I officially started my PhD 3 months ago, and while my current results make me feel like I only started yesterday, I have already gone through quite a bit, and learned a few things about what it means to do Science and to conduct a PhD. Below are a few disconnected thoughts based on my experience — I’m aware that these reflections won’t generalize to everyone. I n fact, it might not even generalize to my future self — but I decided to put them in writing anyway — a Memorabilia for my future self.
During a PhD, figuring out what problem to work on is more challenging and critical than solving the problem.
In the business world, the loss function is generally
There is no absolute way to evaluate what good research is. Is it research that gets cited a lot? Is it research that remains influential decades after its publication? Is it research that gets published in top journals? Is it research that manages to solve very difficult problems? Is it research that manages to solve problems that many people care about? While all of these can somewhat be considered metrics of research quality, they often are not adjacent. Highly cited and top journal papers are not necessarily works that significantly influence a field. Solving very difficult problems is not necessarily something that a lot of people will care about, nor is it necessarily something that people will care a lot about.
Since we cannot find an absolute for what good research is, I feel it’s best to assess it relative to the person conducting it: doing good research is doing the kind of research you look up to. For some, it’s a technical and precise exploration of niche intricacies, for others, it’s information integration and rethinking the direction of a field, for others, it might be real-world-oriented technical breakthroughs.
Our role as researchers is to produce research that aligns with what we look up to. And, not research just for the sake of publishing. This applies to every creative activity I guess - one should seek to produce the kind of work that they can be proud of and that they think other peers might look up to as well.
Now this is hard. Not only do you have to be incredibly performant and perseverant to achieve this
A concrete example that applies to me is to decide on whether to pursue publishing in NeurIPS style conferences, where incremental performance on benchmarks is the general pursuit, vs work that gets traditionally published in journals - slower, more tedious, and more risky too. Both are good in their own way - but one should consciously position themselves.
Successfully navigating this conflict takes a lot of discipline. One key aspect of this discipline is learning how to fight the right battles: the ones we’ll enjoy fighting and actually have a chance at winning.
Academia can be an incredibly diverse and exciting universe, where ideas in all shapes and colors flow from everywhere.
Sometimes you’ll read a paper and think “huh, this could be cool to try”. Sometimes, your supervisor comes to you with their new idea, and you know it would be a good thing for your relationship if you tried it. Sometimes (my personal favourite) you have a conversation with a scientist from a different field in a bar and see some parallels between both of your research and think that this would make a sexy and unconventional collaboration.
It is undeniably a privilege to evolve in environments where this is the norm, but it comes with a price: the burden of choice, and the challenge of focused attention.
The more ideas are around you and the more people are around you, the more difficult it becomes to focus your attention and have a clear sense of direction — especially for the curious and playful spirits — which I’m lucky to be. But I remember the famous Steve Jobs quote from his Stanford commencement address: “you can’t connect the dots looking forward; you can only connect them looking backward”
What’s due diligence for a PhD student?
It’s a continuous dance between 4 core activities:
All of them are equally important and mutually beneficial. One can think of this like a control problem — where optimal state (successful research and timely graduation) can only be reached if these four are well balanced during the span of your PhD.
The last part is probably the most often overlooked. A PhD can feel like a tight corridor that ones tries to escape - especially towards the middle of it (I’ve been told). It’s important to feel like one has the capacity to change their way of doing and their direction - the Sunk-Cost Fallacy can be brutal in Academia…
During my first three months, I made a different mistake — I did very little 1) and only did 2), 3) and 4). Bad too — obviously. But I blame the beginning for this mistake — the future will be different, especially since I now have a bit more momentum on different projects and can prototype quickly.
Choose wisely.
Don’t do too much. Don’t do too much. Did I say you shouldn’t do too much?
Supervising students, even if they are incredibly well chosen, is no substitute for doing science yourself. It is neither sufficient nor even satisfying to only give them a project and let them be. I already have three students under my wing and feel a bit overwhelmed by having to manage them — I simply do not have the time to have a detailed look at the details of their science.
This directly results in the well-known shortcoming of most PIs in the world out there: losing touch with the daily bread of research. I am first hand realizing what this means — it’s a weird thing to just let go of the daily intricacies of experiments, and trust someone else to properly do the details themselves. You eventually lose control over things, and also lose some level of understanding about what is being done and how it is achieved. Of course, in my case, where my students use technologies and experiment ideas that I have played with quite a bit, the issue is less problematic than PIs not having done lab work/coding in years — but one still feels the dangers of it and the struggle of giving up control, and its consequence on one’s understanding of the parts.
So why do people delegate? 1. Because they judge that their time is better spent doing different things, and 2. because they can achieve more in less time. Do I feel like my time is better spent thinking about research direction than doing research itself — I’m not sure. Do I feel like I can achieve more by having other students around me conducting research that I think should be done? I think so, but it hasn’t happened yet. This is where having good students come into play — when they get up to speed quickly and are actually capable of doing more and better than you could have done. This is rare. Most often, one typically spends weeks accompanying a student to set up an experiment that would take the supervisor a day to complete. However, I believe one must do the effort — if you get lucky once, the outcome surpasses the multiple failures.
