Over the last few years, I have worked with many junior students and we have accomplished a lot together. I want to document my mentorship philosophy and record a number of lessons for future students.
My objective has always been to be a scientist and not a manager. If I take on junior students, it is because I believe that their contributions lead to better science. My sense is that a junior student, when properly trained, not only amplifies my efforts but also enhances the rigor of the science. Over the years, I have become more than just a utilitarian. I know that being an educator can be personally rewarding and make the deep pursuit of research less lonely.
The conventional wisdom, to assign students to a non-critical part of the project, has never worked for me. I always ask my students to work on the most important part of my most important project. This forces them onto my critical path. When the requirement for correctness is high, I will carry out the exact same experiments as them and we can doubly verify our results. Since I am not the most detailed person, a strong student helps me identify my mistakes and we get to truth faster.
This is why I describe my mentorship style as apprenticeship. I work with my students side-by-side, and part of their education is to see me perform at my best. I take on students when I am first conceptualizing my project, so that they can experience the exploratory phases of research. Starting out, I plan a curriculum, and try to cover the fundamental bases to study our problem, which can include general ML, law, and inferential statistics. I will also join pair programming sessions, where I will sit down with them and run our experiments together. These sessions are a chance for me to identify their technical gaps and give them the tips that bring them up to speed.
As the project and student start to mature, I find that it is important to educate the students holistically. Beyond the low level basics and how to run the experiments, I start to convey the research ideas and perspectives that could be worthy of publication. For my research, these are often statistical perspectives i.e. could we understand our problem in this statistical formulation? This is also when I start to train them not in the sense of a lecturer, but in the sense of a coach: week to week, I try to issue challenges and identify tasks that are still difficult for the student. I assign tasks that I don’t yet believe they can do a good job on. This could be asking them to write a draft of the introduction to show my advisor, or reading and teaching me about a paper that was just released on Arxiv.
In my projects, there is always some point where I find myself no longer running the experiments. The student, with their youthful energy, becomes the one in charge of bringing in the results. I also see that they start to believe in the direction of the work more than me (partly because they need the publication more than me). This is the stage that convinces me junior students enhance the science. I have the privilege to focus full time on the writing and the scientific story. You may have a chance to experience the joy of complementarity, where students cover your weaknesses and amplify your efforts. I find it important, but also a bit annoying, to allow them to shine. To allow them to take ownership over their work and expertise. To allow them to question my ideas with their outsider perspective.
On the slower weeks after the paper is submitted, I discuss and support the student’s future plans. If a student wants to apply to PhD programs, the discussion is often about the student’s true research interests and vision for the next few years. This vision often becomes their research statement for grad school applications. I actively identify what the student’s own research style is and look for problems and opportunities that suit that style. Finally, I send them off, into the hands of another advisor that we trust. Most of the time, we never have a chance to collaborate again. But I always look back to our tacit cooperation as a special time. I look back at what we accomplished and wonder, how much harder would it be if I had to do it all alone?