This is the application season—many applicants are polishing their essays, personalizing them for individual schools, reminding their letter-writers to avoid the (inevitable?) delay in their submissions, etc. Across multiple discussions, we have noticed that many aspirants are hesitant to apply broadly, and would restrict their Ph.D. applications to “top-4” or “top-10” schools in the world. We often nudge them to explain their rationale for not applying to schools that are not as favorably ranked. Many of them do not have a specific reason (it is just something they never considered seriously). Others arrive at a mix of reasons indicating top schools imply better opportunities, a richer environment, stronger peer group, etc. In short, if they want to do a Ph.D, they want to do it at the best university in the world. Otherwise, why bother?
We do not subscribe to this argument, and rather see Ph.D. as a training in research, akin to fitness training in a gym. You don’t undermine the utility of exercising just because the top-rated gym in your city is oversubscribed. In this post, we would argue that many schools (probably about 100, certainly more than 50) can offer an excellent Ph.D. experience, where you could achieve solid training and launch your career. The crux of our argument is based on the realization that the Ph.D. experience (and success consequently) relies on the following:
- Your advisor(s),
- Everything else.
You are the driver of your Ph.D.
A lot of emphasis of any (successful) PhD program is to ensure that their graduates are independent. The end product of a Ph.D. is not a thesis, or research papers, it is you—a trained researcher ready to make a dent into unchartered territories. The environment around you is set up such that independence is inculcated from the onset. It is on you to identify interesting questions, devise solutions, validate your ideas and effectively communicate them. You meet your advisor, typically once a week, to discuss research but you have to fend for yourself throughout the week. One of the biggest deciding factors clearly then becomes your potential and your outlook to make the best of your Ph.D. journey.
One might point out that many successful people are from top universities. Such correlations, while irrefutable, do not imply a cause immediately and clearly overlook an important confounder—student potential. The top universities in the world have their pick from the application pool, therefore it is not easy to tease apart the role of the institution from the students’ potential in their success. Estimating the role of institutions (by controlling for student potential) is impossible from the data available alone and requires interventions that would force top students to forgo their MIT offers and instead attend lower ranked universities and vice-versa.
Academic jobs are scarce and only 25% of all the institutions in the U.S. produce 71 to 86% of all tenure-track faculty. However, we cannot naively attribute this trend to the role of the top institutions (without accounting for the potential of their students). For a research career in industry in machine learning and allied fields, the supply of students is only catching up the large demand, and hence, the school you attend might not hinder your chances.
Your advisor is your guide.
After yourself, the person who impacts your Ph.D. experience the most is your advisor. While different people look for different things in their advisors, it is beneficial if your advisor is an active researcher in areas that excite you. A noteworthy fact that students overlook is that the U.S. has a thick tail of such active researchers across universities. For instance UNC Chapel-Hill, which is tied at 32th position in the Artificial Intelligence CS Rankings in the US, houses several influential faculty in NLP. The thick tail in U.S. schools is a manifestation of the fact that academic jobs are scarce and therefore many top graduates (from all over the world) join lower-ranked schools. Two (among many) important factors that make someone a good advisor for you are:
Mutual interest in the specific subfield: Ph.D. is about expertise in a narrow specific area, say interpretability of deep learning models. The best advisors in particular fields might not be in the best ranked school. For instance, one of the leading figures in interpretability of deep learning models, Prof. Sameer Singh, is a faculty at UC Irvine which is ranked 31 as per the above list.
Operational style: Advisors come in all shapes and sizes. Some discuss only broad high-level ideas, while others take a more hands-on approach and dive deep into low-level details. For some writing and presentation are of utmost importance and for some the primary focus is on the empirical evidence backing the claims. A few are personable with their students and some maintain a strict personal-professional boundary. Note that in some respects you need alignment with your advisor and in others you need an opposing force. For instance, if you sweat the small stuff, you might benefit from a big-picture advisor. If you need small wins to stay motivated, an advisor who focuses on long-term goals might not be the best for you. To minimize frustration and maximize the quality of experience (for both parties), you need to reflect on your style of working and the kind of advice you would benefit from.
The university rankings take none of the above in account. We suggest, as a rule of thumb, that you should attend a school that has at least two potential advisors—it gives flexibility to change advisors if things don’t work out. Bigger schools are better bets in this regard.
Everything else is tertiary.
Whether your personality type is suited to the size, prestige, culture and location of a school is often overlooked while deciding on a place where you will spend a significant chunk of your life.
A student just about average in a large big-name school might be academically akin to a top student at a lower-ranked school, but their relative standings might affect their self-perception. While some people thrive in a competitive environment, others are prone to imposter syndrome on seeing abundance of high performing students around them. If you belong to the latter group, you might feel confident and work better when you are amongst the top few. On the same lines, if you are prone to seasonal depression, doing a Ph.D. in the University of Minnesota might not be a great idea.
Finally, a PhD is a deeply personal pursuit. No two theses solve the exact same problem. Citations, number of papers, school affiliations are only a soft measure of success, and should be treated as such. Focus on what works best for you!
 Systematic inequality and hierarchy in faculty hiring networks; by Aaron Clauset, Samuel Arbesman, Daniel B. Larremore, Science Advances, 12 FEB 2015 : E1400005