3 Professional Regrets From My PhD

February 7, 2024

As I wrap up my PhD, I've been reflecting on what I wish I did differently. It's a long list, so in this post I focus on three professional things I regret. By professional, I mean specific things that would have helped me wring more value, career-wise, out of the PhD experience. But they also would have helped me stay more motivated, and probably happier, too.

I wish I gave more early-stage talks

During my six-and-a-half-year PhD, I gave three distinct talks on my own work to researchers. I gave each talk only after the work was accepted for publication at a conference. I never gave a talk on in-progress work to researchers. This was a mistake.

I made this mistake because I was insecure. I felt that until the research community had officially deemed my work valuable (by agreeing to put it in conference proceedings), no one else in the community would find it valuable. In my mind, my project had no value until it was published. Who wants to hear about an idea with no value?

Ironically, this insecurity simply reinforced the idea that my unpublished projects were worthless. Consider that research projects in my sub-field often don't get published for 3-4 years. So I worked on each project in a dark room for 3-4 years, wondering all the while if it was worthless. Only when I emerged, submitted the project to a conference, and got an acceptance, could I validate the last 3-4 years of my life. But sometimes that validation never came because my paper never got accepted. How, then, was I supposed to feel about the last 3-4 years of my life?

I realize now that I had it backwards. I waited for conferences to give me validation, to give me permission to discuss my ideas. But I could have given myself that permission. Because presenting work is like planting a stake in the ground and saying, "I did this. It's real. I have been doing research". Periodically emerging from the dark room to show people what you've been working on makes the dark room more bearable. Plus, presentations have tangible effects: discussions and questions. Discussions can lead to ideas, which can lead to collaborations, which can lead to more papers, which can lead to lifelong personal connections.

Furthermore, the insecurities about the quality of my work actually damaged the quality of my work. Not only did I miss out on valuable presentation practice, but I missed out on valuable feedback. Presenting would have given me signals about the good parts (to focus on) and bad parts (to shed) of my work. I could have iterated on my work more thoughtfully and frequently, potentially increasing the probability of acceptance. I'm writing this paragraph thinking "this is all pretty obvious". But that doesn't make it easy to do. It wasn't, for me.

I wish I submitted to lower-prestige conferences

I was completely fixated on the top conferences in my sub-field. I basically ignored lower-quality conferences until my work was repeatedly rejected from the big ones. Part of this attitude came from my peers and superiors, I think. Some researchers feel like if you aren't submitting to the top conferences, it's not worth submitting at all.

Again, I think this is somewhat backwards. Top-tier conference acceptances are important for career growth as an academic. But so is your self-esteem. Consider the following choice. During the first four years of your PhD, you can either (1) get two or three short papers accepted to a low/mid-tier conference in your second year, then focus on expanding upon those pieces of work for later submissions; or (2) get a full-length paper accepted to a top-tier conference in your fourth or fifth year (but no other papers up to that point). The second might be better for your CV, but the first is better for your motivation because you get frequent bursts of positive feedback. In the second, you have a long drought of positive feedback. Speaking from experience, it's hard to stay motivated during droughts.

I think advisors should ease their students into research. Upon arrival, the student should be placed on an existing project with a senior student or post-doc. They should be second, third, or higher author. Their first solo project should be a small paper targeting a low/mid-tier, niche conference. This approach builds up their confidence and allows them to travel, present, and make connections early on. I think they're more likely to stick around after that. It's more sustainable.

I wish I wrote production-grade research code

The code I wrote for research projects is garbage. Many computer scientists say that garbage research code is expected and by design. They believe that time spent making code production-ready is time not spent doing the tricky, more research-y parts.

But some researchers go the extra mile and make their code decent. They open source it, organize it well, and document it. They make fancy landing pages for their research project, and try to make it easy for other researchers (and non-researcher programmers) to use their code.

I wish I made more of an effort to make my code and projects more production-ready and marketable. First, it would have made it easier to talk about my projects with non-academics. It's hard applying to internships when all your research projects are in-progress and closed source until publication. Even after publication, if you want your work to reach a wider audience, you can't just have a research paper---most people won't read it. On the other hand, fancy landing pages, documentation, and demos go a long way. Much of the research community doesn't appreciate (and maybe even looks down on) these deliverables, but I find that mindset short-sighted.

Second, it would have made me happier. As I've already described, hinging the value of your entire existence on whether or not a paper is published is a hard life. I can't control whether my paper will be accepted. But I can control the quality and presentability of my code. I wish I had derived more pleasure from the craft of writing software, even if that software was just a proof-of-concept for a research idea.

Control and feedback

Grad school is full of things you cannot control. One of my mentors, Prof. Dave Levin, often reminds me that you cannot control whether your paper submissions are accepted, but you can control whether you create submissions you are proud of. You can control whether you write good code. You can also control your feedback cycles. You don't need to wait for top-tier conference acceptances.