Tara Safavi data mining researcher

Why aren't CS PhD students trained in research writing or presenting?

21 April 2018
Tags: grad-school  opinion  phd 

ECSEL recently hosted an assistant professor of computer science at Michigan as part of our monthly luncheon series with professional women in computing. One of the most memorable things she shared was that new faculty aren’t trained before starting the job. You’re pretty much thrown into the position, she said. This surprised me because Michigan faculty impact literally thousands of students through teaching and advising. To confirm, I asked several professors and lecturers whether they received training when they were hired by Michigan. No, but I wish I had was the unanimous answer.

I started to think about this when I sensed a similar dilemma among the PhD students. Aside from the college-wide Responsible Conduct of Research and Scholarship (RCRS) requirements, which mostly consist of free food and someone talking at you for two hours, we receive no standardized training beyond the variable and decentralized offerings of our advisers and mentors. To make things worse, grad school is a recipe for isolation. With deadlines competing against work that’s “never really done”, it’s easy to withdraw into bad academic and personal habits. You can go a long time without (in)formal guidance from colleagues.

To keep the conversation going, I asked a professor why CS PhD students aren’t required to take a research paper and proposal writing course. For reference, engineering undergrads in computer science are required to take technical communication, and liberal arts undergrads in computer science are required to take an upper-level writing course. Research writing — in other words, rigorously but clearly communicating the technical details of work no one’s ever seen before while telling a cohesive and compelling story — is hard! The professor’s answer: I agree with you that we should have one, but…

  1. Some faculty would argue that this requirement would take valuable research time away from students.
  2. No one is passionate about teaching such topics.

My response:

  1. A one-credit research writing seminar wouldn’t realistically take much time. Certainly not as much as an assignment on extending the LLVM compiler! Moreover, it’s a one-time thing. Such costs amortize. Better student writing might even lead to higher publishing rates, which every research university wants.
  2. Are faculty really passionate about all the courses they’re required to teach? My five years of experience at Michigan say no. I also think most people can become passionate about topics with which they have personal experience. Not everyone gets excited about speculative hoisting (sorry, compilers people), but we all can remember a great (or terrible) research paper we read and what made it strong (or weak). A recent discussion I led with my classmates on bad research practices like p-hacking suggested as much.

I conclude by emphasizing that computing research is unique and fairly young. We don’t necessarily do well-established “research” things like conducting field work or growing cells or surveying people. And those of us that do via interdisciplinary work face further challenges like following standards across multiple domains. We also publish in different venues. The rest of the research world aims for journals, but we emphasize conferences and even open publishing via Arxiv, itself a contentious topic among CS researchers. My point is that college-wide training like RCRS doesn’t target computer scientists and I believe we need training that does. Until then we’re fending for ourselves.

comments powered by Disqus