Advice for Students

A fun starting point

E. Robert Schulman and C. Virginia Cox, How to Write a Ph.D. Dissertation, Annals of Improbable Research, Vol. 3, No. 5, pg. 8.

Before you begin

So, you are a graduate students and you don’t mind biased information meant to help you out? Read on! If you are an undergraduate student or want to be, go see what Paul Graham has to say.

What is academic research?

Academic research is made of two essential ingredients. Communication and implementing new cool ideas. Both of these ingredients are essential. Great writers or speakers do not necessarily make good researchers. Also, people who can have and put in practice great ideas do not always make great researchers. You need to enjoy communicating your ideas. You need to enjoy looking for news ideas.

So, you are a great programmer, but you hate to write in English? Forget about academic research.

What is a Ph.D. Thesis?

A good thesis topic will simultaneously express a personal vision and participate in a conversation with the literature. Your topic must be one you are passionate about. Nothing less will keep you going. Your personal vision is your reason for being a scientist, an image or principle or idea or goal you care deeply about. (…) At the same time, science is a conversation. An awful lot of good people have done their best and they’re written about it. They’ve accomplished a great deal and they’ve completely screwed up. They’ve had deep insights and they’ve been unbelievably blind. They’ve been heros and cowards. And all of this at the same time. Your work will be manageable and comprehensible if it is framed as a conversation with these others. It has to speak to their problems and their questions, even if it’s to explain what’s wrong with them. A thesis topic that doesn’t participate in a conversation with the literature will be too big or too vague, or nobody will be able to understand it. (Source: MIT AI research howto guide.)

Are you sure you want to earn a Ph.D.?

Firstly, a little test. How many research papers have read in the last 12 months? Magazines, even serious ones, do not count. How many boring research papers have you read? You know, the ones with a formal introduction and the painful reference section? If you have read none, a Ph.D. is probably not for you. I would say that having read 5 research papers is an absolute minimum before you can even wonder whether you want to spend your life writing these.

Here are some bad reasons to start a Ph.D.:

• I’m building a business as a consultant and a Ph.D. sounds cool. Do not even think about a Ph.D. for this reason. Consultants with a Ph.D. do not, in my experience, earn more money.
• I already have a job and my current project would make a great Ph.D. thesis. You think so? Very few industrial projects make a good Ph.D. thesis. In fact, it is extremely hard to reconcile industrial interests with Ph.D. work.
• I would earn more money if I had a Ph.D.: this is false. Earning a Ph.D. means you are using up a lot of your time to make an entry into the academic world. Financially, it might not be a complete disaster, but it is hardly wise.
• I want to impress my friends. A Ph.D. might impress your friends, but so would earning 1 million dollars by the time you are thirty, or becoming a top lawyer or saving lives in a hospital. Consider that many people who go for the Ph.D. have to go live in cities they wouldn’t have otherwise chosen. There are many ways to impress your friends, but a Ph.D. is probably the worst option.
• I’m so smart, it is my destiny. Actually, really smart people choose easier paths. If you are down to the Ph.D. path, well, something is a bit wrong with you to begin with. But that’s ok, there are many of us.

So, are there good reasons to get a Ph.D.? There are some, but I will not list them. If you can’t find them, then a Ph.D. is not for you.

You can try filling out Mihir Bellare’s form about the Ph.D. experience to see if you should pursue this further.

You are not yet a graduate student but want to be?

Download recent research papers. Study them. Try to imagine what work you could do. Write it down. Sketch a research plan. Yes, do this on your own. If you can find a way to start a paper on your own, do it. No, you do not have to wait to have a supervisor to do research. And if you started some research and got some early results, you are much more likely to convince the good supervisors to take you in. This does not mean you should be not seek a supervisor early on, but prove as early as possible that you have what it takes to do research. And please, read research papers as often and as early as possible.

Musicians often start to practice at an early age. Most actors start acting in high school. The earlier you start doing research, the better your chances are.

Research is art

Research is art, there is no right way to do research. So far, I only believe in three general rules:

1. don’t be isolated… be part of a network
2. be open minded
3. don’t copy or try to be fashionable, always try to go further and lead

You might want to compare them with the Three Dijkstra Rules for Successful Scientific Research (but please, don’t compare me with Dijkstra):

1. “Raise your quality standards as high as you can live with, avoid wasting your time on routine problems, and always try to work as closely as possible at the boundary of your abilities. Do this, because it is the only way of discovering how that boundary should be moved forward.”
2. “We all like our work to be socially relevant and scientifically sound. If we can find a topic satisfying both desires, we are lucky; if the two targets are in conflict with each other, let the requirement of scientific soundness prevail.”
3. “Never tackle a problem of which you can be pretty sure that (now or in the near future) it will be tackled by others who are, in relation to that problem, at least as competent and well-equipped as you.”

