Getting started with a Ph.D.

Tags: musings, research

Published on
« Previous post: A Method(ology) to the Madness — Next post: Delayed gratification and gratitude »

If you are reading this and are about to take a deep dive into your Ph.D., let me welcome you to your new chapter and your new adventure. Many guides talk about ‘surviving your Ph.D.’, but I want you to thrive in your Ph.D. This article covers a few of the basics that will help you get started. I have written this for a typical STEM Ph.D., because I lack familiarity with other subjects. In general, I would hope that all concepts—but maybe not the examples that I chose—translate quite well to other disciplines. Let’s get started!

Prepare and plan for the long haul

I cannot stress it enough: ‘A Ph.D. is a marathon, not a sprint’. This means that you need to shift to long-term planning, something we humans are notoriously bad at1. If it helps you, you can visualise the time it might take you to finish the Ph.D. and start treating that time as a budget that you can allocate to things. You will see that there is a slot for almost anything you want to do—so there is no rush. Your objective is to finish your Ph.D. with expert knowledge about your field and some quality publications to boot2. You do not get quality if you rush everything. Hence, missing a conference deadline, for example, is much less serious than it might sound and feel at the beginning. Let no one tell you any different.

A quick digression here about working towards a deadline: some poor souls think that I am advocating for complacency here when I am saying that missing a deadline is not important (cue the collective gasps of the audience), but if you are planning for the long haul, it is kind of irrelevant whether your glorious publications are published in 2020, 2021, or right before you start finishing your thesis. A good publication is worth a lot more than low-quality ones. One of my favourite quotes from Epictetus summarises this brilliantly:

How long are you going to wait before you demand the best for yourself and in no instance bypass the discriminations of reason?

Demand the best for yourself by taking time to develop into an awesome expert:

  • Plan out certain phases of your Ph.D. in order to figure out what you want to do or achieve in each one of them

  • Do not rush blindly into deadlines. Most, if not all, deadlines are arbitrary.

  • Never compromise the quality of your work.

Moreover, if you are in it for the long haul, take care of yourself. This involves paying particular attention to your sleep and to how much you work. I once had a colleague who used to appear and disappear at highly irregular hours: on certain days, they would work from, say, 10am to 5pm. On others, they would work from 10am to 10pm, while on a few days, they would also just work from 2am to 5pm. These irregular patterns are bad for different reasons, but mostly because they make it easy to fool yourself into thinking that you are working more than you actually do, at the expense of your mental and physical health. Instead, do the following:

  • Treat your Ph.D. like a job: in most jobs, you will be fired if your hours are too irregular. Use the flexibility of academia but do not abuse it (because you will hurt yourself the most). Try to work more or less fixed hours every day and make sure to tell yourself that this should be enough.

  • Never compromise sleep for work. The idea of staying up all night and finally finishing that paper quickly loses its appeal when you have done it multiple times3. The paper quality also does not get better, trust me on that. If you absolutely must sacrifice something, be sure to take some time off to account for it.

  • Do not get participate in fights about who is working more hours every week. In particular, do not believe people when they talk about working long hours. They are probably wrong.

Let me reiterate the last point because it seems to be a badge of honour to work much more than anyone else in the room. I am no stranger to hard work, but if you work longer hours, remind yourself that you are also less efficient as time drags on. If by 6pm, only half of your working day has gone by, you might not be inclined to work efficiently and quickly on your projects because you can fool yourself into believing that there is always more time. This is bad—strive for efficiency and getting stuff done instead of counting hours4.

Read and establish a bibliography

Whatever you are doing your Ph.D. in, aim to get some quality reading time. Reading is a another secret weapon in your arsenal of skills. The more you read, the more you will understand about your topic, and you will be able to form new connections, which in turn will lead to good ideas. Here are some suggestions for reading:

  • Well-written expositionary papers that influenced your field. It can take some time to find them, especially if you are just starting out, so let me suggest at least one: A Mathematical Theory of Communication by C.E. Shannon is one of the classics that every computer scientist should have skimmed at least once. Shannon takes so much time to lay out the ideas behind his theories, that it is a pure joy to read. I wish more papers were written like this.

  • Published papers that are somewhat related to your field or topic. Reading these helps you understand the open problems and make it possible for you to ask questions that no one has asked before.

  • Text books on subjects tangential to your Ph.D. topic. It always helps to broaden one’s own knowledge and find out something new about mathematical tools, for example.

  • Articles about tools that you might need for your research. If you are developing software, read something about professional software development, such as The Pragmatic Programmer. You do not have to strive to become a professional software developer to benefit from these books. As a scientist, it is your responsibility to learn how to use the tools that you have to use daily. By the way, this does not necessarily have to be a software tool—even something like ‘statistical analysis’ can be a tool. Stock your toolbox well and learn how to make use of it.

In particular when reading papers, understanding all the material might seem daunting. Hence, let us briefly discuss how to read after having decided what to read. In general, I am an advocate of ‘reading for the gist’:

  • Make multiple passes of the paper, going from very coarse (just the abstract, just the figures, etc.) to the very detailed (what does this symbol mean in that proof?)

  • Try to figure out what the main ideas of the paper are. Most, if not all, papers have a central idea or hypothesis, and some ‘minor’ or ‘supporting’ ideas. Learn how to spot those.

  • You do not have to understand every nook and cranny of a paper to draw some use from it! Often, it is sufficient to get a general idea of something and then move on.

