This post’s intended audience is extremely narrow: prospective (computer science) PhD students who are accepted to multiple top-20 schools and have to decide where to go.

Should you choose where to do your PhD based on how well you jive with your advisor, or based on the school’s prestige?

Warning: talking about prestige, and social class in general, makes people feel uncomfortable. Choosing your school is an extremely rare, weird moment of social class mobility. It’s especially weird in times like visit days, where everyone has a current level of prestige and is eyeing those boundaries. Weirdness discussing class is a general phenomenon spanning cultures and centuries. I wrote about it anyway.

Advisor #

Choosing your advisor is very important. Many people will suggest you value the advisor match over the prestige of the school. I mostly followed this advice.

Assessing this match can be hard, because in the extreme case, you’ll only have fifteen minutes to talk to someone who you might be signing up to work with for six years. This was the especially case at the most prestigious schools. That situation itself factored into my decision. Do I actually want to work with someone who is so busy they can only afford fifteen minutes of time with such a potential commitment? Such a signal is indicative of a place’s culture: that students will work so hard, and be so self-sufficient, that they will make the research work, even if the advisor can’t devote much time to them.

Even though I didn’t fully conceptualize it, I think I intuitively knew I wanted someone more hands-on, who could spend a few hours a week working with me. I was switching fields (from robotics to NLP), so I had a lot to catch up on, and I wouldn’t be able to jump into a breakneck pace on my own.

In my particular case, this worked out. My advisor (Yejin) was able to meet with me for multiple hours per week when I started out, going over really low-level fundamentals about statistics, linguistics, and graphical models. Rather than publishing a paper every year, or even every quarter, I took two years to publish my first one, with enormous help from my advisor. Over the next few years, as she gradually got busier, I also needed less help.

So that’s time, which was an important factor for me. Others include:

• Jr vs Sr — correlates with other factors, like hands-on-ness, publication “hunger,” availability of career placement track record, presence.

• Research topics — what’s their academic brand? Where do they fall on the spectrum from “always publishes on-brand” to “willing to publish on any topic a student likes?” (If always on-brand, you’d better be able to learn to love that topic.)

• Publication venues — are they publishing where you want to publish? (Roughly: are they exclusively at top-tier conferences, if you care about that?)

• Career placements — where do their students end up? This one is tied up with a school’s prestige, too. I think advisors vary in how directly they optimize your career for your goals.

• Collaborations — do their papers have 2 or 20 people on them? Is there a lab hierarchy of senior students and postdocs? With big enough labs, this is almost a requirement purely for the sake of the advisor’s time.

• Funding — how much must their students teach?

• Presence — are they actually around? How are they managing their 20% – 100% time at whatever company they’re also employed by? (Ask their students.)

It is hard to weigh these factors, especially soft ones (e.g., seniority) and career planning that you won’t care about for several years. I think you’ll likely have a couple factors you really care about. For me, I didn’t even know what I valued until I met all my potential advisors. The biggest factor of all for me ended up being an undefinable “click,” where with just a few people I felt we could really communicate well.

But we must now address the thousands of micro-dopamine hits you’ll get from saying a prestigious school’s name to strangers for the rest of your life.

School #

Prestige is real. The moment you tell an everyday person where you go / went to school, they will ascribe you an according level of social status. The same is true with the letters “PhD.”⁰¹

Prestige also correlates with things that matter during your PhD. If you’re in a thriving (prestigious) research group with multiple professors, you’ll have other advisors to potentially switch to.⁰² Prestige correlates with good funding, which means minor perks like free food and major ones like not having to teach to pay for yourself. And prestige gives you a leg up interacting with the outside world, like applying for internships and jobs.

Surprisingly, not only are there different layers of prestige, but the layers are audience-dependent. By visit days I had basically memorized the top-20 computer science schools the US World News Rankings,⁰³ in-order, including ties. I was yapping at some poor old professor at Cornell’s visit days. I don’t remember the context, but Carnegie Mellon’s (CMU’s) shared #1 computer science ranking came up, and he said,

“No matter what it’s ranked now by some source, CMU will never have the same prestige as MIT or Stanford.”

In other words, realize that prestige—like basically everything—is made up in our brains. It is more complex than a single list.

To demonstrate, here are some rough rankings based on who is ascribing them.

Every random person in the public, and probably your parents:

1. MIT, Stanford
2. (any other very famous name even if it’s actually irrelevant in your field, like Harvard)
3. (all other schools)

People in tech

1. MIT, Stanford, Berkeley
2. CMU
3. (all other schools they’ve vaguely heard of)
4. (all other schools)

People in computer science PhD programs:

1. MIT, Stanford, Berkeley
2. CMU
3. Cornell / UIUC
4. UW / UT Austin / Georgia Tech / Princeton
5. (all other schools they’ve vaguely heard of)
6. (all other schools)

Within your field (e.g., at a conference):

1. Doesn’t matter. They know about your advisor.⁰⁴

This means your whole world temporarily flips upside down. When you’re choosing a PhD, you’re probably telling your friends and family where you’re visiting, reading the micro social signals they send when they recognize a school and assign it high prestige. You’re weighing your upcoming social class (it’s very real), future diploma, etc. Then, after all this, you’re in the PhD grind for years, nobody around you cares about your school because they’re all going there as well, and nobody you meet at conferences cares because they just care about your advisor and your research. It takes popping out of the system entirely (e.g., meeting normal humans,⁰⁵ applying for jobs) to remember your social reputation.

Another small thing to remember is that, like with master’s degrees, it seems easy to hop to a higher-prestige school for your postdoc and brand yourself with their colors. So that’s an option if you’d like a name boost.

The Meta #

You’ll get advice from professors at high, medium, and low prestige schools whose conclusion matches their school’s reputation. Surprisingly, the advice from high-prestige and low-prestige schools sound similar.

• Profs at lower prestige schools will tell you the advisor match is the most important thing, and that you’ll do great research with them. They’ll generally devote the most time and effort to recruiting.

• Profs at medium prestige schools will tell you to balance prestige with advisor match.

• Profs at high prestige schools will basically tell you to go where you feel you’ll do your best research. On the surface, it’s strange because it sounds like the advice professors at low prestige schools would give. But there’s two things going on.

  1. So many people pick based on school prestige that these professors will always have a plentiful supply of extremely motivated students to work with them, so they don’t actually care if you go to their school.

  2. If you pick their school because you genuinely think that you’ll do your best research there, it works well for them, because if they’re going to work with you, they want you to do great research. If you picked only based on prestige but the environment was bad for you to do research, then everybody ends up losing.⁰⁶

One kind of meta-advice I heard on this front is that if you’re going to go into industry, you should care mostly about the school’s name. If you’re going into academia, you should care more about the advisor match. Realistically, I think both matter for academia. The rub is you might not know what you want to do next. I didn’t know my future plans for about five years, and even then I had many doubts. (I was always jealous of people who knew their whole PhD what they wanted to do next.)

Probably the best general advice I heard was from the head of our department:

“I know you have your spreadsheet, you have all the different factors and scores and weights, and you’re tweaking all the weights to make the result line up with what you want. At the end of the day, there’s no perfect choice, only tradeoffs. You have to go with your gut.” ⁰⁷