There are thousands of articles (online and in print) on advice for people in academia, especially for graduate students. Unfortunately, I find the vast majority of them kind of unsatisfactory: they are mostly conventional answers to conventional questions. For example, regarding
"How to come up with a research topic"
The most common answers are:
Read the "future directions" section of a paper
Ask your advisor
Discuss with other people
But everyone knows these, so what's the point of repeating?
My own approach to these questions drastically deviates from the mainstream. It certainly works for me, but I am not sure whether it will work for others — I suspect that you will find it useful only if you have a specific personality/ideology. Still, I figured it would be fun to write it down and see whether it will be helpful to some readers.
A1: Avoid those "winning-brained" people, namely those whose entire meaning of life is to "win" and be superior to others. Indicators include:
Listing all kinds of awards/fancy achievements at the beginning of their homepage (having a list in their CV is fine)
Stating that their undergraduate GPA was 4.0 or 3.9 in their CV (a 2.2 GPA in their CV is fine — in fact I would consider that a plus)
You can (somehow) infer that they don't have any non-academic hobbies
These people make the worst advisors, as they will induce a tremendous amount of anxiety in your brain, which will haunt you even after you graduate.
A2: Most soft skills in academia can be practiced in non-academic settings. This is obvious to me but somehow I haven't seen anyone else say it.
Imagine you are in a fight with someone on Reddit: You believe that we are on track to pass the 1.5 degree Celsius threshold of global warming, while they think things are not that bad. You only want to spend 20 minutes on this fight: within this amount of time, you need to find 2 or 3 scientific articles on this topic, summarize them, organize them into a coherent and convincing argument, and hopefully earn 50 upvotes. Now if you think about this whole process... it is very similar to writing the introduction and technical overview of a paper!
I always think
Being a good keyboard warrior ≈ being a good paper writer
But it doesn't have to be some Reddit drama: You can practice your skills everywhere. For example, when I walk down the street and see a building that looks old, I sometimes wonder when exactly it was built. After going home, I'd try to look up some databases and figure that out. Sometimes it's easy to find; sometimes you need to dig into some obscure stuff. This is basically the process of doing a literature review.
As an exercise, you could try to describe some complicated and boring governmental procedures (e.g., obtaining a visa) on Reddit or Quora. This is great training in academic writing.
A3: The key to time management is to reduce the amount of busywork. I first observed in China that people do a huge amount of busywork, but this is the case in the U.S. as well.
There are two ways to avoid busywork. First, in many cases doing the bare minimum suffices, and putting too much effort into something could be unnecessary or even negative. For example:
Many students pay close attention to a talk, even if they cannot understand 90% of it (and the content of the talk is not super relevant to their own area).
While preparing for a talk, many students have super fancy/complicated slides, in which they cram too many definitions/formulas/proofs. A good talk should highlight the intuitive ideas, not the technical details.
While reading a paper, many students try to understand it word by word. Sometimes this is necessary, but most of the time you only need to gain a high-level understanding of the main contribution of the paper — which should cost you 30 minutes at most. (Also see A9 below.)
Second, it is generally NOT the case that your learning outcomes improve as long as you put more time in it. Rather, it is crucial to have a thorough understanding of stuff and know what you are talking about — which takes only a little time. The best example I can think of is security proofs using the (Bellare-Neven) forking lemma: It involves a very specific process, namely
Building a wrapper W from the adversary A, where W accepts or not based on whether A wins its game.
Building a forking algorithm F from the wrapper W, where F outputs some specific information by running W twice. The forking lemma establishes the relationship between Pr[F outputs something useful] and Pr[W accepts].
Building a reduction R from the forking algorithm F, where R solves some hard problem if F outputs something useful.
You can understand all of the above by going over just one concrete example, namely how forking works for the Schnorr signature scheme. However, I've seen many people trying to write a "forking lemma proof" with only a vague idea of "the reduction should run the adversary twice", and their "proofs" don't even specify the A → W → F → R chain! This is doomed to be busywork.
A4: Burnout mainly comes from two sources:
There are a lot of stuff that you are not interested in but have to do, and
You constantly face failure/rejection on stuff that you are interested in.
For (2), see A7 below. This answer will focus on (1).