Research is time consuming

Research does take time. Even the most brilliant scientists need time to get results. If you have a job while you are working on your Ph.D., always allocate research time first. Do not worry, you will find time for your other duties.

I do not think it can be easy to complete a Ph.D. if you do not spend at least three days a week working on your thesis. Yes, this means working week-ends. Most professors work during week-ends (though I do not), so nobody will feel sorry for you if you spend your week-ends reading boring research papers.

Success is completion or you have to get things done!

Brian Bialkowski says it best:

It won’t be easy and it won’t be pretty, but eventually you will finish. Sure, some of your work may be unreadable, but other parts will surprise you at their quality, and will be more concise and polished than anything you ever expected. It won’t be perfect, it might not even be good, but it won’t matter. In the end, you’ll have words on paper and your degree in hand.

To be clear, at some point, in any research project, you have to sit down and write. Whether what you write is great or just good, matters less than you think some of the time. You need to get a passing grade, but most of your work will not be insanely great. You need to know what the rules are so that your work is considered good, but not everything you do can be excellent.

Christoph Koch says it best:

The secret of success is to DO THINGS, e.g. solve problems, prove theorems, build systems, and the like, as early on as possible. Apart from good grades, publications make a very good impression.

To sum it up: don’t be nice, be productive! Many great researchers are nice people, but not all of them. And all of them have a nasty side.

How to get noticed by important folks

Professors are pretentious. If you pick up their work, study it carefully and extend, improve, or document it, they may notice especially if you did very hard work. Don’t try sucking up to them, it won’t work. Don’t be nice, work hard.

On quality… and content reuse or why the bad sticks with you

I just wrote you could get away with doing a half hearted job, didn’t I? Not really. Quality matters more than you think. Basically, you must complete your work, finish this thesis you started, complete the last run of experiments for this paper you are writing. If this means not getting the Nobel prize this year, fine, but don’t try to skim on quality too much. Just do not sweat it if it is not perfect.

• The secret to success in academia is content reuse. Most people, in a given year, will only create very little new content. Most of what we output is recycled. Courseware is recycled all the time. We even reuse other people’s work (such as when we use a textbook in a course). So, if you produce bad content, it will stay around with you for a long time.
• A lot of what we do in academia sticks around. The technical report, the thesis or the student evaluation from 10 years ago can still be around. If you have too much of it which is of bad quality watch out!
• If you really like what you are doing, you will go the extra mile and produce good quality: if you hate what you are doing, producing high quality will seem like hard work. So good quality content is a way to test your heart out.

Be determined

Is being smart important? Quite possibly, but I don’t think it is the most important quality. Desperate and determined researchers win every time. This implies two things:

• Being smarter is not as much of an edge as you may think. If you are too confident, over time, the desperate lesser researchers will pass you by.
• If you are not the cleverest person in your grad school, you can make it up by your determination.

Publish early and often

Whenever you have a really good idea, don’t spend too long before publishing it. Don’t sit on ideas until you can perfect them. For one thing, bean counters can’t count ideas you have in your drawers. Second of all, if you had the idea, someone else can also have it and publish it first. Not all publications need to be peer reviewed. You can publish technical reports which can be as simple as merely posting a PDF file on a web site. There are clearly cases where it is best to delay publication, but for most short and simple ideas, publishing early is the best strategy.

Research is about marketing your ideas

Marketing is extremely important in research. You found a new algorithm which improves by epsilon over another algorithm? Why should I care? You found a new algorithm which does not generally improve over another algorithm, but for a set of problems, it does much better, then by all means, go out and convince me that this set of problems is important and focus on it. Be honest about the flaws in your ideas, but do not be shy: explain to us why your results are interesting.

Recycle, recycle, recycle

It takes a long time to learn about a new field. Changing application area is fine, but you should always try to recycle what you already know. Don’t continuously start from scratch. The secret to success is building on what you have done and what you know better than anyone else. This means that you should be extra careful about what topics you pick: pick hard problems.

Research is a business

It doesn’t work like other businesses, but research is a business of sort. It is not free of politics, it doesn’t happen in an ivory tower, it is full of nasty people and extremely competitive. It is not about getting passing grades. It is not about doing what you are told. It is really like starting a small business in a crazy but not always nice universe.

Christoph Koch told prospective students: “If you think a PhD position is just a job, please look elsewhere.” As a Ph.D. student you are not an employee, the prof.’s helper or his secretary. It is most certainly not a routine job. You are not a soldier on the field following orders. You are someone building “something” on your own (mostly).

