Philip Guo (Phil Guo, Philip J. Guo, Philip Jia Guo, pgbovine)

Reflections and advice on life as a mid-stage Ph.D. student

Summary
This is a collection of reflections and advice on Ph.D. life, written at the start of my 3rd year of graduate school. It is unique in the genre of "Ph.D. advice" articles since it was written at a time when I was in the middle of my Ph.D. and without any tangible achievements. In contrast, similarly-themed articles are often written by successful senior researchers in hindsight, not by struggling mid-stage Ph.D. students such as myself.

Update on 2012-05-07. Now that I am about to finish my Ph.D., I want to add one more piece of advice:

Here is one possible strategy for new students. For the first half of your Ph.D. (2 to 3 years), your only priority should be to help out on projects where somebody else has a huge incentive to publish. Who has a huge incentive? Assistant professors who are fighting for tenure, postdocs who are competing to get faculty jobs, and senior Ph.D. students who need to publish in order to graduate. Join onto their projects as a humble apprentice, make yourself useful, and learn how publishable research is done by actually doing it. Don't worry about defining your own original project or being the first author on papers.

If you follow this advice, then most of your problems with procrastination, getting stuck, feeling lost or overwhelmed, work-life balance, etc. will naturally disappear. And by the time you begin the second half of your Ph.D., you will have the experience necessary to define a respectable and publishable dissertation project.

{An interesting take on the above passage from my friend Shaddin:

I will emphasize one point that is perhaps implicit in your writing: the best way to reach your goal of defining your own research agenda is to work with as many people as possible more senior than you on projects that are roughly in the vague general direction of your interests, possibly as a "tag-along" collaborator at first. The main reason is that this is a very efficient way to learn the skills that you need, by watching others perform them. Another reason is that sampling different points in the "research problem" space gives you the layout of the terrain, which you can then use to identify where in the terrain you would like to play.}

In contrast, who does not have a huge incentive to publish? People who are working on follow-up "nice to have but not crucial" papers to accompany their original already-published paper. Also, tenured professors, especially those who are already famous. As a new Ph.D. student, the worst-case scenario is if a senior tenured professor suggests for you to play around with their "pet ideas" or "side projects." You will likely feel lost or stuck, because they have little incentive to be an effective advisor on such projects. I don't blame them—they no longer need to publish to survive, and when they do want to publish, they will probably focus more energies on their main projects.


I wrote this loose collection of notes in September 2008 just prior to starting the 3rd year of my Ph.D. in the Stanford Computer Science Department. At that time, I had just gotten my first few paper submissions rejected and was trying to pump myself up to start a new academic year. While riding on the train, I took some time to reflect on my past two years in the Ph.D. program and internalize bits of advice that older students had told me.

This article is unique since other "Ph.D. advice" articles are often written by successful senior researchers in hindsight, not by struggling mid-stage Ph.D. students such as myself.

Here are some standard disclaimers:

  • I wrote these notes after only finishing 1/3 of my Ph.D. and without succeeding in publishing anything yet, so some of this advice might be incomplete, naive, or misguided.

  • Some of these observations might not make any sense for your particular situation. Ph.D. students in different schools and even in different departments within the same school often have very different experiences, so it is hard to provide truly generalizable advice.

On boot camp

The best way to prepare for war is by going to boot camp and doing military training, but you can't really be psychologically prepared for what you'll face in a real war-zone until you actually go.

Similarly, the best way to prepare for doing research as a Ph.D. student is by doing research as an undergrad, but you can't really be psychologically prepared for how you'll feel as a Ph.D. student until you actually live full-time as one for a year or two. Even if you are doing the exact same tasks day-to-day as when you were an undergrad research assistant, you'll likely feel much different since your primary purpose for being in a Ph.D. program is to produce research (vs. in undergrad when it was only one of several things you did).

On the paramount importance of grinding

If you talk to people who have already gotten their Ph.D.'s and ask them to trace backwards in time from their dissertation topic back to the beginning of their time in the program, you will likely find that they did not take a direct path towards their dissertation topic right away. Most likely, you'll find that they spent their first few (typically 3 to 4) years working on various projects in related fields or areas but on topics that don't seem to resemble their final dissertation topic at all, and then in their final 1 to 2 years, they figure out the topic and work on it, maybe publish a paper or two, and then finish their dissertation.

