Rambles around computer science

Mildly profane meta-advice for beginning PhD students

A while back I chanced upon yet another “advice for PhD students” article, which cajoled me into finally writing my own. I should mention that I don't really like this sort of article; as a side-effect of this cognitive dissonance, the text below will be somewhat profane.

(Just as many such articles contain unacknowledged US-centrism, so mine contains some UK-centrism. I hope you can deal with it.)

Figure out how to make yourself work effectively. If you're relatively young when you start, a surprisingly large fraction of your PhD could go by before you get there. For me, eating properly and sleeping properly are the hardest things to get right, and (of course) a lot flows from them. It might sound silly, but I was in my third year before I mostly cracked this. (Ten years later, I still struggle from time to time.)

Read. Knowing the prior work means really knowing it; not just the last ten years' worth or the high-profile stuff. To merit the PhD degree, I'd argue it's actually more important to be an expert in your field than it is to advance human knowledge. (Some institutions' regulations even reflect this. Not that that's relevant... in practice, your examiners will be guided far more by the folklore of their particular research culture than by any regulations. There's more about examination below.)

Learn how to give good talks. The easiest way is to give talks, and get frank feedback. Also, attend talks and pay attention to which ones work. (Flip side: even after learning this skill, not every talk you give will be good. Don't beat yourself up.)

Attend talks. Cultivate your breadth as well as your depth. This can be easier or more difficult, depending on the particular academic environment you're in. I was lucky on this one.

Don't let anyone tell you “how research must be done” or “what good research looks like”. There are a huge number of different styles of research, and most people know only a few. The more prescriptive someone's take, the more narrow-minded that person and the more loudly your bullshit detector should be sounding. Example: once I was advised that good work should generate “graphs that go up and to the right”. If we only had the kind of research that does generate such graphs, the world would be a poorer place, no matter how good that work. To this day, there are few graphs in my work (and those that do exist don't look like that).

Find your research community (or communities). This relates to the previous point: even people claiming to be doing similar research and/or have similar goals might actually have very different attitudes and methods. Communities have their own vibes and personalities even beyond those issues. It took me a few years, and a couple of wrong attempts, to find the community for me. I was glad when I did. No community is perfect, of course, and I still feel like an alien from time to time.

Don't let the bastards grind you down. Chances are, if your work is at all interesting, some people will want to shit on it. More generally, the research world attracts a lot of extreme personality types (including narcissists, bullies, hypercompetitive arseholes, manipulative gaslighters, etc.). Learn how to see them for what they are—even the ones who are well-regarded famous people.

Get access to good feedback. It doesn't have to be your supervisor (who, realistically, you might well have chosen based on limited information and/or without really knowing what you were doing). It doesn't have to be one person; it's probably better if it isn't. Despite good intentions all round, I found that my supervisor's feedback wasn't very useful (I have no hard feelings). But I found that through my surrounding research group (groups, in fact) I could crowdsource a lot of extremely valuable feedback.

Know how to disregard unhelpful feedback. Some people won't have anything constructive or insightful to say, but will still be happy to pick imaginary holes, make you feel small and/or make themselves feel clever. Don't let them sap your energy.

Learn how to defend against probing and/or hostile questions. In a talk situation, or perhaps a viva/defence situation, you will get these from time to time. It's rarely a fair fight; often, the hardest questions to answer are the most sloppily formulated or the ones founded on wrong assumptions that remain implicit. There are some tactics for handling these. The simplest is batting them back for clarification or re-statement. Another is offering back the question that you think was intended (and, hopefully, one that actually would have been a sane question). The more expert technique is on-the-hoof picking-apart of exactly what those wrong assumptions were. Some of my most awkward moments as a young researcher were when my lack of skill at this was exposed. So, the other point is not to sweat it when this happens.

Don't be afraid to talk to people whose work you admire—even by sending cold e-mail. Even “famous” people are mostly approachable and are happy to hear if you like their work and have something interesting to say. (And if they're not, then it's their problem not yours.) I was far too timid about this. You never know; it might lead to something. And even if not, there is value in just saying hi. I held back too much.

Socialise your work, and yourself. Get yourself invited to give talks at other universities or research labs. Talk at workshops, give demos or posters; take the opportunities. Although it's possible to overdo this—getting some time for actual work is important too!—it's easy to underdo it, especially if you're shy or not a natural self-publicist. (Sad but true: the more people know you personally, the more citations your work will attract, especially earlier on.)

When communicating with others, learn how to focus your thoughts, and tailor them to the audience. This is partly about having not just one but many “elevator pitches” (or, as we Brits say, “lift pitches”). But it's more than that. One of the things I used to get wrong when talking to people, especially by e-mail, was to let my message degenerate into a brain dump. There's always more you can say... but often, less is more. Select the thoughts or points that you're most interested/excited about, and filtered by relevance to or expected resonance in the recipient. Have faith that the other stuff will come out over time if it's fruitful. There will be a lot of implicit cues, in how you're writing or talking, hinting to the recipient that there is more where that came from.

Learn when administrative rules are not really rules. Even the best institutions will have over-narrow admin-defined ideas about how research students' journey should go. They might assign you to one research group even though your interests span several. They might refuse to fund your travel to a particular event that's considered out-of-scope. Towards the end of your PhD, they might neglect to provide extension funding, or not connect you with the help you need in applying for your next funding. In all cases, the rules often appear inflexible when you read them, but they are often fluid in practice, so be bold enough to ask the question anyway. (One example: I was told I had only a three-year studentship that could not be extended. But I asked for extension money anyway. The rules didn't allow extending me, but they did allow retroactively increasing my stipend, paying me effectively as additional back-pay; this got me an extra 4.5 months of funding.) Usually it helps to be on good terms with the relevant administrators—go and talk to them in person, be unassuming, try to understand where they're coming from and what the real constraints are. It can also help to have a supporting professorial type (again, need not be your supervisor) who will be your advocate and help you poke the administrative gears.

Know how your PhD will be examined, and do what you can to influence and optimise for that. Under the UK system, in practice it is your supervisor who will choose your examiners, but you may have some influence over that choice. Sadly, “examination-friendly research” and “good research” are overlapping but non-equal sets. This is particularly dangerous in the UK system where the examiners are all-powerful, and doubly so if your supervisor's social capital is low enough that the examiners do not see fit to moderate themselves on that account. It's pragmatic to see your thesis as a thing to be examined, rather than a thing you're writing for yourself (even though the latter may make you feel fuzzier and more motivated). This also comes back to the “many styles of research” point: if an examiner isn't used to work of your particular style, you will get a rougher ride than you deserve. Make sure the choice of your examiners is made with due consideration of this; it shouldn't simply be whoever your supervisor randomly bumped into and asked on a whim.

Be sceptical about advice. There's a lot of it about, given freely and often smugly by people who took a single path. This is generally well-meaning, but the people concerned don't necessarily know much about the other paths. Successful people often don't truly know why they were successful. I received huge amounts of bad advice during my PhD, which has made me a cynic about advice-giving and mentorship to this day. (Can you tell?)

Clearly I'm a hypocrite too, otherwise I wouldn't have written this. My so-called career path, though not without its good points, has definitely not been one to emulate. Sadly, I can't blame all of its failings on following other people's bad advice. Still, I hope this article has been different enough that it makes a useful counterpoint to advice in the usual vein.