The key to handling stuff that you don't like (but have to work on) is to treat them as a LARP game. For example, perhaps you feel you are a theorist down to your heart, but you frequently have to sell your work as applied. Well, then you just LARP as a practitioner: You are never actually a practitioner; rather, you are merely playing a game in which you simulate the words and actions of a practitioner. And once you switch to this mindset, the LARP game is actually fun: It's very interesting to see other people's reactions to your simulation, and you need to adjust your simulation strategies based on them. I find this much more interesting than playing video games.
A5: I am against the common approaches to finding role models, as they glorify people who are conventionally successful. This is problematic for multiple reasons:
Paying too much attention to a single achievement ignores the broader context which enables this achievement.
In a capitalist society, conventionally successful people tend to be bad people, since the primary way to succeed is to exploit others.
Focusing on other people's success — or your own success — will make you "winning-brained" (see A1 above).
Rather, true role models are those who stand up to the establishment and defend their principles, potentially at the cost of their conventional success. Some random examples include:
Alexander Grothendieck
Grigori Perelman
The teens who organized the Luddite Club in NYC
At the end of the day, what keeps your peace of mind is not those conventional achievements, but rather the fact that you've never been a sellout.
A6: Most types of success are not worth celebrating. If you made some sort of conventional achievements, that's most likely because you are lucky/privileged.
I was a pretty successful student in China (which means that I ranked well in the college entrance exam). This is not because I was particularly talented or something; rather, the most important factors are
I grew up in a family that values education
I am from Beijing, which arguably has the best educational resources in China
I was a boy, and boys are generally encouraged to study STEM compared to girls
By and large, the society is very unfair, in that people who are already privileged tend to succeed much more easily. The focus should be on improving this unfair system (or at least pointing out that the system itself is unfair), rather than celebrating individual success. It's not an honor to win a rigged game.
A7: A6 pretty much already answers this question: It's not an honor to win a rigged game, and it's also not a real failure to lose a rigged game. However, there are also many cases where your failure is just out of bad luck, rather than the whole system working against you.
It is very common nowadays that the same paper is rejected by (say) CRYPTO this year, and then get into CRYPTO next year upon resubmission (without much revision). The only explanation is that you sometimes just fail to get reviewers who appreciate your work. It has long been observed that peer reviews are highly random, and from what I've seen, it has become more and more so in recent years (in cryptography at least). It happens very frequently that the fate of your paper was decided by the whim of one or two reviewers. I'll just mention two things that are less commonly talked about:
When the reviewer enters the score of a paper, sometimes there is a "neutral" option, but for most IACR venues there isn't (it varies from year to year). So when a reviewer is truly neutral on a paper, will they give a "weak accept" or a "weak reject" score? This mostly depends on the reviewer's personality/habit, yet sometimes it is a key factor that shapes the final outcome.
If two reviewers suggest acceptance and one suggests rejection, will the dissenter fight for rejection in discussion? Of course, part of it is how the dissenter exactly feels about the paper, but in many cases it's (again) a matter of personality: some people are just more argumentative, while others tend to be silent. But the discussion often decides which papers can get in.
So, if you get rejected, sometimes it merely indicates that you are out of luck this time and nothing else...
A8: The only effective way to actually beat your imposter syndrome is to realize that other people are full of shit, too.
To begin with, I don't find the conventional approach of "(somehow) inducing self-confidence" very useful. If you feel you are incompetent, that's probably true: at the end of the day, you are the primary expert on yourself! So if you have some concrete evidence why you are incompetent, you'd better live with it.
However, I think most people severely overestimate other people's capabilities. Over the years, I've made the following observations:
The majority of PhD students in cryptography don't know how to do a reduction properly. How did I know? Well, I wasn't able to do a reduction properly when I was a PhD student, and I was above average...
There are many papers that promise a "full version" which cannot be found anywhere: sometimes it's because the authors didn't make it public, and sometimes the full version simply doesn't exist. There have been papers like this as early as the 1990s; I haven't seen anyone defend this behavior, i.e., this is universally considered a bad thing — which means that the community has not come up with a way to deal with this issue even after 30 years! Without risking being mean to others, I'll just give two examples of my papers: example 1, example 2 (just search "full version" in the papers).
