One of the main benefits of a PhD, or an academic career, is the freedom you have to follow your interests. As a PhD student you have a remarkable amount of freedom. Very few people ever have so much time and freedom to follow their interests, and you probably never will again during your working life. (Your retirement is another matter and some academics continue to do research after their official retirement.)
On the other hand a PhD is pretty much guaranteed to be deeply frustrating at times. This is little wonder, since a PhD is an intrinsically big and difficult undertaking. There will be setbacks and aspects of the work you just don't enjoy. In fact it can be difficult just to get used to the nature of the work and having so much freedom. Many new PhD students also get a bit of shock when they find they are no longer the most clever person in the room, and that results are not as easy as they were on taught courses.
Perhaps worst of all there can be a lot of uncertainty about all sorts of things: how to proceed, whether there's a flaw in your work, whether you know enough about an area to make decisions, whether your results are significant enough to merit a PhD, whether it has all been done before, and so on. The trick is to deal with such uncertainty rationally and then avoid unhelpful worrying. In other words, split your problem into two sub-problems: the underlying problem and your counterproductive emotional response to it. Then deal with each in turn. If you're worried that your new algorithm may not be new after all, decide how much time to spend on a literature search, then accept that you've dealt with it in the way that seemed best at the time, and then move on. It may later turn out that you were wrong, but that is the nature of research.
In addition to uncertainty, I suspect the other big source of misery for PhD students is misguided objectives, which I will deal with later. If you have the right objectives, and if you can focus on what you enjoy about your work and leave the worrying aside, you'll be much happier. It can help to remember why you wanted to do a PhD in the first place.
Incidentally, a bad reason to do a PhD is to prove how clever you are. That motivation doesn't make the work intrinsically rewarding and spending years proving such a point doesn't seem very healthy.
Clearly it's useful to be able to compare students and schools. Students are also generally motivated to achieve good marks and feel satisfied when they do. Nonetheless there's a serious downside to the increased testing which many writers have noted in recent years. Students feel stressed by constant testing. Teachers are constrained by highly regulated course specifications, the need to test frequently, and the pressure to "teach to the test" in order to maximise results.
In these circumstances it's easy to confuse getting high marks with successful learning. If your objective is to pass a course so you can get a qualification and a well-paid job in an unrelated area, then high marks are a reasonable measure of success. But if you define successful learning as gaining understanding of the subject then high marks and success are not the same, the reason being that tests are often poor measures of understanding. My computer, for example, can perform many mathematical calculations better than I can, but it understands nothing about maths. Similarly, you can teach people to carry out calculations with little or no understanding of the underlying maths. (In fact many scientists will use statistical tests without much thought about how well their data fits the assumptions of the test, or about what the results really mean.) Because it's easier to teach and measure procedures than to teach and measure understanding, students end up with a counterfeit education.
As a PhD student you suddenly escape constant explicit testing; you won't receive marks on your work and in the end you will either pass or fail your PhD -- there's no such thing as a 2.1 at PhD level. Without marks, what should you try to optimise? That is, what should your objectives be? An obvious answer is to publish papers and produce an impressive thesis. You will definitely get positive feedback for doing so, but I think making this your criterion for a successful PhD is just as misguided as making high marks your objective in earlier studies. Such misleading objectives are everywhere. You might be tempted into a higher-paying job only to realise the long hours, the commute, or the work itself make you unhappy. So be careful what you wish for! In the next section I argue against over-emphasising the thesis and in the section after that I list the objectives I think a PhD student should have.
Furthermore, few people continue to work on exactly the same thing after graduating -- it's rather unlikely you'll find a postdoctoral position on a project which is effectively an extension of your PhD. Many people have a substantial change of direction in their research after graduating, or even a major change of direction in their career. As a result much of the most specialised technical knowledge you gained during your PhD is not likely to be of much use, but luckily the transferable skills you've learned can serve you well no matter what you end up doing.
