Skip to content

Latest commit

 

History

History
71 lines (52 loc) · 6.7 KB

what-working-with-me-is-like.md

File metadata and controls

71 lines (52 loc) · 6.7 KB

What working with me is like

tl;dr - see Juggling Analogy.

  • Unlike most advisors, I don't have regular, fixed meetings with students. The rationale is that progress doesn't happen at an even pace. Sometimes you're churning out experimental results every two days and need my help interpreting them and formulating next steps—making you wait until our next weekly meeting reduces your productivity. Sometimes you're in the middle of a complex implementation that may take a couple of weeks—you'll have nothing to tell me other than "I'm working on it".

  • Instead, I try to keep as much of my schedule as open as possible and give you individualized attention. I might spend an entire afternoon brainstorming an idea with you, but then not meet again for three weeks while you're implementing the idea. Toward the end of your Ph.D. career, I might say, "Go write your thesis!" and not meet with you for a month until I see drafts.

  • I am a heavy Slack user and use it as my primary channel of communication with students. My favorite thing about Slack is that it fills in the "little crevices in time" that allows me to effectively multi-task. For example, when I'm waiting in line at the checkout, I can use the 30 seconds productively on a quick issue.

  • I don't particularly care where you work, as long as you're productive.

  • I don't particularly care when you work, as long as you're productive. Want to work on Hawaii time (which is six hours behind Waterloo)? Fine with me. (For some time, a student actually did this.) The only caveat is that you might want to have some overlap with my schedule for meetings, discussions, etc.

We asked our employees where they want to work:
7% want to go to the office full time
31% want an office-remote hybrid
62% want to work only from home

So I told them: sounds great. Do whatever you want. As a CEO, what do I care? If you get your work done, that's all that matters

— Dan Price (@DanPriceSeattle) March 15, 2022

How much should I work?

For reference, I estimate that I work around 40–50 hours a week. I try to get 7–8 hours of sleep every night. See my writeup on what my personal schedule is like.

Additional useful links:

What should I be working on?

In terms of research, I expect you to tackle items in one of two categories:

  • Research problems that I've formulated or novel approaches to existing problems that I've devised. In some cases, these "ready-made" research ideas are contributed by a post-doc or a senior student. These items can be found in our group's internal research board, which contains a holding area for "unclaimed ideas". They vary from vague observations ("this is interesting, we should dig into it a bit more...") to concrete suggestions ("I think we can improve this model with a small tweak...") and span lots of different sub-areas of NLP, IR, etc. I generally have good intuition about what's interesting, and these ideas are usually good starting points that will lead to something publishable. In formulating these research ideas, I also try to build in a "fail fast" property so that you can do some initial explorations to make sure it's not a dead end.
  • Your own ideas, e.g., a problem you've formulated yourself or a novel approach to an existing problem that you came up with.

The tl;dr is that you should be making demonstrable progress executing "my" idea or "your" idea. I put these in quotes because it's an over-simplification: obviously, you'll contribute to "my" ideas and I'll certainly help you refine "your" ideas. Since the bulk of the research conducted by my group is empirical (i.e., experimental) in nature, by demonstrable progress I typically mean new experimental results (or concrete steps towards generating them). Of course, I don't expect all experiments to actually "work", but "failed" experiments also offer valuable insight into the problem.

The "anti-pattern" I don't want to see is you struggling to come up with something to work on for extended periods of time. To put it somewhat bluntly: either come up with a compelling research vision of your own (and I'll say, great, go ahead) or competently execute a research vision that I've laid out.

Generally, I will nudge newer students to "my ideas" to give them concrete things to work on and leave more senior students "to find their own way". A few common "advising patterns" I frequently adopt:

  • For particularly enthusiastic new students, I sometimes give them "a lot of rope" to come up with their own ideas. Some do succeed, but to be honest, the success rate isn't very high. New students have a lot of energy, but lack the experience of knowing what's interesting or doable.
  • In the above case, after I let the student "spin their wheels" without getting anywhere for a while (for example, the first semester), I'll start nudging them towards more concrete ideas... the nudging becomes more forceful the longer I don't see demonstrable progress.
  • For Ph.D. students, I actually take the opposite approach: I will let students "wander in the wilderness" for a while to develop their own ideas, because that's an important part of the Ph.D. experience.

I recently learned (Feb 2022) that Austin Henley has a much more pithy way of saying this: "lead or be led". His lesson of "managers as input/output machines" is also quite complementary and consistent with what I've written here.