A lot of published works — including those at "top venues" — contain major flaws (usually a wrong proof, but sometimes it's an assumption that is false). I don't even need to give concrete examples in cryptography, since there are frequently works that point out errors in other works. And if you look at those errors... sometimes they are indeed subtle, but in most cases it's just because the authors weren't paying attention to what they were doing, and any careful reader should be able to spot the mistake. However, three (or four, or even more) reviewers couldn't!
Let me give an example outside cryptography. Around 2000 the editors of the BMJ (formerly the British Medical Journal, one of the top medical journals) did several experiments; quoting a later paper citing the studies,
"At the BMJ we did several studies where we inserted major errors into papers that we then sent to many reviewers. Nobody ever spotted all of the errors. Some reviewers did not spot any, and most reviewers spotted only about a quarter."
The general takeaway is that the modern system of academic publishing is severely broken: authors make major mistakes in their papers, and reviewers are not able to detect them.
So, to overcome your imposter syndrome, you just need to switch your mindset a bit: from
"I am a fraud, I fear being exposed by others"
to
"Sure, I am a fraud, but so are you — so why don't we expose each other?"
A9: I generally don't read papers; I think I've read maybe 15 papers in my entire career. Compared to reading papers, what's much more important is to skim through a paper in 15 minutes - 3 hours and understand what it does at a high or intermediate level.
I have a concrete example of this, in my own area of research. There are around a dozen PAKE protocols based on Smooth-Projective Hash Functions (SPHFs); which one should you start from? Starting from the first one that was published and trying to read all of these papers is a very bad idea. Rather, I started from the one that sort of concluded this topic (in the classical setting and for PAKE only; extensions such as asymmetric PAKE are another story), namely
Round-Optimal Password-Based Authenticated Key Exchange
and carefully read the first three sections of it, which is the game-based proof of their PAKE protocol. This took me an afternoon.
As I said, there are around a dozen works on SPHF-based PAKE prior to the one I linked to (henceforth Katz-Vaikuntanathan, or KV). What's the difference? The main difference is KV is 1-simultaneous round, whereas all prior works require 2 or 3 message flows. This is because KV uses an SPHF that has stronger properties, and in this sense, all prior works can be viewed as relaxations of KV: use a weaker SPHF, get a PAKE protocol that is not round-optimal. Once I fully understood the security proof for KV, I was able to reconstruct the security proofs for prior protocols without reading the papers themselves.
So again, I believe the ability to find good materials to read is much more important than "how to read a paper". Once you have a firm grasp of the "core" materials, you only need to spend a very small amount of time on related works.
A10: Contrary to popular belief, a large portion of academic works is actually clerical work. I am not a super creative person, either (anyone who has worked with me for an extensive period of time can attest to this); my advantage is that I pay close attention to a lot of stuff, which allows me to see some things that others fail to see.
Among all of my publications, around one third of them are re-analyzing existing protocols. For example, if I see a protocol that has been proven game-based secure but not UC-secure, I'll ask questions like
Can you prove that the protocol is NOT UC-secure, and
Can you prove the UC-security of the protocol in some stronger idealized models.
Of course, this requires me to have a thorough understanding of what's going on with the protocol, and what's going on with the security definitions — something that I feel I am better at.
As another example, say you see a security proof under some assumptions. Nowadays, many papers don't write down the reductions in detail: at best they will include a sketch, or in many cases they just say something like "a reduction to [assumption] shows [...]" and that's it. I like to think through those reductions in detail: how they exactly embed the challenges and simulate the games, and so on. Sometimes I figure out that the reduction (or a variant of the reduction) works even under a weaker assumption, which allows me to improve the existing result. Even better, sometimes the published work is wrong!
Another third of my papers are minor tweaks to existing works. I have a paper that (among other things) introduced what's called the Strong Algebraic Group Model (SAGM), which is between the AGM and the GGM. The idea is very simple: the AGM does not have the concept of "steps" in group operations, and you can reuse how the GGM models "steps" to impose the same concept in the AGM. This almost trivial technical idea nevertheless yields a new model that is useful in some contexts such as time-lock puzzles.
The last third of my papers are (more or less) truly creative works, involving constructions of new protocols. I personally find these works to be the most challenging, yet they were NOT published at better venues in general. So overall, I feel the community has decided that clerical works are as important as "creative" works.