Think of this iterated reading process, which is masterfully outlined in How to Read a Paper, as sharpening a knife and cutting progressively deeper until you reach the ’truth’ of the paper. It is normal that you will take a long time initially, but as with all skills, you will get faster and better if you keep at it! At some point, you will be so fast that you can ‘sort’ through stacks of papers from your field and quickly decide whether you want to read them or not. This brings me to my next point: when you read, try to read critically, but with kindness. No paper is perfect—yours included—and every one of them will have some flaws. Try to find those, as well as the gaps in the arguments. In some cases, this might teach you something new, or give you a nice insight that might lead to something. However, do not forget to be kind when doing so—your goal is not to rip the work of others to shreds! Instead, think of the words of Francis Bacon:

Read not to contradict and confute; nor to believe and take for granted; nor to find talk and discourse; but to weigh and consider.

Having now selected what to read and how to read, we need to talk about what to do with your notes. I might dedicate a separate article to the specifics of this, but my recommendation is to establish a bibliography by storing your notes about your reading materials in a central location. Mac OS X users have their work cut of for them when using something like Notational Velocity; if you are more of purist, a single Markdown file will do as well. The important thing is to distil your notes into something that you can revisit—for example when summarising a paper later on5. Your papers will help form your bibliography, i.e. the set of materials that you will use when writing your dissertation and conducting your research. Since every academic field is different, it is hard to give good numbers here, but I would estimate that every computer scientist Ph.D. student should be familiar with something like 100 papers at the end of their Ph.D. A few of these papers will be well-read, while some of them will have only been skimmed. Given that the average Ph.D. takes you about 4.5 years, there is ample time to reach this number—see also the first point about planning for the long haul.

One caveat before moving on to the last point of this essay: reading is so good and useful that I would recommend every Ph.D. student to take at least half a day every week to go through some materials. However, if you start feeling pressured by your reading list, or you have the feeling that there are no new ideas any more that you could be working on, it is time to stop reading or at least consider that you might be reading the wrong things. Reading should always leave you a little bit ‘fuller’ than before; if you feel ’emotionally drained’, something is wrong and it might be time to switch to something different, such as implementing something. Lest this caveat makes you afraid, let me alleviate these fears: of the many Ph.D. students that I have encountered so far, only a single one experienced this feeling of being stuck, and it was related to external circumstances of their Ph.D.6 rather than their reading habits (moreover, I helped them get unstuck by trying to reproduce another paper; this is always a fun activity). Hence: tolle et lege!

Get to know yourself

The last point I want to make in this essay concerns your personal introspection skills. It is the point that is tough to make and tough to understand because everyone is different. Let me try: doing a Ph.D. requires learning some organisational skills, which in turn require you to get to know yourself better. Find out what motivates you. Find out what makes you afraid. Find out why you procrastinate. Find out why you get into the zone. Treat yourself as a research project and collect a lot of data about your personal preferences. Then, having collected these data, structure your work environment around them. Here are some examples, taken from my own working style7:

  • I always have a battery of projects to work on, and I keep track of things that require a lot of ‘creative juice’ (writing a paper, writing a grant), and things that are more ‘playful’ (learning something new about visualising data, trying out a new framework for Python programming, etc.). Whenever I start to ‘drift’ from one project, I ‘procrastinate’ productively by switching to something else that also needs to be done at some point. This also gives me additional energy to ‘recharge’.

  • I tend to prefer writing text and reading papers in the morning, while the afternoons are more for writing code or re-writing text. If your attention span is similarly preferential, figure out your best times for doing certain things. In my case, my preferences mean that I prefer meetings to be in the afternoon because I am itching to get some writing and reading done in the morning.

  • Whenever I am completely stuck, I just talk to my colleagues8 and try to learn something about their current work. This either gives me new impulses for my own work, or helps them get unstuck. Everyone wins!

Knowing yourself also means being honest with yourself—if you slept badly, you might not be in top form, so you should not expect to be able to, say, read as many papers as on a good day. Developing a model of yourself is crucial to have a fulfilling time during your Ph.D. Find out what motivates you, what projects you are drawn towards, and so on. In particular, treat the beginning of your Ph.D. as a magical ’land out of time’ and do small steps, taking your time to learn something new and exciting.

Enjoy this time and see it as a stepping stone toward discovering something new about yourself and our marvellous universe. All the best to you and your research, until next time!

  1. But at least everyone is bad at it initially, so just knowing about this issue will make it possible for you to avoid it. ↩︎

  2. The number of publications depends on your field. In pure mathematics, it is typically 0–1, while in computer science, it is more like 3–5. Let’s not get bogged down why it is a crazy and stupid idea to assess the act of building knowledge by counting publications, but just accept it for now. ↩︎

  3. To the people who know me personally, I can only say ‘Do as I say, not as I do’. Looking back, the papers that were least stressful were also the papers that were easily accepted by a conference. ↩︎

  4. If you have to code a lot, think of it like this: you should not equate writing a lot of lines of code with writing good code. Remember the adage of one of the grand masters of Unix, who apparently said ‘One of my most productive days was throwing away 1,000 lines of code’. ↩︎

  5. Again, the specifics of how to summarise and which systems to use will be part of a future blog post. ↩︎

  6. The usual dose of a professor who was unwilling or unable to invest more time in their students, coupled with a few circumstances beyond the control of the Ph.D. student in question. ↩︎

  7. I do not consider this style superior to anyone else’s style, I am merely using myself as an example because I know my style best and can assess the rationale behind some decisions. ↩︎

  8. Much to their chagrin, I suspect. ↩︎