Philip's Guide to Grad School

by me
This article is intended for people who've asked me to write them recommendation letters to graduate school. It may also be of interest to others applying to graduate school at a research university.

This Is Not A Meritocratic Process

If you're reading this it is probably because you finished 6.171. Congratulations. In addition to being a capable Internet application engineer this also means that you got into MIT as an undergraduate. Undergraduate admissions are a meritocratic process. The gnomes in the admissions office do their best to get 1100 of the world's smartest 18-year-olds to show up at 77 Mass Ave in September. Your worthiness as a human being is assessed from test scores to high school grades to extracurricular activities. Those who are most worthy get in.

Forget all of this as you apply to graduate school at a research university. It might be true at a professional school where you're paying tuition, e.g., at a law or medical school. But at a research university graduate students are paid and are part of the revenue production system. If an undergrad does a bad job, the university pockets his tuition money, gives him an F and makes him repeat the course. If the undergrad does a really bad job, the university may eventually kick him out after pocketing $50,000 or $100,000 in tuition. A tragedy for the individual? Perhaps. A problem for the institution? MIT reported $9 billion in assets in 2002 and Harvard College reported $25 billion (source www.guidestar.org).

As a graduate student you are an employee. If you are TAing a class and doing a noticeably bad job, the university will have to pay someone else to take your place. If you are doing research for a professor the situation is even more delicate. Jane Professor promised the NSF or NIH that she would produce Results X,Y, and Z in exchange for $3 million. If you do a bad job for Jane she may be in default of her promises to the federal government. NSF might decide to revoke the grant or not issue a follow-on. Jane might have to lay off her other research assistants. If Jane does not get grants she will not get tenure. As an undergrad, your incompetence got you an F. As a grad student your incompetence will get your professor fired.

Professors do not like to be fired. In fact, university professors are some of the most risk-averse people in the world. A risk-loving person drops out of college to trade commodities, deal drugs, or sell arms. Half of these people are so rich that you'll never meet them unless you are from Greenwich, Connecticut. The other half are dead. A risk-averse person asks "What kind of a credential can I get that will ensure that I'm never unemployed?" A risk-averse person asks "Where can I get a job that, even if it doesn't pay so great, has the potential to give me a lifetime guaranteed income [tenure] no matter how incompetent and out of date I become?"

Suppose that I give you a briefcase with $1 million in cash and say "You get to keep this if you will give me a computer program that does X,Y, and Z but you aren't allowed to touch the computer yourself." If you want the cash you're forced to hire someone to write this program. It looks pretty straightforward and you think you could knock it out yourself in two weeks. Now I give you a stack of 100 applications. You must admit one of these people to your grad school and then he or she is going to write this program. You're in luck. The 99 smartest people in Mainland China have applied for this position. The remaining applicant is your old roommate Fred. Fred was a reasonably bright guy but certainly he would not stand out as one of the 99 brightest in a group of 1.3 billion. Imagine how much you could learn from one of these 99 Chinese applicants! Imagine the new worlds that you would discover together! Imagine the stimulating conversations once this person learns to speak English really fluently. Hmm... what if communicating the specs to this Chinese genius is difficult due to his or her not being a native English speaker? And what if some of the numbers on this application were fraudulently obtained? Can you be sure that the GRE and TOEFL tests are administered in China in a fraud-proof fashion? And how about these recommendations? They say that the Chinese genius is the biggest genius ever to live in China. But you don't know the recommenders personally. What if they're not very discriminating? What if they are biased because they are good friends with the applicant's parents? Do you really need a genius anyway? You get the $1 million as long as the program works. You've seen Fred crank away on his UROP at the Media Lab. You've seen his running system from 6.171. Are you going to risk losing a guaranteed $1 million on the chance that the person from China will look just as good in person as he or she did on paper?

