Should I Go to Grad School?

25 Mar 2018

If you were to ask me, “On a scale of 0 to 1, how much do you think you will want to go to grad school rather than straight into industry when you finish undergrad?” and plot the answer over time, it would probably look like this:

Very underdamped oscillations

I’m on the third oscillation right now. They’re not small oscillations, either, and if asked to estimate a damping coefficient… well, let’s just say it looks like Google’s life extension project had better make some real breakthroughs if I’m ever to converge to a decision.

In the hope that writing it all down will reveal some patterns in the chaos, I’m going to use this page to document my thoughts on grad school. I expect this to be a long work in progress.

Table of Contents

  1. Background
  2. Why Go to Grad School
  3. Analyses
    1. Pascal’s Wager: Grad School Version
    2. A More Thorough Cost-Benefit Analysis
    3. Predicting the Future
  4. Pieces of Wisdom from Other People
    1. If you have no interest in remaining in academia afterward, then you should not get a PhD
    2. You should not go into industry first because you will find it very difficult to come back later
    3. You should get a Master’s degree because it will protect you from lay-offs
    4. There are two types of people in the world: those who will be happy either way, and those who will be unhappy either way
  5. Further Reading
  6. Final Result

Background

My attitude, when I first decided to go to engineering school, was “I just want to learn some cool stuff then get a job.” Computer Engineering was my initial goal, since that seemed to be pretty good on both fronts. But students don’t choose majors until second year in British Columbia, so I took the generic first year science and engineering courses.

As it turns out, I’m a lot better at school than expected. Also, I ended up really liking physics and linear algebra. Suddenly grad school became appealing. At the end of the year, I chose Engineering Physics, partly because it had the reputation of being “the grad school-oriented engineering program.”

This is where my oscillation begins. By the end of my first year of EngPhys – second year of undergrad – I will have gone from being pretty certain I want to go to grad school, to pretty certain I definitely want nothing to do with grad school, then to pretty certain I will live a miserable sham of a life if I don’t go to grad school. 1

It was pretty clear that I needed to figure some stuff out about myself and what I wanted out of life.

Why Go to Grad School

Why, when I’m studying things that are valued in industry, would I possibly consider throwing away 6 years of my life for a degree that, everyone knows, will never pay me back?

The answer is that I want to leave this world feeling like I’ve made a lasting impact on it, that I have pushed the frontiers of human knowledge or brought into fruition things that were previously thought impossible and that will outlast me. I want to walk a path that no one ever knew existed.

I want my work to be challenging, because I have frequently found “challenging” and “rewarding” to be synonymous in my life. And I want to be among a community of like minded people who similarly find it thrilling to tackle problems that may take years to solve. And I think grad school may help me achieve these things.

“Being able to devote myself to challenging, rewarding work that helps advance humanity’s collective knowledge” is a bit of a mouthful, so from this point forward I’m going to call this “the holy grail” or “HG.”

Analyses

These reasons seem perfectly convincing to other people when I explain them, but for some reason, it’s not so clear to me. For one thing, it may be true that grad school will give me a better chance at achieving these goals, but how much better?

I’ve met plenty of PhD holders who work in jobs I would not, personally, classify as holy grails. In at least one case, it was because they could not find the sort of work they wanted. 2 To be fair, most of the others seem pretty happy, which is great – but I don’t think I would be among that group.

Pascal’s Wager: Grad School Version

There are two choices: go to grad school (GS) or don’t go to grad school (~GS). To keep it simple, we’ll assume that the outcome space is neatly divided in two: find the holy grail (HG) don’t find it (~HG).

If I finish my Bachelor’s and end up ~HG, then that is a hit to my overall happiness. Perhaps I’ll feel bored and always be considering going back. But this would be much worse if I had first spent 6 years in grad school in the hope of finding the HG – then I will always know that I failed my quest.