What is often more interesting to learn, though, is how their earlier projects actually often led them to eventually discover their dissertation topic, exerting influence in indirect ways such as exposing them to ideas, tools, related work, or even people that guided them towards their eventual dissertation.

Therefore, during your first few years of Ph.D., it's very important to be grinding (working intensely) on research that you feel will get somewhere in terms of publishing, but not to worry about whether this work will define you as a grad student and be your magnum opus of a dissertation. Don't be afraid to tag along on existing projects led by professors and senior grad students, and don't rush to try to be the first author on papers right away.

Remember, actively grinding—even on research that won't eventually make it into your dissertation—is always better than being idle.

{Some caveats from my friend Kevin:

True, but in labs where lots of grunt work must be done (biology labs with real animals, psychology labs running subjects, chemistry labs, ocean engineering) you need to be conscious of ...

a) "pulling your weight" in terms of grunt work, especially in the first few years or you'll get on the bad side not only of your advisor, but of everyone in your lab who you will need to rely on in the future to help you get data, etc.

b) not pulling more than your weight. everyone (research staff included) is trying to get stuff published and move ahead in their career. that means if you blindly grind, you could end up doing tons of grunt work that won't lead to publication.}

{Extra notes from my friend Neal:

now that I'm nearing the end [of my Ph.D.], I'm shocked at how all of the things I did in my early days, which at the time seemed like random unrelated things, and what i'm doing now magically came together to make a coherent story. i didn't realize it at the time, but the stuff i did in my first year, though not related DIRECTLY to what i'm doing now, gave me a framework in which to think and conceptualize. different projects related by a THEME...it happened accidentally, but it's amazing how that accident seems to happen to everyone. you don't need one big project for your dissertation; instead you can connect 2-3 small projects by a single, central theme. (and be creative about formulating that theme to include everything.)}

On paralysis

The worst thing you can do as a grad student is not doing anything at all; doing some sort of research-related grinding is always better than being paralyzed. Paralysis can come pretty easily to grad students because the task that you need to do is so ill-defined and long-term by default, so you need to always keep trying to move forward, setting small, concrete goals along the way.

One bad sign that you are proceeding deeper and deeper into paralysis is when you don't even want to think about or to talk about your current state of paralysis. Without directly confronting the source of your paralysis, there is simply no way to overcome it and you are probably better off dropping out of Ph.D. altogether.

On impedance matching

In order to get something non-trivial published and to earn a Ph.D., you need to make sure that your advisor is interested enough in what you're doing and wants to make an effort to push you to publish.

The more ambitious but far more difficult way to accomplish this goal is to try to formulate an idea by yourself and simply hope that it's so damn good that your advisor will instantly jump on-board. Sadly, this rarely happens. A more realistic approach is to find something that your advisor is already interested in (or find another advisor who is) and figure out how your interests can match with his/hers.

The reason why it's often difficult to "sell" your advisor on your idea is that you need to first produce proof that it's actually promising, which can be difficult to do alone with no support. Since professors have seen many half-baked ideas in their years of experience, they are probably very skeptical about hot-shot junior grad students pitching ambitious research ideas to them without any evidence that those ideas have potential. Be warned that the burden-of-proof can be huge, since as a lowly Ph.D. student, you won't have a prior reputation or track record to help your case when making your proposal.