Late in August of 1978, a doctoral candidate named Theodore Streleski walked into the office of his thesis supervisor at Stanford University, and bludgeoned the professor to death with four blows to the head. Streleski, who had been enroled in graduate school for almost half of his forty-two years, had completed a baccalaureate in engineering physics, a mastership in mathematics, and an all-but-typed [sic] doctoral dissertation on "approximations and interpolations of function [sic]." He had also entertained the thought of murdering his thesis supervisor for years, taking that bloody plan as the only workable means at his command of attacking the graduate program he held responsible for depriving his life of meaning. "Stanford, with de Leeuw's help," he said grimly of the institution and the professor, "took nineteen years of my life with impunity and I decided I would not let that pass." Obsessed like Dostoyevsky's unbalanced protagonist Raskolnikov, Streleski set out to enact his own version of Crime and Punishment. "I decided I'd do something about it." he explained. "I did. Under the same circumstances, I'd do it again."
[source: wire service story]
What do you know about the personality of the applicants who look so great on paper? The undergrad with the annoying voice and Star Trek T-shirts; you only had to suffer with him for 3 months and then he was gone, done with your course and off to make some other professor's life miserable. The depressed undergrad who drops out after 7 months? We express concern for his or her welfare but mostly we deposit his or her tuition checks. It is all different with a grad student. We're going to be living with this person for 6 years, possibly in a rather small office. If he gets depressed and drops out, we lose our grant. If he has an annoying voice, we lose our other grad students. If he is unlucky with the ladies and makes unwelcome advances to some nearby females, we get sued for sexual harassment (he's our employee). And if that isn't enough to get us worried there is always the possibility of violence: "Graduate school students have been killing their professors and related personnel since before murders committed by students in lower levels of the American educational system caught the public's attention." ["Murder in the Department: Factors of Organizational Deviance in Graduate School-Related Homicide" -- Dr. Marilyn Dudley-Rowley]

You're going to hire Fred, the person you know and trust to do a reasonably good job. You're not going to admit the most qualified applicant. If there is an admissions office intervening between you and Fred you're going to lean on them to get Fred in. This is why so many grad students at School X were originally undergrads at School X; they became known to faculty as undergrads. If you run out of undergrads from your own school you start to consider undergrads who've been working with your former grad students who are now professors at other schools. If you run out of qualified candidates whom you know or whom are known to your friends and still have some slots left over you might look to the admissions office to fill them with people who sound promising, but otherwise you're not interested in taking a chance. [As long as we're on the "taking a chance" subject, note that you hear little about affirmative action for grad students at research universities. Nearly every university bureaucrat is in favor of giving minorities with poor test scores a chance at the undergrad level. If Bill Favored-Minority flunks out, well, that's a shame but at least we pocketed $75,000 of his parents' dough before he left. And affirmative action is okay for the tuition-paying mobs at law schools, business schools, and medical schools. But when it comes to grad students who are really employees in disguise, any social goals are tossed out the window in favor of getting the job done with the least possible risk at the least possible cost. Risk is something that you take with other peoples' money.]

What does this mean to you? You can assemble a beautiful application with a recommendation from me saying that you are God's gift to computer science and fight it out with 10,000 other applications that have been tossed "over the fence" by eager beavers worldwide. Maybe they'll fill 20 slots from this pool (6.041 grads note that this is a 0.002 probability of getting in). Or you can get to know a couple of people (professors) who need workers (students) and convince them that you're the right person for the job and have them pull you in from the inside.

You're Applying for a Job

As noted above, this is not a beauty contest. You'll be spending most of your graduate career working on a small team for a single supervisor. If you want a job at GE you don't send your resume to the personnel department at headquarters. You try to get to know some of the groups within GE, figure out which is the most interesting to you and the best match for your skills, then try to meet some of the people there. Treat a research graduate school application in the same manner. You are applying for a job.

Ideally you want to pick a research group rather than a university. Suppose that you were stuck in the basement of Building 36 for 6 years with a bunch of people that you didn't like, working on a problem that bored you and struck you as unimportant. Would the fact that you were at MIT make that life more appealing? Would you even be aware of the fact that you were on the MIT campus? Even though MIT has many wonderful people and problems and a great reputation, none of that is helping you in any way during an unlucky grad school experience. Think back to the very worst professor that you had during your years at MIT. Imagine spending 6 years in a small room with him or her and having that person totally control your intellectual life. Sound appealing? Well, that's MIT grad school for at least a handful of people. Wouldn't you rather be at University of Massachusetts working for someone more interesting and supportive?

How do you pick a research group? If you're applying somewhere locally you start by wandering around the halls, asking grad students questions about their lives and projects and asking admin assistants for copies of recent papers by faculty members. To explore schools nationwide start by looking at the most respected journals in your area of interest. Every paper has authors and every beside every author's name is an institution. Eventually you will uncover clusters of people whose work you find interesting. These are the schools at which you should be applying. Because you're at MIT you have another avenue of attack: the visiting speaker. University researchers are irresistibly drawn to the coach section of airliners. Even after they get tenure they can't resist getting on a flight to Boston and standing up in front of the assembled mob to talk about what they have been working on. Join the mob, listen to the talk, and if it interests you get the speaker's business card. Then proceed as per the next paragraph.