If I go to grad school and it turns out as planned, I will gain some happiness. But if I go directly into industry and end up finding fascinating work without an advanced degree, then I’ll feel like I’ve hit the jackpot.

My Pascal’s wager looks like this:

  HG ~HG total
GS +1 happiness -2 happiness -1 happiness
~GS +2 happiness -1 happiness +1 happiness

~GS wins in this extremely simplified analysis.

A More Thorough Cost-Benefit Analysis

Suppose there are two students both graduating this year, and one chooses to go directly into industry while the other chooses to pursue a PhD.

The average entry level software developer in Vancouver earns $60k a year, while a computer science PhD student at UBC earns a stipend of $24k per year.

It’s probably safe to say that our industry student, since they had the freedom to choose between the two paths, is probably doing a little better than average. Also, they will be getting raises, while PhD stipends are fixed. So, let’s estimate conservatively that in the first 6 years after graduating, the student will be ahead by about $300k.

Worse, these are the prime saving years. Those $300k, if invested at 6% per year, will result in… oh, something in the range of $1-2M depending on when they retire. It is very unlikely that their PhD-pursuing counterpart will ever catch up.

None of this is particularly new or interesting, but in the tradition of liking to put numbers to things that people usually prefer to avoid putting numbers to, I wanted to quantify exactly what the difference is. Because, ultimately, the equivalent of about $1.5M in intangible future value is what a PhD needs to provide, if it is to be definitively a better choice than industry. This means that we need to figure out how much, exactly, the holy grail is worth, and what my odds are of finding it with only a Bachelor’s is vs with an advanced degree.

Suppose I value the HG at $5M and my likelihood of finding it is 20% without grad school and 80% with grad school. Then my expected gain from choosing grad school is $3M. I should definitely go to grad school.

On the other hand, suppose it turns out I had a 50% chance of finding the HG and going to grad school increases it to 75%. This makes it a not so great choice.

I have no way of predicting these numbers. I don’t even know if $5M is a reasonable value. Maybe it should be $100M. Maybe it should be $0 because some of the world’s happiest people don’t even have jobs – they just have, I don’t know, lots of pleasingly plump grandchildren.

This strategy is not going to work very well.

Predicting the Future

I’m a bit of a podcast junkie, and one of my favourite listens in the last year was Decide Already!, a podcast episode about why humans are often so bad at making decisions. Of course, it immediately got me thinking about grad school.

According to the podcast, one of the reasons humans suck at making predictions is that we’re really bad at predicting the future. And the reason we’re bad at predicting the future is because:

1. We fail to account for the details of what the future will contain.

That is, we try to imagine the future, but we’re really only imagining one or two things at a time, while in reality there will be thousands of little things that can completely change the balance.

Details about grad school that don’t make it into imagination

For example, in my imagination, as a grad student I will get to choose a small handful of domains to focus on, based on what interests me the most. I will take some classes related to those domains, meet people with related interests and share ideas, pick some projects that look interesting and see where they lead. Some might fail, but others might take delightful twists and turns. One of them, over the course of hundreds (thousands?) of nights, will eventually be molded into a thesis.

But this is only the highest level description. Since I started spending two days a week at the Systems research lab at UBC this term, here are a few more details I can now add that would not necessarily have occurred to me before:

  • the amount of bureaucracy at a university is out of this world
  • TAing can take enormous amounts of time, which sucks since you want to be researching
  • but on the other hand, some people also find it highly rewarding to teach
  • grad students can develop super close relationships with each other, much more so than usually happens in a workplace
  • grad students can develop super close relationships with their supervisors

And, obviously, there is a lot I’m missing because I’m not actually a grad student, but this is a more detailed picture.

Details about working in industry that don’t make it into imagination

TODO fill this section later this summer when I’ve spent a few months at my next internship.

And, obviously, there is a lot I’ll be missing because being an intern is not the same as working permanently, but this is a more detailed picture.

