insight. For another, these moments are only possible because of all the preceding hard work. So who’s to say that the grunt work isn’t part of the research process too?
Another common experience that Phil illustrates nicely is the difficulty of coping with truly open problems. Our high schools and universities have not prepared us for this. I have a hunch that some of the brightest people from the best schools may be especially vulnerable to this problem because they’re so used to winning. Nothing has ever made them feel stupid before. Research makes you feel stupid—if it doesn’t, you’re not doing it right.
Although Phil shows far better than average judgement throughout his career as a PhD student, he seems to have made a subtle error regarding the role of publications. He clearly describes the Stanford culture: publish or perish, even as a student. It’s much the same elsewhere. But it’s easy to buy into this idea a little too much (this is very similar to the disastrous phenomenon where students become too invested in grades and test scores). For example:
At this point, I had given up all hope of getting a job as a professor, since I still had not published a single paper for my dissertation; competitive computer science faculty job candidates have all already published a few acclaimed first-author papers by this time in their Ph.D. careers.
This may be true, but even if so it’s largely a matter of luck. Most of the time, a student who publishes a solid top-tier paper early in their career does this because they were handed a meaty project in a hot area. Another way to say this is that even very talented students flounder when faced with the problem of selecting a topic to work on—so they rely (rightly) on their advisor. But here’s the thing: we, on the faculty hiring committee, are aware of all this. We know that there’s a fine, but crucial, line between the student with six top-tier papers who was fed every one of the ideas by his advisor, and the student who published only three or four papers but on her own ideas and agenda. Top-10 CS departments produce many of the first kind of student each year. Phil (who I know slightly in person) is one of the second kind. Even top-tier schools produce only a few of these people and they are exceptionally valuable. On a hiring committee, I would fight to interview this kind of person, though this doesn’t happen very often—these people are rare in the first place and then many of them do not continue on to academia. So, when Phil says this:
… I sensed that my current publication record wasn’t impressive enough to earn me a respectable tenure-track faculty job. My hunch was confirmed months later when I saw that, sadly, fellow students with slightly superior publication records still didn’t get any job offers.
He’s assuming that we just count top-tier papers instead of looking at the individual and their context. This happens sometimes, of course, but we try not to do it. The best hiring committees do not do this. Alas, as the bar for new hires continues to rise, it’s getting harder and harder to hire people like Phil. This is not good for our field.
Here Phil gets at a separate, and perhaps deeper, problem:
Of the 26 Stanford Computer Science Department Ph.D. graduates in my year, I consider myself fairly mediocre from an academic perspective since most of my papers were second-tier and not well-received by the establishment. My dissertation work awkwardly straddled several computer science subfields–Programming Languages, Human-Computer Interaction, and Operating Systems–so it wasn’t taken seriously by the top people in any one particular subfield.
The kinds of research topics I’m deeply passionate about aren’t very amenable to winning grant funding, because they aren’t well-accepted by today’s academic establishment.
There’s a large element of truth here. The research process (and particularly its funding aspect) is unfortunately very conservative. But this doesn’t necessarily mean that a person should accept defeat. Rather, take it as a challenge. The question is: How can I impose my vision on the paper committees and the funding committees? It’s hard, and it’ll take some time, but it can be done. People on these committees want to make the right decisions—you just have to give them the ammunition they need in order to actually do it instead of funding or accepting the next piece of boring, incremental work.
Anyway, I am not criticizing Phil’s decision to give academia a miss. It’s his decision and there’s every chance it’s the right one. But I do think he might have overestimated the importance of the top tier paper and underestimated people’s ability to appreciate the subtleties of a budding research career and to cope with novel, interdisciplinary work.
Finally, Phil has a nice answer to the question: “If you are not going to become a professor, then why even bother pursuing a Ph.D.?”
Here is an imperfect analogy: Why would anyone spend years training to excel in a sport such as the Ironman Triathlon—a grueling race consisting of a 2.4-mile swim, 112-mile bike ride, and a 26.2-mile run—when they aren’t going to become professional athletes? In short, this experience pushes people far beyond their physical limits and enables them to emerge stronger as a result. In some ways, doing a Ph.D. is the intellectual equivalent of intense athletic training.
This rings true.
7 responses to “The PhD Grind, and Why Research Isn’t Like Sex”
It is very tough, though. Mostly, successes are advertised and not failures. You compare yourself with your successful colleagues. You wonder whether you have taken the right decisions and most of the time there is no reassuring sign. Your solution, trusting your advisor is probably the best. Nevertheless, it is very hard to follow when you see that if you had opted to work in industry a few years ago, you could lead a much less stressful life. Moreover, having nothing tangible for a considerable amount of time while you spend your most productive years is not exactly encouraging.
Great post (and great memoir by Phil). It is very true that success in academia – especially when trying to get a faculty job – depends as much on sheer luck as technical skill. But it also depends on research taste. My research taste as a professor ran counter to most of the systems community, and as a result I never even bothered sending most of my work to traditional systems venues: It wouldn’t have been accepted, nor would it have been the right place to publish it. For a PhD student, though, bucking the trend and working on arcane topics can make it very difficult for your work to be known. There’s a reason there’s so much “me too” research in academia: Students flock to the hot topics in the hopes that their work will be published in the top venues. After a while the community tires of a given topic (recall MANETS? DHTs? And more recently, variants on MapReduce scheduling). I firmly believe that your fortunes as a PhD student depend on your being an early contributor to an up-and-coming area, so you’re one of the first to publish in that area in a major conference. It’s a bit like surfing: You need to know when to catch the wave.
Hi Matt, agreed on all counts.
Anon, yes it’s tough. But so is doing anything worthwhile. I doubt that being a PhD student or junior professor is as stressful or difficult, on average, as creating a startup.
The ironman analogy breaks down when you realize the financial disadvantage associated with doing a Ph.D “just because”
Suresh — good point, but by making oneself mentally stronger, it potentially improves one’s chances of succeeding professionally (and financially) in jobs outside of academia, at least for applied science/engineering fields.
Suresh — I think you underestimate the cost of excelling at a sporting event. Becoming a really good endurance athlete (or any other kind) most likely requires significant outlay of money and (more significantly) time, unless you’re a rare specimen.
Disregarding the economics of sports, the test for a good analogy isn’t whether it breaks down. All analogies break down, because no non-trivial analogy is perfect.
The test for a useful analogy is whether it helps us understand something that is otherwise difficult.