Peter Feibelman in “a Ph.D. is Not Enough” warns you that there are bean-counters who think the number of paper matters. This means you are compelled to demonstrate some level of scientific activity. Notice that scientific activity is not the same as greatness. In other words, a great scientist who hasn’t published in a while, will lose out to a mediocre scientist who publishes routinely. It is not so bleak as it seems: you should feel free to offer your own measures to the bean-counters, so that if you gave 20 invited talks in the last 2 years but published only one paper, you can stress that the total number of communications was 21. But you have to be able to count something, anything, and show that there is a constant flow of something (publications, talks, citations, etc.). I have never seen it done, but you could count the number of word or the number of pages you have published.

Choosing a research topic is a puzzle

Scott describes the act of choosing a research topic as a puzzle. I believe he is right! Don’t despair!

My own advice in this respect would not be too hard to be an oracle. You can’t foresee what needs to be done. You can’t foresee what the largest breakthrough in your field will be. You should be on the lookout, but don’t aim straight at it.

Instead, focus on little things that you can do better than most people and listen to what people around you are saying. Other researchers are your customers. If several of them say that they want this one extra element to be taken into account by your research, then do it! If you are not getting feedback, then you are not submitting enough papers, giving enough talks and chatting enough with people.

A good strategy for a thesis, I think, is to learn to do one thing very well, one thing where you have no competition, and then you apply it to different problems. As long as all of this is novel and non-trivial, and you produce enough work, it qualifies for a good Ph.D. thesis.

References are your friend

People cite each others in papers often not because they have to, but as way to cast their paper in a context. This establishes a web of trust: pay attention to often-cited papers and authors.

Search, search and search again

Keyword searching is a potent tool. Try the ACM library, Google, Google Scholar, and a collection of search tools by Todd Veldhuizen. Not everything is on the web and not everything can be found through keywords, but if it can be found and you didn’t find it, shame on you.

Don’t underestimate tools like wikipedia.

In a given field, seek existing bibliographies like the OLAP bibliography or the collaborative filtering bibliography.

When you find something interesting, organize it. Owen Kaser has online slides covering the search and organization of references and he points to Todd Veldhuizen’s slides on a similar topic.

Become a member!

While not all societies are useful, join the important societies related to your field. In Computer Science, I strongly suggest you become a member of ACM. Try out a few interest group. Network as much as you care for!

Setup a web site! Be visible on the web!

People may come across your name and they may google you: make sure that they find the information you want them to find. Make sure that all your publications are available on the web as freely as is possible.

Work on your presentation!

Presenting well is an important skill. There is a lot of advice out there on the web. There are several great pages of advice. Stephen Downes suggests the old “Winging it” book, maybe you can find it in a library. I like Bob Geroch’s suggestions for giving talks.

Remember that there are lightweight solutions for preparing slides these days: PowerPoint is not your only option. For example, you can use plain HTML together with some scripts and it even supports TeX.

Be a good student

You know you are a bad student if

• you cannot keep track of tasks assigned to you and be responsible for such tasks;
• you lie about what has been done and what hasn’t been done;
• you repeatedly ignore phone calls or emails.

Become a hacker

If your research has anything to do with IT, become a hacker which means:

• learn Linux, BSD or MacOS: Windows is not a good hacker platform;
• learn LaTeX, then learn to use aspell and style-check: it is a major pain at first but tools like LyX and Kile can help ease the learning curve, don’t try to avoid learning LaTeX;
• learn XML, RDF, RSS and all the nifty XML stuff: you can learn 80% of the important XML stuff in a week and it is will make you more credible;
• learn to use CVS or other version controls such as subversion, use them extensively, keep all your stuff in backed up repositories: it is a bit of a pain at first, but it will help you a lot later;
• learn Java, C++, Python, Ruby, PHP, SQL, Regular Expressions… as many high level languages as you can: you can learn Python in an evening so there is no excuse not to learn it.

Being a hacker will mean that you’ll be able to quickly build prototypes or, at least, determine how hard it is to build prototypes. It will also help you bring your inventions, if you have any, to industry and maybe, make some real money.

Don’t expect too much from the job market

The job market may or may not smile on you. The truth is that, in many instances, there are just too many people with graduate degrees. You can earn as much money without a graduate degree. You may even end up working for someone who has lesser degrees: this is quite likely.

Again, desperate people win every time: if you did many good things because you were nervous about your job prospects, you are possibly better off than the very smart guy who just sat around thinking he would get several offers easily.

If you go for a Ph.D., in most countries and disciplines, you are unlikely to get a professorship on the day you graduate in the city of your choosing. This may never happen. You may have to take jobs you didn’t know existed in cities you would never have otherwise visited. Your supervisor tells you otherwise? Be worried.

Most definitively, currently, a law or medicine career is a much safer bet on the job market.

For more comments, see my post on the death of the invisible adjunct.

Myself? I chose to live dangerously! (Actually, I had no idea what I was doing.)