{A caveat from my friend Kevin:

Be more clear that [while] it is rare that your advisor will jump on board, it is not-so-rare that you might try to formulate something yourself. It is important to work on a specific problem that your advisor is interested in if you want his or her full support (and not just a problem in the general field s/he is interested in). That said, I think it's a very good exercise to try to formulate an idea on your own, pitch it to your adviser to make sure it's not totally wacky, and spend a 3 or 6 months on your pie-in-the-sky idea. It will make you appreciates others' (like your advisor's) ideas more, and you will learn a lot in the process.}

{Another note from Kevin:

You may want to make the point that advisers are people with emotions, not machines of research. In the same way that you are not interested in some particular chapters of your "discrete mathematics" book, and you are interested in other chapters of it, your adviser is interested in some problems in his field and not others. Think about what your attitude was towards learning things you didn't really like from your discrete math book (e.g. learn it just for the test, then forget it), as opposed to something you really liked, which you were driven to really understand it.}

On what skills Ph.D. graduates possess beyond typical bachelor's or master's graduates in the same field

For instance, if a company like Google is hiring software engineers, why would they prefer to hire Ph.D.'s over bachelor's or master's graduates, even if the Ph.D. grads are often doing a job unrelated to their areas of specialty? Here are some general skills that Ph.D. graduates usually possess that not as many average bachelor's or master's graduates do:

  • The ability to cope with ambiguity and uncertainty in task requirements.

  • The ability to set achievable concrete goals and to work towards them, even without much external influence.

  • The ability to grind hard and to focus for long periods of time, even in the face of failures.

  • The ability to actively seek out information and to learn new skills, tools, technologies, and knowledge on-demand as the given task requires.

  • The ability to take constant criticism and discouragement from things not working properly.

  • The ability to learn lessons from failures and to recover gracefully.

  • The ability to properly frame the "big picture" importance of one's work and to communicate it with peers in the form of oral or written presentations.

(Of course, exceptional bachelor's and master's grads will also possess these qualities, but they are by no means necessary for earning those degrees.)

{A note about oral presentations from my friend Neal:

another thing i'd add is that one should take as many opportunities for oral presentations of research as possible. preparing for presentations forces me to reflect on and appreciate what i've done, contextualizes it, and tells me where to go next. on top of all that, it's a very useful skill to speak eloquently about your own work. and it's also nice to have beautiful slides prepared in case you ever need them (and you need them more than you'd think).

these are some reasons and me and a coworker have begun a student-run seminar series for grad students to talk about their own research in the chem dept. it's going well so far, and i think people are getting something out of it.}

[Addendum about oral presentations added on 2011-07-30: One of the most important skills I've gained as a Ph.D. student is the ability to confidently deliver a technical presentation and, more importantly, to answer tough questions on-the-spot. Nobody is born with this ability; there's simply no substitute for repeated practice.]

On why it's harder to grind as a Ph.D. student than as an undergrad

Undergrad grinding (e.g., on homework and class projects) is pretty much linear, so that every step along the way, you grind for N units of effort and get incrementally closer to your goal of finishing your class, semester, or degree. Even if the path isn't totally straight, at least you won't have to do much backtracking. It's a steady, fairly predictable march from beginning to end. Of course, it will require effort and grinding (and you might face failures along the way), but the path is laid out for you.

In contrast, Ph.D. grinding is like doing a breadth-first search over an unknown space. You could grind for N units of effort, then your attempt fails, and then you have to backtrack to your starting point, tweak your direction, and grind again for however many more units of effort. If that still doesn't work, you will need to keep trying until something does start to work. Then you repeat again and again, constantly with the prospect of backtracking.

However, the backtracking doesn't mean that your efforts were wasted, as long as you are conscientious enough to derive lessons, wisdom, and concrete skills from those failed mini-attempts so that you will increase the probability of success on your next attempt. Thus, you should expect many of your experiments to "fail" in the sense of not producing tangible, useful results towards your research goal, but you should always be trying to learn from such attempts. The real failure comes when you ignore your failed attempts and avoid thinking about them because they are too discouraging.

On what is a bad reason to do a Ph.D.

A really bad reason to start a Ph.D. is because you want to simply learn more about a subject in more depth. A Ph.D. is not a 6-year extended master's degree; it is qualitatively different. You are producing "products" of knowledge+grinding+creativity, not simply absorbing more information. The learning you do as a Ph.D. student is a by-product of producing research. You are learning-by-doing; nobody directly teaches you how to do research. It's an active type of learning where you need to take the initiative both to learn and to figure out what exactly to learn in the first place. It isn't a passive type of learning where you know what you need to learn and simply need to put forth the effort to do it.

