A year in PhD


So, it’s been a bit more than a year since I took a flight to Montréal to start a PhD on mathematical optimization & game theory for smart grids.

After the rush of summer conferences, and my return to France for a part of my PhD at INRIA Lille, it is a good time to take a step back and think of what has been going on and what is coming next. I’ll also answer some common questions I had in a more thoughtful way than I can in a bar conversation/family dinner. Maybe this can also help other PhD students seeing we are not in the same boat, but they all still look alike.

TL;DR: a PhD is not studies in the sense you think, and it’s not a job either, these mental models will not help much.

So, when are you going to finish?

I don’t know, when are you going to finish your job? It doesn’t look like it’s been moving that much recently. Or when will this company you’re building be “finished”?

These questions are similar, really. A research subject is rarely isolated, don’t see this as emptying a 4-year bowl of soup. It’s more like picking berries: you grab one, then the next, which is close enough, oh and this one is nice, but a bit further, I’ll have to stretch my arm a bit more to reach it.

[1]

I had some interesting discussions in Montréal about when and how to know you should bring your PhD to a conclusion. And the answer should always be that it depends what your objectives are, if you want to include this last project in the PhD. So no, I don’t know when I will finish, because if every step was predictable in terms of duration and success, it would not be a PhD or even research, what I do know is that I don’t want to block interesting projects or leave only half-explored research trails because “3 years is plenty”.

It must feel weird, getting back to university

It does, but not how you imagine. I was previously at a startup for a while. What I was used to is a great autonomy in execution. What the PhD is about is adding self-determination of the objectives, expected results, and means. It does not mean I’m working alone while I was in a team before, it means the degree of ownership of successes and failures is much higher, try to picture the three following sentences in a conversation:

  1. “I was at a startup before, it failed and I moved on to XYZ.”
  2. “I started a PhD but didn’t get through, then moved on to XYZ.”
  3. “I built a company, it failed, now I’m working on XYZ.”

It depends on the relationship to failure the person has in front of you, but for those I know, (1) is just an external cause, while (2) and (3) are personal failures, that’s ownership.

The biggest conclusion I made roughly after 6 months in is that a single-founder startup is one of the closest mental models to keep during the PhD, which explains several things, like inability to explain exactly what you do to your family and friends, imposter syndrome or procrastination.

So you get paid enough to buy noodles?

Yes, I’m living quite well thanks, I can even afford fancying my noodles, but let’s dig deeper on the matter of €/$/£.

Disclaimer: my PhD is between applied maths & computer science, I know all majors are not that financially comfortable.

[2]

I also know it’s considered rude to talk about money in some cultures, including France, especially if you’re not complaining; so yes, I’ll be rude.

When I’m in France, I’m paid slightly less than some engineers with the same level of qualification. The difference is higher if I’m comparing to what I would have had on Data Science, applied maths and software development positions. The difference between what I would earn and the scholarship is higher in Canada. Still, like I said, I can live without watching my bank account towards the end of the month.

The biggest danger of getting money monthly is thinking of it as a salary, meaning you’re thinking of the PhD as a job, meaning you’re thinking of yourself as an employee. On the paper, the money I get in France is a salary from my research institute, but one should keep in mind this is only on paper, the danger is to get the wrong mindset: think of yourself as a single person carrying a project, not an employee.

People often argue that they have a research director, who is de facto their boss. I don’t think this is the case, directors choose project proposals and people to carry them out (the order of this choice varies). They choose to invest time, effort and money from their structure into this person+project pair, without dictating to the letter what the outcomes of the projects are. Their retribution for this exchange is a partial ownership in the outcomes of the project (publications, conferences, software). Sounds familiar? Yes I’m looking at the Wikipedia page on Venture Capital. Let’s dig deeper: thesis directors invest this time, money and energy in areas they are familiar with, they have worked in and/or have mentored other people on. This sounds like the VC firms’ investment theses. Read these two articles to see a more proper definition and examples of investment theses but I hope you’ll get my point: PhDs are not R&D employees and directors are not “bosses”. If you have friends familiar with how startups work, this should be fine to explain. If you’re talking to people who have been employees their whole life in traditional structures, I have not yet found a clear and simple way to explain the situation in a casual conversation, let me know if you have something.

So, back to being a student heh?

This image is much easier to correct. On the paper this is true, a PhD has a student status, even when working close with/in companies like in Germany, Denmark or France CIFRE theses. Some people will ask this genuinely because they still picture their undergraduate years and think you’re back to this. So, it’s true, I’m taking courses, mostly because I find them interesting and keeping learning things is a bit of a raison d’être. But I was also doing so while working, using online courses platform like Coursera or France Université Numérique, going to meetups, reading and writing blog posts. So the thing that changed is maybe the part of my time dedicated to learning which got higher than when employed.

Now about the second category of people asking this question, yeah those one. In general, the points discussed above are enough as an explanation, but if the tone is really about trolling, which can happen, reverse the question: “Yeah I’m a student, I’m learning stuff every day. Wait you’re not? Exactly the stagnation I don’t want in my life."

Is it mean? Maybe, but the point is not to “win” a conversation, it’s about shutting down the perpetration of imposter syndrome, own the uniqueness of your status, both the good and the bad bits. None should be made feel inferior because they didn’t take the conventional, safe and socially accepted path of the 9-17 workday.

Re-thinking my values & priorities

I’ve written about values two years ago in the context of a company trying to define who they are. This is a complex topic, about discovering and understanding what the group prioritizes collectively.

As the owner of your project and of your time and resources, it’s up to you to define what is important. This begins with what is measured for success. Coming to academia from another milieu, I was not used to the process of publishing in peer-reviewed journals. First position I’ll take and try to maintain: the measures of academic success are broken, or at least imperfect (reach out to future-me if I change my mind). One example of this is research software.

If you can, watch Mike Croucher‘s excellent talk at JuliaCon 2018 for more depth on the subject.

The consequence of ignoring software as a valuable result of research is pretty straight-forward: proprietary software all over. It’s changing in scientific computing, statistics and other fields with the rise of Julia, R, Python. My domain, mathematical optimization, is still behind with dominant solvers (the software doing the actual work) and lots of Algebraic Modeling Languages (the front-end to interact with the solver) being proprietary. The last part is changing, the first one is still tough. My point is that people behind software you’re using everyday contribute way more to the success of your PhD than this obscure paper you cite because the review committee asked you to. If your university is giving thousands in commercial licenses and millions in access to paper, maybe you should make them donate both time and money to the tools you’re using. Yes all of them.

I have been involved in the Julia community, especially within the JuliaGraphs and JuliaStats ecosystems, mostly because these are subjects I understand (at least a bit) and/or used. Key take-away:

Reporting issues you have and contributing to improve the documentation is as valuable as writing code.

So… it’s been a year

And I’m still learning (understand making mistakes, getting stuck, etc), one of the reasons I had to learn more is not coming from a background in maths, nor from a research-oriented degree.

Reach out any way you prefer, Twitter, email. Of course some things I’ve written are related to my situation, I’d be interested to know how it relates or not to yours.


Image sources:
[1] https://pxhere.com/en/photo/571187
[2] https://pixabay.com/en/noodles-thailand-food-thai-2693009/

Avatar
Mathieu Besançon
PhD candidate in mathematical optimization

My research interests include bilevel optimization, convex and mixed-integer convex problems, demand response and pricing.

Related