The next step is going to visit potential employers. Before visiting any professor you want to make sure that you read his or her most recent publications. These folks live and die by citations, i.e., how many people are paying attention to their work. The most flattering thing that you can say to a professor is "I read your paper in ...". Take notes on each research paper and try to answer the following questions: (a) What is left unanswered by this, i.e., what would the next interesting experiment be? (b) What implications do these results have (if your field is engineering, concentrate on practical applications)? Now you have a good excuse for asking for a meeting with the paper's author. You've read the paper and would like to discuss what further experiments are worth pursuing. You can segue from that discussion into the question of whether or not the professor might like to hire you to pursue that further experiment right here and now (remember that "hire you" = get you admitted to grad school or actually hire you in as a staff member (more on that later)).

If you have a few months between your discussion and the admissions decision, try to start a collaboration with the professors that you think are most interesting. Get a small problem from them and start working on it, giving them periodic written progress updates (this may not be practical in laboratory sciences such as biology). Now they can tell that you're a self-starter who is easy to manage.

If you actually get into graduate school you're going to have to do all of this anyway. You have to make an informal survey of the research projects going on at your institution and try to get into a group that seems interesting. If you start this process right now you not only get started earlier (which means you'll finish earlier) but you also have the advantage of the professors getting to know you.

You might not get to know any of these professors well enough to ask them to write you a recommendation but at the very least you can ask them to send a note to the admissions committee saying "I met this applicant and he/she seems like an easy person with whom to work".

What I Can Do For You

As noted above the opinions of those with whom you're about to work are much more important than the opinion of an old instructor like me. Nonetheless a letter from me may be of some use in meeting the admissions office's formal requirements. There are two kinds of letters that I can write. The standard letter includes a ranking: "This student is the best that I've seen in 10 years" or "This student is better than 95% of his peers". You really don't want me to play this game. First of all, you were at MIT. The average student at MIT is pretty able. Second of all, you were in 6.171. The weaker students in Course 6 hear about how much work it is and don't register for it. The mid-grade CS students often falter getting their PCs set up with the tools and drop out before the first week. So the only MIT students that I personally see are those who were fearless enough to sign up for 6.171, persistent enough to get their tools working, and tough enough to survive the semester. Remember what a bunch of hardcore nerds your classmates were? Can you say in good conscience that you were the best person in the class? If not, you're in trouble for the standard letter.

Fortunately, it isn't clear that the standard letter is the best. Admissions committees see a million of these. They are the hallmark of the undergrad who has done nothing except what he or she was told to do. They say "I signed up for some classes; I did some problem sets; I got some grades." As you saw in 6.171 that kind of person may or may not be successful when asked to display initiative, when faced with an open-ended challenge, when called upon to describe succinctly his or her work. Grad school is more like 6.171 than like 18.01 or 6.041. You're going to be asked to push forward a loosely defined project, not plow through problems on a problem set.

The alternative to the standard letter is the narrative. I tell a story about our time together. The story illustrates what a capable person you are. Ideally the story shows rather than tells. The reader is left to draw his or her own conclusions about your suitability for graduate work. The letter does not say "Joe Foobar is brilliant" (telling) but rather "With no supervision over a two week period and never having used a relational database before, Joe Foobar rebuilt our genomics database so that it ran 100 times faster and added a data warehouse so that biology graduate students could explore our results without being SQL experts" (showing).

Examples of narrative letters:

What stops me from writing one of these for you right now? The first is senility. When I was your age my memory was near perfect. But if you're reading this at 6:00 pm it is a fairly good bet that I can't remember what I had for lunch. For you, your semester in 6.171 stands out among your memories like an especially painful visit to the dentist, perhaps a wisdom tooth extraction. For me, the database-backed Internet application that you built was one of about 500 that I've been involved with in some way since 1994 (yes, they had computers back then). So you really need to draft the letter for me! I can embellish it with some information about how hard 6.171 is and how smart you are but you should lay out the basic facts of what we did together: the challenges presented by the course, the problem set forth by the client, what you did to attack the problem, etc. Write it as if you were me, e.g., "Billy Bob conducted a user test of his software and, after measuring the amount of time it took for people to accomplish various tasks, redid the page flow to streamline the most common tasks." If you're applying somewhere other than MIT just the basics of what you did in 6.171 will be reasonably impressive. The calendar was short, you were taking other classes ("While taking 4 other courses, Bill Bob..."), you took a lot of initiative, you offered useful criticisms to other teams in class ("Billy Bob demonstrated an ability to think on his feet when he suggested to Team X that they ...").