Advice for getting research/academic jobs

Inside Higher Education has great advice for managing your early career.

Here is my own advice:

• Publish early and often. Always publish something each and every year after you receive your Ph.D. even if it is only a so-so paper. The length of the publication list is not the determining factor, but if you stop publishing for a long time, even if you are working on a brilliant paper or have other excuses, it will look bad.
• Publish at good places on a regular basis. Not all papers need to be great, but if all your papers are published at second or third tier journals or conferences, it will show.
• Never stop completely your research, even if you need to do research part time, on week-ends, keep at it. It is very hard to get back to research when you’ve stopped for a year or two.
• Teaching every year, even if it is only 50 minutes, gives you a small edge in the academic job market.

In my experience, people will judge you on the following factors (disclaimer: this is my opinion):

• Your current position: if you are tenured professor at MIT, you are very likely to make it on the short list. I have a theory that the current position one holds can be the single most important factor into getting to the short list. Basically, people don’t like to feel like they are promoting you or giving you a chance. They prefer to feel lucky you would leave a great job to come to them. Unfortunately, if you are currently unemployed, you might be in trouble. This is probably why people accept unacceptable working conditions just to hold decent enough jobs.
• Location of your current position: if you have a position in a foreign country or far away, most hiring committees will only short list you if you are extremely good. This means that your best bet is to have an impressive sounding position near several good academic or research institutions.
• Quality of the publication list: do you have some recent papers published at good places? Poor or average researchers can have long publication lists, but poor or average researchers don’t regularly publish in the best venues unless they work with great people. Make sure your publication list proves to them that you are either a good researcher or that you, at least, work with good ones.
• Research topic: do they care for the type of work you do? Working in a narrow field nobody cares about is not good.
• Clarity: surprisingly, some c.v.s are very confusing. Make sure people understand what you did, what are you doing and so on. Omit the marketing talk.
• Letter of interest: include a nice letter with your c.v. explaining clearly why you want the position.
• Interest: how badly you want to job? Calling up some of the professors to ask about the position can help you out. Any effort you can make to show you care for this particular position will play in your favor.
• References: did some famous person known to the committee recommends you? I find that if a respected researcher says a given candidate must be on the short list, it can make a huge difference. The reference doesn’t have to written on paper: you can ask someone to call up a friend on the hiring committee.
• Networking: are you connected? Do you know the right people?
• Honesty: when answering questions, don’t hesitate, don’t walk around the issues. If you have a weak point, come forward, with it. If people ever get the feeling you are hiding something, you are in trouble! Try to show you can prove everything you say. You will not get hired if people feel you cannot be trusted. By the way, this means that you should not be too creative with yourr c.v. If you feel you have to lie, you are probably in trouble and will not get the job. Better play it honest.
• Funding: prove that you can earn research grants. If you can’t apply for research grants, try applying with someone else who can. If you cannot get a grant, do some research to find out how you could get a grant. For example, check out how the current faculty members got grant and make sure you could also get the some of the same grants. Know exactly what you would do with the research money. Do you want to buy a powerful server? How many CPUs? How much RAM do you need? Do you need several workstations? Which brand? How much funding would you give any given student? Think about these issues.

Don’t do like I do, travel smartly

You’ll have to travel, eventually, if only to present a paper in a conference. Presenting your work outside your school is important: try to do it once. Michael Nielsen has a nice page of tips on how to organize your trips.
I also wrote a page on this subject.

Finally: get all the advice you can get!

Whether you like it or not, working toward a graduate degree means that you have to worry about life in academia and in particular, you must listen to what others have to say about what makes good research; make sure you read what the experts have to say about it:

• How to finish your PhD in a reasonable number of years
• John Baez’s advice page
• Academic Careers for Experimental Computer Scientists and Engineers
• Networking on the Network: A Guide to Professional Skills for PhD Students
• Principles of Effective Research by Michael A. Nielsen (a must!)
• You and your research (by Richard Hamming)
• Advice compiled by Michael Ernst
• The Ph.D. experience
• Yuhong’s tips
• Non-Technical Talks by David Patterson including How to Have a Bad Career in Research/Academia
• Advice on research and writing
• A Ph.D. is not enough (book)
• The Fisher files: a Physics professor at MIT, has produced a set of podcasts about how to get a job in academia, how to do good research and so on.
• The changing art of Computer Research
• How to Write a Lot: A Practical Guide to Productive Academic Writing (book)
• Advice for students starting into research work
• Resources for Grad Students by Owen Kaser
• Useful tips on how to succeed in graduate school and your subsequent research career by the UCSD VLSI CAD LABORATORY

When in doubt, ask for advice! You can’t get advice? Change supervisor!

Subscribe to this blog
in a reader
or by Email.