Now supervision has an additional layer compared to mere “delegation” — teaching. This is one part I value a lot. I feel a responsibility towards the students I supervise. I feel like I must, to the best of my abilities, show them the way, explain them things the best I could, and most importantly, inspire them to do good Science. This is a humbling experience, that takes human discipline, but also scientific rigor and knowledge. While I know that a lot of student project I’ll oversee won’t lead to ground-breaking research, I know I have the opportunity to positively influence one’s scientific career and attitude to work — and that is a privilege and responsibility I’m happy to carry.
Oh, before I forget - Choose your students wisely. And don’t get too many. Really, don’t get too many. Did I say you shouldn’t get too many?
There are a number of interesting problems that one experiences while mentoring students. A couple of weeks ago, I realized that I had sent a student in the wrong direction (dead-end method). No big deal, “only” a few days of work were lost, but I should have done my homework better. In the moment I realized, I was torn between apologizing or slightly downplaying the time-loss into a “it’s fine, we learnt about X and Y in the process”. I picked a mix of both.
It’s also a difficult to challenge to find the best way to motivate students. Should one be very demanding, or should one let them do their thing and only put as much as they are willing to put. After all, we do not pay students, and the cost of a badly performing students is “only” some time lost. I so far have chosen to let them be, and be as here as possible whenever they need me — let’s see if the results follow.
To get through a PhD, you need to be in a stimulating lab, with stimulating people around you — that push you, towards whom you feel some kind of accountability. There is no way you get through this sane without it — no way.
I feel it — some days I feel a push because of a good conversation, or because someone I look up to asks me a difficult question or expects some good work from me. I also have days where I don’t feel that at all — and these days, not only am I not motivated to get anything done, but my work also feels significantly less meaningful.
Non-PhD/Academic friends often ask me (and other PhDs) how do PhDs navigate the freedom of working on what they want, and the freedom of not showing up to work on a random Tuesday. Truth is, it completely fucks us up. There is this impression that a single day of work doesn’t matter - and it’s completely true: a single day of work doesn’t matter in the greater scheme of your PhD
This makes the PhD experience very close to “creative” kind of careers, rather than “corporate” ones. Does it matter if you don’t compose that song? Does it matter if you don’t write that essay? Of course not - no one will know that you haven’t done it, and no one will care nor ask for it. But that song that you didn’t compose, that essay that you didn’t write, that side-project that you never got to try out, that paper that you never took off your reading list - it could have been the start of your best work. It could have been the start of a work that actually is the 10x kind of achievement that we are after in life.
And why did you not do it? Because you didn’t show up on that random Tuesday - thus not putting time on anything else than strictly urgent work.
Doing a PhD requires a special kind of discipline and organization. - to get you to show up every day, and master the exploration vs exploitation tradeoff. Every student who doesn’t manage to do so ends up depressed, and I think that even students who kind of manage to do so also end up depressed. Caring about your work makes life more difficult I guess
One can also think of this as a delayed gratification problem. The work you do today has very minimal influence on the outcome of the end of the day. You’ll probably keep getting paid even if not a single gram of good work comes out of your mind for 6 months during your PhD
“Most people overestimate what they can achieve in a year and underestimate what they can achieve in ten years.”
Context: Due to various circumstances, I happen to be involved in a project where I have been struggling to contribute. The work is most often far-away from my expertise and current scientific focus, and I unfortunately do not have the time to catchup enough to manage contributing as I’d like to.
After one project meeting where I had been particularly useless (in the presence of a top ETH professor), one of the other PhD students present in the meeting messaged a few minutes after the call ended to tell me that I shouldn’t feel left out, and that I should keep showing up, despite not managing to contribute yet. He told me about one project which he also got fortuitously involved in, and where he had no expertise either. But in his words, “I was part of the discussion and asked the stupid questions that only people without experience would ask, and that ended up being an actually helpful contribution.” The project has now been written up (with him as co-author) and sent to NeurIPs - not bad for a project where you have no experience!
I found his message incredibly inspiring because I never saw it through this light - instead of asking the useful stupid questions, we always feel self-conscious about not sounding stupid and choose to shut up. Things don’t have to be this way.
Even more so, I found his message inspiring because he perceived the struggle and embarassment I was going through, and was kind enough to take the initiative of encouraging me - just because. I’m eternally thankful for that - I really needed it at that moment.
I’ll try to keep these lessons with me: ask the stupid questions, and most importantly, show kindness to people who are struggling.
There is no reasonable justification to pursue a PhD program, except to enjoy the daily bread of one’s research. You won’t make the world a better place by getting your PhD. You won’t save lives through your PhD research (if you think that’s the case, think again). Oh, and you won’t earn money by doing your PhD either. In the best-case scenario, you make your grandmother a bit more proud of you than she already was. So why are you here, if it’s not to enjoy the process?
Am I enjoying the day-to-day so far? Yes. Is it a bit painful? Also yes, but in a good way, the kind of painful that makes you grow as a scientist and human being.
I truly feel blessed to be part of the Institute of Neuroinformatics, and that I chose to pursue this PhD instead of going to industry. I love science, and I love being surrounded by people who care about producing interesting work, and discuss ideas constantly. I love that ideas are discussed, rather than money or means to money.
The road will, however, be bumpy. Looking ahead, I must do a better job at focusing my attention on the things that matter and getting things done - if not for the sake of my PhD, rather for the sake of my enjoyment of it. Well actually, aren’t both equivalent two sides to the same coin?