The second thing that keeps me from writing you a great narrative letter is that I don't know exactly what you want to do with your life, near-term or long-term. It is always better if it sounds like you have a plan. Universities don't like to think that people are applying merely because they can't find a job. It is always much easier to get support from a university to finish a project than to start one. So if I can say "From the very first week of 6.171 Billy Bob said that he wanted to study skin pattern evolution in guppies, which is why I think he'll do a great job in your evolutionary biology program" that's much better than "I'm glad to see that Billy Bob, after drifting aimlessly for years at MIT and spending a lot of time with his Playstation 2, has finally found a goal in life."

So drop me an email with, in addition to the draft narrative, answers to the following questions:

  1. What's your 15-year plan? (Every good researcher has a 15-year plan. This is about problems that you're going to attack and solutions that you're going to develop, not about positions that you're going to obtain.)
  2. How is the particular graduate program to which you're applying going to help you in your 15-year plan?
  3. What are you going to work on in grad school? For whom?
  4. What have you done to search for a compatible research group? Of the papers that you read, which were the most interesting and have you talked to their authors?

What If You Don't Get In

If you don't get a job at Microsoft or Oracle you can't simply show up and start working anyway. They have guards posted at the front desk of every building to keep you out. Suppose that you get into Stanford. You're going to show up, wander around the research groups trying to find a reasonable supervisor with interesting goals, sit in on some lectures, and maybe do some programming or writing. Suppose that you don't get into Stanford. What stops you from showing up, wandering around the research groups trying to find some interesting work, sitting in some lectures, and doing all the programming and writing on your laptop or PC at home? If a high school kid tried to do this at MIT, there would be some pushback. What does his mom think of him hanging around? Is he going to get his nose into a beer keg? Will his complete ignorance make him a liability? But you're an MIT graduate. You ought to be a useful resource in almost any technical situation. Maybe a group doesn't have budget or inclination to hire you for a big salary. Maybe a group doesn't have the necessary pull with the admissions office to get you into school. But nobody is going to kick you out (esp. if you volunteer to do some sysadmin/dbadmin work!).

Grad school at a research university can been done almost as efficiently without being an enrolled student. In fact, suppose you were to get hired as a junior staff member in a research group. You'd be working on research, just like a grad student. This research could be used as the basis for a PhD thesis, just like a grad student's. You could attend classes and do problem sets, just like a grad student. The only difference between a staff member and a grad student is that the staffer gets paid 3X as much. In some countries, Germany for example, the concept of grad-student/research-slave does not exist. A person gets a PhD by becoming a staff member in a research group and, after a period of time, writing up his or her research results in thesis form. Instead of the enforced poverty and celibacy of the American grad student, the German PhD-seeker is earning a normal salary and living a relatively normal lifestyle among neighbors. The German-style plan doesn't work indefinitely in the U.S. Eventually you must register as a graduate student to finish your PhD. However, after a year or two as a loyal and productive staff member in a research group your chances of admission as a graduate student are enormously improved. And remember that the year or two hasn't been wasted; you're pretty much just as far along towards your degree as if you'd spent that time as a registered student.

So if you're serious about solving a problem and you've found a research group where you'll get support for working on that solution, just start. Show up. Show up as a volunteer and try to get hired as a staff member. Simulaneously apply to grad school. If you get in, great. Continue what you're doing and adjust your spending to fit your new (derisory) salary. If you don't get in, fine. You'll get in next year.

What If You Do Get In

Congratulations. You're a grad student now at a prestigious research university. One of our colleagues was just like you once, an eager beaver starting his first semester in MIT EECS. Unlike you (I hope), Mr. John Beaver (not his real name) had pled guilty to a federal drug possession charge. During IAP he went home to appear before a judge and was sentenced to 1 year in prison (joining 2 million other Americans; we have the highest incarceration rate of any industrialized nation). He served his time and came back to MIT. In his final year of graduate school he was complaining about how much he hated his life, hated being poor, hated his thesis, hated his advisor, and hated MIT. His officemate, trying to cheer him up, noted "Well, John, at least it is better than being in prison, eh?"

John Beaver the grad student leaned back and reflected for a moment. Slowly he responded "... actually when I was in prison I had a lot more optimism and zest for life than I do now."


philg@mit.edu
Add a comment | Add a link