On Reference Letters

I write a lot of reference letters. I have had a lot of reference letters written for me. Like a lot of people, I hate asking for reference letters. This webpage is intended to make it a little easier for people to ask (or not ask) for reference letters from me.

Academia runs on reference letters. You need them to get into grad school, and then to get your first academic job, and then again for most promotions and moves. No one likes asking for reference letters, so if you feel hesitant about asking for a reference letter, you are far from alone. Writing reference letters is part of the job for me, and anyone with an academic job has had lots and lots of letters written for them, and has an obligation to do the same for people who are just starting their career. Don't be afraid to bother people for a reference letter. That said, you don't just want a reference letter — you want a good reference letter.

In this webpage, I try to outline situations where I can write a good letter. There are definitely situations I did not cover here. When asking for a reference letter, it is useful to ask yourself: what can the person say about me that will make me look like a good candidate for grad school or an internship? (This is not an easy question to answer — it's easier to know what it takes to be a good grad student after you went to grad school.)

This webpage reflects my views only — other people will have other views and policies regarding reference letters. I am also teaching rather than research faculty. That means I do not supervise graduate students doing research, and so I have never decided to accept or not accept a graduate research student. (However, I have supervised an Applied Master's student and have served on the admission committee for UofT's MSc in Applied Computing program.)

Undergraduate students who need a reference for a first job

If I can figure out who you are and you did a reasonable (or outstanding) job in my class, I'm happy to say that you did a reasonable (or outstanding) job in my class.

Undergraduate students who took my class and are applying for competitive internships

Examples of letters I've written include letters for students applying for summer research internships at UofT and at places like MIT and Harvard. An OK letter for an internship like that would say that you are a student who performed very well in my class and in the program in general. A good letter for an internship like that says something interesting about you. Examples of things I said about students are: the student improved my class by asking interesting questions; the student came up with an original approach to some problem we were discussing in class; the student came up with an original approach to a problem on the exam; the student impressed me with their drive and ability to solve problems in office hours; the student's overall performance was in the top 10% of the class.

An OK letter might do, and I will be happy to write it. However, you should think about who is the prof who is in the best position to write a good letter for you.

Students applying to research-based graduate programs in Computer Science and Statistics

First, a caveat: while people do seem to be give my opinions of students some weight, a letter from me is not going to be as influential as a letter from a tenure-stream faculty member, all things being equal.

A good letter says I think you will be successful as a graduate student researcher, presents concrete reasons for why I think that, and contains specific examples of your work or your interactions with me that make me think that you will be successful. (This applies to any good university. Stronger letters are needed for more competitive universities.)

All this should be easy enough if we successfully worked on a project together. The best thing you can do to get admitted to a graduate research program (and to figure out whether you want to go to grad school) is to work on a research project with a faculty member.

I have written what I think were good letters for students who have not worked on projects with me. Examples include students who showed me extra work they did for my class that evinced superior technical skills, imagination, and curiosity about data analysis or machine learning and students whose performance in my classes was outstanding overall and who I could see were taking a lot of challenging courses.

I have written OK reference letters for students as well. Unfortunately, a letter that simply states that (for example) a student got a A+ in my course is unlikely to be very helpful for gaining admission to a good graduate research program.

It helps if I can write about, for example, a bonus problem that you completed. By itself, it probably would not be all the helpful (unless you did a really great job and showed a lot of originality), but perhaps it might help.

Most graduate schools will ask me to say whether the student is in the top 1%, top 2%, top 5%, top 10%, top 20%, or top 50% students in terms of things like originality, academic potential, maturity, writing, speaking, etc. If I don't have better information, I have to rely on your work in my courses, which means that I cannot give really top scores to people who haven't outperformed nearly everyone in class. I suspect that even top-10% scores are not that useful for admission to most good graduate schools (though they won't hurt). This does not mean that you shouldn't write to me to ask for a letter unless you got a 100 in my class. If I remember interactions with you that indicated that you would be a good graduate student, I will be happy to write about them.

If we haven't interacted much when you took my class

If you did good work for my class, I generally want to help you. Almost always, before agreeing to write a reference letter, I will meet with the student for 30-60 minutes to discuss their work, studies, and goals, and to form an impression of whether they seem like a potentially good graduate student. In my letter, I will have to mention that my impression is based on at most a hour-long conversation and looking at the student's coursework. This is not ideal, but I believe those kinds of reference letters can be useful as well under some circumstances.

Students applying to course/internship-based graduate programs in Computer Science and Statistics

A OK letter says that I think you will be successful in the program. I will try to determine that based on your transcript and my familiarity with UofT and your performance in my course. In most situations, an OK letter won't help — especially if you are applying to UofT, since UofT admission committees already know all about the difficulty of the courses you took. In some situations, an OK letter might help.

A good letter will say that I think you will excel in the program. I will try to determine that based on your work in my course. I try to assign projects that approximate the kind of work that an advanced student might do in industry, so the quality of your work in the projects is important (not just the grades — I try to look at your actual work, and figure out if you did the kind of work that indicates that you're prepared for advanced coursework and high-quality industry work).

Students applying to professional programs outside of CS/Stats

I know less about admission to those programs than about admission to programs in CS/Stats. I still occasionally write letters for those programs, but it is harder to determine what makes for a good letter.

Former/current TAs

I usually have something to write about TAs. Send me an email.

If you'd like to ask me for a reference letter

Please send me an email with

Valid HTML 4.01 Transitional