On understanding what academic research is and isn't

There is a big difference between what seems like research and what is actually acceptable, publishable academic research in your field at this particular moment in time. Out of the set of all cool, seemingly-innovative, and intellectually-stimulating ideas, only a small subset of those count as publishable research in top-ranked academic venues (which are reviewed by top-ranked academics at various universities and industry research labs). Sadly, an even smaller subset of publishable ideas are within reach of what you can feasibly do as a Ph.D. student, since you are bound by your advisor's area of expertise (which can greatly help to get your papers published) and also by the equipment, tools, and frameworks available in your lab.

A good way to seek out research problems as a Ph.D. student isn't by starting from a vacuum, reading papers, and generating ideas (because most ideas that you generate will not fit within the definition of publishable academic research), but rather by taking into consideration existing constraints such as the tools, frameworks, and people available to you and then figuring out how/where you can contribute within those constraints. Start first with constraints, not with a blank slate.

On avoiding working in a vacuum

Be very wary when you start your work session with a pen and a completely blank piece of paper. While this might work for some forms of research, the problem with starting tabula rasa is that it becomes too easy to brainstorm pie-in-the-sky ideas that are too unfocused. Research (especially for a Ph.D. student) never occurs in a vacuum and must be built upon an existing infrastructure and/or related work.

On the importance of setting small-scale incremental goals to make sure that you consistently grind every single day

You should train yourself to be uncomfortable if you wake up in the morning and don't have a good idea of what you are going to work on during that particular day. Or if you find yourself sitting in front of a blank piece of paper theorizing on what great idea you might be able to come up with in a vacuum. Or if you find that you've been utterly unproductive for the past few weeks. Inaction leads to paralysis!

Set small-scale, concrete goals for yourself and work hard every day to meet them (with occasional well-deserved breaks). Not knowing what you are going to be doing at the start of each day will quickly lead to you feeling unproductive for long stretches of time and then starting to feel paralyzed.

Remember, even though the Ph.D. experience consists of semi-blind breadth-first searches where you will have to face many mini-failures and frequently backtrack, those backward steps are really where you learn the most and gain your skills and wisdom. So if you don't keep on grinding (and likely failing), then you won't ever have a chance to truly learn how to get better at research.

Passive learning by reading papers, taking classes, and listening to talks can only get you so far, but the best kind of learning for how to produce publishable research is by attempting to do so and hopefully receiving useful feedback from others along the way.

A friend once told me that she heard from a famous professor that if you're completely in a rut during grad school, then just try to work for three hours every day ... no more than that. Consistently putting in a few light hours every day, combined with talking with and getting feedback from your professors and fellow grad students, should hopefully be enough to get you out of your rut.

On your choice of projects as a Ph.D. student vs. as a more senior researcher

When you're a Ph.D. student, it may be a good idea to tackle well-established problems that you don't need to "sell" too much, because you need to first gain your "street cred" by doing work with very clear and indisputable metrics for success.

When you're young, you need to figure out the rules of the game and how to play well and gain recognition, but when you're already well-recognized, then you can try to break some of those rules and to change the game ... fields often move forward by breakthroughs led by experienced researchers, not the opinions of hot-shot grad students.

For example, a 25-year-old Ph.D. student and a 55-year-old tenured professor might have the same wise long-term view and opinions on a particular research field, but nobody is going to invite the Ph.D. student to write position papers or to give invited talks about his/her viewpoints. The only way for the young Ph.D. student to prove himself is by playing the game well and excelling in the metrics that the current leaders in the research community have set.

