Collected advice on the mechanics of academia. Things I tell other people, should they ask, and things I tell myself, should I forget. Things specific to student-advisor are to be found here.

Caveats: this is all personal perspective, in case it’s not obvious enough by the fourth paragraph. I found it had a positive impact in how I go about my work, and I think it could help other people. It’s neither exhaustive or unchanging.


I try to read Hamming’s “You and Your Research” once a year. Pay attention to the “work more” bit, but ignore the part about stress and ulcers: it’s useless chest-thumping. Work on important problems, steadily; don’t fret if you see a contradiction: learn to keep conflicting views in your head; keep your doors open, literally; learn to sell.

Someone once told me that the only questions you should ask a researcher are “what is the most important problem in your area?”, followed by “what are you doing about it?”.

Academics is about building a better, shared understanding of something

Sometimes the best way to make an academic point in computer science is to build software; often you need to build something in other to understand it. But the questions are always of understanding, why and how: why does this matter? why is it bad? how can it be improved, and why?

The way I see it, academia exists to improve how people understand things. Each of these is actually important, and helps unpack how other important things are not academic.

If I understand something better than everyone else but I fail to communicate that to other people, then I’m not a good academic (I pass the “understand”, but not the “people”).

If I deeply care about doing something and sharing it with the world, but it doesn’t help understanding things better, than I’m not a good academic (“people”, but not “understand”).

So in an academic setting, you will be spending equal amounts of effort trying to understand things, and trying to find ways to make other people around you understand things the same way you do. Of course, other academics are doing precisely the same thing, so you will also spend time trying to see the world the way they do, if only so that you can explain to them why yours is better in a way they can see it.


At some point in time I used to recommend Strunk and White’s Element of Style, together with pretty much every other piece of writing advice you’ll find online. But it turns out it’s not such great advice. S&W was the first thing I read when I started caring about my own writing, and my writing did improve after reading it. On a third or fourth reading, though, I started noticing that S&W is contradictory and sanctimonious (“be clear”. Well, duh. But how?). So what should you read?

Start with Gopen and Swan’s Science of Scientific Writing: it’s short, sweet and it actually explains why you should write in any particular way. Then go on to buy Gopen’s Sense of Structure, and Williams’s Style: Toward Clarity and Grace.

If you’re interested in how language works, you will love (although you probably already do) the Language Log.

The Chicago Manual of Style is handy for grammar questions, but take it as a descriptive grammar reference, not as a manual of style.

The research paper

Research communication currently happens primarily through papers. This is where you show the world that

  1. you understand how the field got to where it is
  2. what you see to be a problem
  3. what you’re doing about it
  4. why it works

The context of a paper

The “context” of the paper is all the stuff, roughly, about where you connect with the expectations of the community reading the paper. Remember that you are writing your paper for other people, who have some set of ideas in their heads, likely somewhat distinct from yours. The introduction, discussion, and related work sections are where you get to make these connections. Pay attention to these sections!

The content of a paper

When it comes to the content of the paper, the platonic ideal is that every bit of it should either be

  1. a theorem, a proof, or an argument
  2. an experiment, a description thereof
  3. a statement followed by a citation

The basic rule is: anything you say in the content of the paper (that is, the actual things you have done, discovered, or invented) needs to be substantiated. Theoretical things require arguments and formal proofs, experimental things require experiment descriptions, and anything else requires citations.

This is, incidentally, a simple way to remove fat from a writeup.

Unlearning bad habits

Since you’ll do a whole lot of reading before you actually do any writing, chances are you’ll pick up writing ticks and idiosyncrasies, and you’ll tend to repeat them (I know I do, and I’m still unlearning). There are a lot of bad habits in math and computer science. Pay attention to what Shewchuk says.

Be careful with what you write, even on drafts.

Sometimes bad things slip past proofreading. And peer review. And they end up on twitter. Do not make this mistake. As so many wise mothers say, if you cannot say something nice, say nothing at all. This goes twice as strongly for the written word. One of the hardest things you will ever do as an academic is to learn to resist the temptation to be negative. There’s plenty of bad writing and bad research out there, don’t get me wrong. But in a few years you will go back and read some of your own papers and realize that they are not the masterpieces you once thought they were. Remember that when you find a bad paper.

What’s a non-native speaker to do?

The Corpus of Contemporary American English.


Do not think that you have to read the entire textbook, paper, or even section. Find the right algorithm, sentence, theorem, or proof you want to understand, and then make sure you understand everything about it. Fight the words that you’re reading: are they really true? How would you tell? Can you prove it? To yourself? To a friend? To a reviewer? What happens if you change the words one way or another? Is it obvious? Is it a new paper waiting to be written?

It’s far more important to know how to go find the right resource than to read everything from beginning to end.


You will over the course of your career review many things. Remember this very important rule: in very few occasions you will be reviewing people.

If you are reviewing a paper, it is never the authors that are wrong; rather, you only know that the paper might be wrong. Before submitting every review, make sure you check what you wrote. Very rarely you should need to write “author”. If you can’t fight the urge, at least save “author” for praising, and “paper” for criticizing.

This is a matter of making the process less of a personal judgment, but it is also a matter of making it more precise: you truly do not know if this was just a screwup. And writeup screwups that look bad do happen. I’ve made many of them myself, and you’ll make them too.

Hence, making your reviews about the paper and not about the authors is a good habit to get into: it makes your review much more likely to be listened to, and it focuses your thinking on those things that matter.

Here’s a neat trick I learned: the principle of charity also works as a great source of ideas. If the manuscript is wrong, you can tell the authors how to fix it. They will be thankful for it, you will have helped make the world a better place, and will have worked towards leveling up your “turn bad ideas into good ones” skill.

External resources