Students sometimes start a PhD thinking that they'll solve a substantial problem and have the last word on it. This is not likely, to say the least. It's much more realistic to expect your work to add modestly to the body of knowledge on a subject, building on what was done before and being built on by later work. This is all that's required or expected. So the scope of your contribution is expected to be modest, and so too is its impact. Have you ever thought about how many people are likely to ever read your thesis? In most cases it will be less than ten. How relevant do you think it will be in ten or twenty years? In most cases it will only be of historical interest, if that (although foundational and theoretical work tends to have more staying power). So, given the modest scope and impact of most PhDs, what is their value? It's that by the time you complete your PhD you should have become a competent, independent, mature researcher. Having a PhD means that you can be expected to carry out good quality research on your own and that you have all the associated skills. The thesis is just the tangible evidence that you've acquired these abilities. Again, your PhD is just part of your journey. This is partly why you must pass a viva, and do so without input from your advisor.
Suppose you are able to write well but don't have the other qualities described above. If your advisor gives you detailed instructions throughout your PhD you could generate a good thesis in the end, but it would be your advisor, not you, who deserves a PhD and who has shown him or herself to be a mature researcher. (Incidentally, not many advisors are willing to put this much thought into a student's thesis.) During the viva the examiners will want to evaluate you as a thinker. They will want to satisfy themselves that you are a competent, independent, mature researcher. If your advisor has supplied all the ideas and insights, you can expect difficulties during the viva. Conversely, if you have fallen short of your objectives, or your results are disappointing, you should still be awarded a PhD if you can convince the examiners that you are a mature researcher.
Once again, a PhD is just part of the journey. If you apply for an RA position, everyone else applying for it will also have a PhD. Most of them will have a couple of good references. What will set you apart?
Learning isn't so much a process of adding information as critically integrating it with your current understanding. If you were to add information uncritically you would end up with lots of contradictory statements with few links between them and little understanding.
In contrast, critically integrating information involves iteratively revisiting assumptions and revising your understanding in the light of new information. This leads to changes in opinions, modification of old assumptions, and creation of new connections between ideas. This is how deeper understanding is achieved, and where the insights and progress needed for PhD study arise. Writing, presenting and teaching about a subject are surprisingly good ways to stimulate this revision and integration.
Much more could be said on the subject but suffice it to say critical thinking is the foundation of the scientific method and that the opposite -- rigid, dogmatic, and fundamentalist thinking -- will not get you very far.
I often find that when someone can't explain something clearly it's because they don't understand it very well themselves. Oddly enough they often understand it less well than they think they do, and it's only when they try to explain it that the confusion becomes obvious. This is certainly true of me.
For these reasons I doubt it's possible to be a good researcher without the ability to communicate clearly and coherently. To invert things, you can develop your ability to think critically by developing your communication skills, and critique your reasoning skills by critiquing your communication skills.
Furthermore, even if you - somehow - come up with great ideas, if you can't explain them clearly and convincingly, and in a way that suits the audience at hand, you'll have trouble getting them funded, or published, or taken up by others. Although some of your conclusions may be proofs, which should stand on their own, you will often have to persuade others to accept your views using less rigorous arguments. This is more subtle than writing proofs and often more useful. Scientists are neither purely objective nor unemotional. You should be able to interest people in your work, entertain them with your presentations and even excite them. Presentation skills are also very important when applying for jobs. Your CV may get you shortlisted but your presentation and interview will often be the deciding factor. It's common for the ranking of candidates to change completely after interviews. Your research group should provide a supportive environment in which to practise oral presentations, as departmental seminars do. Both are less formal than presenting at a conference and you are less likely to face aggressive questioning. Lecturing is again somewhat different and also good experience.
If there's one fundamental point on writing it's the following. It's essential that your writing consists of putting your own thoughts into your own words. You must not write by paraphrasing other sources. Occasional and brief paraphrasing is appropriate, e.g. when giving an example or a definition, or when presenting a summary of another person's position. However, writing an entire report by paraphrasing an existing one is not appropriate. This generally produces less clear and less relevant results than your own writing, and in any case you need to develop your own writing through practice. You must not write by editing text from other sources into something new. This is bad for the same reasons as paraphrasing but is also plagiarism, which is a serious form of academic misconduct.
