Monday, October 13, 2014

How to Write an Academic Research Statement

It's that time of year when junior researchers are preparing applications for academic positions. One of the largest uncertainties that many people have is how to properly write a research statement that is typically part of the application package. This post contains my thoughts on what a good Computer Science research statement should look like when applying to US and Canadian universities. Please keep mind, though, that everyone's research profile is different so what worked for me may not exactly work for you.

To be perfectly honest, the research statement is not the most important part of your application package -- the letters of recommendation are. Your letter writers are accomplished researchers in your field of study, and can place your work in context as well as compare you to other researchers (when they were at your current career stage). Whether or not a hiring committee seriously considers you for an onsite interview is largely a function of your recommendation letters and any other research reputation you've managed acquire while disseminating your work (**).

Nonetheless, the research statement is still important, especially once the hiring committee gets down to a short list and are basically trying to figure out which of the strong candidates seem like they would be the most interesting and impactful additions to the department.

For reference, here's my research statement when I was on the job market in Fall 2012. I want to emphasize a few points:

(1) As my history teacher Dr. Skinner would always say: "Pithy and Erudition!" In other words, keep it short and to the point. You have to optimize for the case when someone with very limited time is doing a quick read of your research statement. No convoluted sentences, and no long paragraphs. As a general rule, I'd say it's probably too long if it's more than 3 pages. Optimize your wording to be as concise as possible.

(2) Tell a Story. Academics like to get excited by the potential of new research directions -- after all that's why many of us chose to pursue this line of work. So make sure you have an overarching vision in mind. For me, I chose to talk about machine learning with humans in the loop as my central theme. During my onsite interviews, I was repeatedly asked to describe what my NSF CAREER proposal would look like. The purpose of the question is so that the interviewer can get a sense of my research vision. I quickly realized that I can just re-emphasize various aspects of my research statement as my answer. This also helps create a consistent image of who you are as a researcher.

(3) Don't Regurgitate Your CV. Your letter writers will do a far better job of describing your previous accomplishments than you will in your research statement. Trust in them to do that. Only describe your previous work to support the story you're trying to tell. For me, I used my previous work to demonstrate that machine learning with humans in the loop is both a broadly practical and an intellectually deep research area. But I kept it to a bare minimum -- previous work took up just under 1 page in my research statement. Your research statement is your one chance in your application package to describe your vision to the hiring committee. Don't waste it all on dwelling in the past.

(4) It's OK to Stretch the Truth a Little Bit. Because you're trying to keep the research statement concise, you can't accurately describe all the details of your previous work. For instance, when I described my prior work, I did not include all the caveats that necessarily come with any such research result. That is OK; everyone understands that your research results have caveats. People not in your area don't want to read a laundry list of assumptions and conditions that your result must be couched in. And people who are interested will read your actual research papers. You can explicitly highlight the more interesting limitations of your previous work when you talk about future research directions.

(5) Don't Bullshit Too Much. Of course, you must be somewhat speculative when you're laying out your research vision and describing future research directions. But make sure that your speculations are grounded in some kind of sound reasoning. The easiest way to do this is to demonstrate that you've already done some preliminary work in the future directions you want to pursue. For my research statement, I listed one piece of preliminary work that I've done for each future direction. This is also a nice way to incorporate the more interesting peripheral parts of your CV into your research vision.

(6) Get Lots of Feedback and Iterate. I had many great mentors and colleagues who contributed significantly in helping to sharpen the wording and focus of my research statement.

Again, not all of these points may work for everyone, and I'm sure there are plenty of other good tips that I didn't mention (examples here and here). But hopefully this was useful to some. Best of luck everyone!

(**) This is not to say that your actual accomplishments are not important. If there is no substance to your work, then your letter writers won't write you strong letters, and your research won't have garnered you much recognition and reputation. Having done substantial work is assumed by default in this post.


Anonymous said...

If a PhD student wants to transfer to another PhD program at another University, what should he/she do? In this case, it is not easy to get a good reference letter from current supervisor.

Emma Glour said...

Good Information.