{A note from Kevin:

I see your point, but this sentence isn't the whole truth. Several big breakthroughs in fields occurred to people [who] are naive enough to ask the right questions (cliche, I know, but true). I think instead of telling people to be conservative in their research goals, just tell them to be conscious of the trade-off between guaranteed but not-too-exciting results, and risky but exciting results. It's just like if you are a high school baseball MVP, should you major in accounting with a 90% chance that you'll becoming and accountant, or pursue a career in major league baseball with a 1% chance of success. Except in grad school you can actually pursue both. Ed told me his adviser always puts grad students on two projects: one safe but not too exciting, and one not-so-safe but very exciting if it works. [...] Just emphasize that the probabilities are pretty low to make a big breakthrough, but if that's what your heart really wants to work on and you don't have other constraints [e.g., funding, family, lab resources, lack of advisor permissions], then go for it.}

On dealing with civilians

Civilians (people not in Ph.D. programs) don't understand at all the Ph.D. mindset, so it's often not productive to try to explain your Ph.D. struggles to them, since you will likely come off sounding aloof, arrogant, and like you are mentally masturbating in your Ivory Tower. To the outside world, grad students have the privilege of doing "research" in some Ivory Tower while civilians need to actually work for a boss 9-5 for a living! Civilians will not sympathize with you or relate to you well.

When civilians ask about your research, it's probably a good idea to simply talk about very high-level goals of your entire research area, preferably with motivating real-world applications that they can envision, rather than talking about your specific projects. If you start talking about your project, either they will tell you that it seems lame or boring (at which point you will start crying inside) or they will just politely say something like "wow sounds interesting and hard" followed by an awkward silence (at which point you will need to quickly change the subject). Also, don't say stuff like "it's really hard" or "it's really difficult" because then you will appear arrogant and to be implying that they are too dumb to understand all the smart stuff you do.

Civilians also tend to clump Ph.D. students together with Master's students, law students, medical school students, and any post-baccalaureate program, all under the umbrella of "grad student". The most salient difference between Ph.D. students and all other post-bachelor's students is economic; Ph.D. students are paid salaries (e.g., by fellowships, research grants, or the department) while all other types of grad students pay tuition. Because of this, Ph.D. students are actually employees expected to be producers of goods (mainly in the form of research papers) and services (in the form of assisting professors with grunt lab work and other stuff to impress funding agencies, helping them earn tenure, and being teaching assistants), whereas all other types of grad students are expected to be consumers of information. This is why I always prefer to use the term "Ph.D. students" rather than "grad students" when facing civilians.

Finally, when civilians ask you about grad school, they will typically ask about how much you like TAing and also about what kinds of classes you are taking, since the only exposure most undergrads get to Ph.D. students is by them being TAs, and logically since you are still a student, you presumably take classes. They will likely ask how well you are managing to keep your grades up and what kinds of interesting classes you are taking, since their only reference point is that grad school is like undergrad.

On taking breaks

Being a Ph.D. student is a full-time job, literally, not just full time as in 9-5, 40 hrs per week, but literally FULL time, every waking moment. It is possible to agonize about your research all the time, night and day, since it is never "good enough" until you get all of the signatures on your dissertation.

Because of this, Ph.D. students often feel guilty taking breaks from research, feeling "so unproductive" during times when they are not working. For civilians with regular jobs, there is no need to feel guilty during non-work hours since nobody else is working either and nobody is supposed to be working during off-times. But Ph.D. students can technically work all the time, 24/7, so any time that you don't spend working can be thought of as "unproductive" time.

You need to definitely get over the guilt of taking breaks, as long as you feel like you've worked consistently and worked hard during your bursts of productivity. You can't possibly be productive 100% of the time; if you are productive several hours at a time in bursts, you should take well-deserved breaks without feeling guilty, and then recharge your batteries.

That said, though, it is tempting to do a burst of "fake work" such as responding to personal emails or updating your blog, saying to yourself that you deserve a break, and then taking a long break. You want to make sure that the work you're doing before taking breaks is laser-focused on your research goals.

On Seeking diverse sources of advice

Finally, it is a good idea to seek advice from diverse sources—professors, postdocs, fellow Ph.D. students, and even the civilians in your life (i.e., friends and family). As a starting point, here are some links to webpages that provide advice for Ph.D. students:

Note that, unlike this article, all of the above links are from people who wrote articles in hindsight after finishing their Ph.D. and becoming successful researchers, so survivor bias is definitely present.

Created: 2008-09-20
Last modified: 2012-05-07