There are simple techniques which can help you write, for example putting away your sources when you start so you're not tempted to follow them too closely, making slides before writing text in order to get the structure right, or recording yourself explaining things off the top of your head and typing them up later. Finally, although I consider them much less important than clarity, don't overlook the technicalities. Badly formatted references, sloppy layout, and bad spelling, grammar and punctuation do not leave a good impression, although the last three are more easily overlooked when the author's first language is not English.
I've used the word "advisor" here rather than "supervisor" to emphasise the independence of a PhD student.
Initiative is also good for the quality of teaching in your department. I think that if you're given the responsibility to set an assignment you're more likely to think of ways to improve it than if you're only given the task of marking it. This is one reason I'm in favour of PhD students taking ownership of coursework and lectures.
Similarly, you should take ownership of your PhD. Even if you are funded by a scholarship, you are working for yourself more than you are working for your advisor or whoever is funding you. You have secured a position that will fund you for a certain amount of time and you should use this time in the way that benefits you most. Suppose you become very interested in a particular specialised area and decide that you want a career in it. If there are five active researchers in this area in the UK you should not only get to know their work but the researchers themselves, and you should make sure they know you. Depending on the circumstances you may interact with them by email, meet them at a conference, invite them to give a talk in your department, offer to give a talk at their University, organise a workshop, or a research visit, or collaborate on a paper. The are a lot of possibilities but odds are none of those things will happen unless you initiate them.
A word of warning about filling in the form, and for that matter about any summary of your work, such as a CV. As mentioned earlier it's easier to pursue misleading objectives, and one example is to make adding to your CV an objective in itself. This can result in "CV pumping" which, in the worst case, involves things like publishing low-quality work just so you can add it to your CV. Don't sleep through seminars just so you can list them on the form.
Each item in the following list should contribute to one or more of the objectives listed earlier. As a PhD student you should normally aim to:
Of course it's hard to finish a PhD in 3 years as it is and taught courses take up time, but I don't think time itself is the issue; it's how the time is used. My PhD took almost 5 years (including a year's worth of lecturing), but if I had had the experience I have now, I could probably have done it in 2 years.
In general there are diminishing returns to be had from any one strand of research. In fact the more time you spend on one thing the more likely you'll be unable to see the wood for the trees. My best guess at making PhD time more productive is to get involved in a series of short collaborative research projects, and I'll say a little more on this later, but, in short, work smarter, not harder or longer on one thing.
If you don't already have a degree in computer science you should make sure you're familiar with the fundamentals of the subject. Many people write terrible code which gets the job done, but just barely. This may allow you to complete your thesis, but unless you're doing very little coding you're better off devoting some time to learning how to do it properly - it will pay off in the long run. Similarly, you ignore core subjects like complexity theory at your peril. Again, you should be a well-rounded computer scientist by the time you graduate.
If you already have a degree in computer science there may not be many relevant courses left for you to take but consider whether further study of mathematics or statistics would be useful. You may also consider a philosophy of science course.
If there are useful courses consider how much time to spend on them. It would not be unreasonable to take 10 credits per semester for the first two years for a total of 40 credits. MSc students take roughly 60 credits per semester so this should take only 1/6 of a PhD student's time (or a little less than one day a week), and only during teaching time (20 weeks/year), and only for two years. This is not an excessive load, and it doesn't impose on the third year when it's understandably difficult to concentrate on anything other than finishing the PhD. For comparison 40 credits is 40% of the taught component of an MSc.
You should register on courses so you will receive emails and be registered on the forum. For courses which are most relevant to your studies you should do all the coursework and the exams. This will ensure you get the most out of them and will give you a measure of how you're doing on the subject. In rare cases taught marks can also help assess whether a student should continue with PhD studies. In other cases doing all the coursework and exams is not appropriate -- you may only want a passing familiarity with the subject and the coursework and exam add considerably to the burden of taking a course.
Students in the Machine Learning and Biological Computation group. You should do the new parts of the advanced topics in machine learning course each year: new guest lectures and the new data mining challenge. You can also use the "student conference" as an opportunity to practise presenting.
I'm sure this document can be improved. If you have any feedback I'd like to hear it.