2. We fail to account for how much we will change.

A really fascinating piece of research that the podcast presented involved groups of 18-year-olds and 28-year-olds. In this study, the 18-year-olds were asked to predict how much they will change in the next 10 years, and, overwhelmingly, they did not think they would change very much.

Of course, all the 28-year-olds said that they had changed enormously over the previous 10 years. This result is probably completely unsurprising to anyone over the age of 18.

The part that was surprising is that they found a similar difference when performing the experiment on groups of 58 and 68-year-olds: the 58-year-olds thought they had basically done changing, while the 68-year-olds felt that they had become completely different people (but that they probably wouldn’t change much more from now on).

So, I currently want very badly to find the holy grail. But in 10 years, will I still want it so much? What if future-Lise ends up preferring the pleasingly plump grandchildren?

But fortunately, there is one more saving grace, which is…

3. We’re really good at rationalising our decisions after we’ve made them.

There’s another example in the podcast where they ask a group of student photographers to select from their own portfolios one photograph to keep and one to give away, and… well, I’ve already spoiled way more of that podcast than I should, so just go listen to it – it’s well worth the 20 minutes.

The gist of the story is, after we’ve made a decision, and particularly if it’s an irrevocable choice, we immediately start rationalising it and very quickly convince ourselves that our decision was absolutely optimal in every possible way.

Is the best answer, “Lise, just shut up and flip a damn coin”?

Not necessarily, because…

4. The best method is to ask other people who have already made the decision and choose based on their outcomes.

If the majority of people think choice A is better than choice B, then choice A is probably a better choice. We all want to think that we’re unique and our situations are completely different, but that’s usually not true.

The only problem is that most people who talk about the decision chose to go to grad school. They write articles like these.

More to the point, they fill the hallways of my school and teach my classes. They are my and my classmates’ mentors and role models. Almost everyone with whom I’ve discussed the notion of grad school is either a grad student, a post-doc, a professor, or someone who hasn’t yet decided. A few are people who worked in industry and then came back to pursue a PhD.

What I wish I could find is some examples of people who really thought about grad school and almost went, but decided instead to go directly to industry. But nobody really seems to talk much about this. Are they all happy and fulfilled in their choice? Do they feel like they have found one holy grail among many? Do they feel like they found the holy grail?

TODO I don’t know yet which way I’ll choose, but I promise that if I choose industry, some years from now I will come back and give an answer.

Of course, future me will have had the benefit of years of rationalising the choice, so it will probably be very tempting to say “But the answer is obvious.” Maybe (probably) I’ll be reading through this whole thing and cringing at my naivety. But I promise to at least try to recall just how agonising the decision was, so I will hopefully give it a serious answer.

Pieces of Wisdom from Other People

Collected over the last few years, with some commentary from me.

TODO Add more as they come in.

If you have no interest in remaining in academia afterward, then you should not get a PhD

I’ve heard this one quite a few times and have come to the conclusion that it is probably not true. Perhaps this advice makes sense for people in the humanities, but I don’t think it is relevant for people in technical fields. There is room in industry for people who can build new technologies.

Also, considering how rare academic positions have become, maybe it’s actually better to be more interested in industry.

You should not go into industry first because you will find it very difficult to come back later

Also one I have heard many times. My counterargument is that if you find it difficult to come back because you like working, then isn’t that a good outcome?

Frequently when people say this, it is combined with the phrase “golden handcuffs” – the idea that you are forced to stay because you enjoy having a comfortable middle class income too much. This seems needlessly judgmental. If someone values the income from working more than the value of going to grad school, then they have made a valid choice.

As a further counterpoint, there are currently two PhD students in the NSS lab who worked full time before starting their PhDs.

You should get a Master’s degree because it will protect you from lay-offs

Also one I have heard many times. Personally, this feels like a weak reason to go to grad school if you are otherwise uninterested in it, but perhaps for someone who is on the fence it makes a convincing tie-breaker.

There are two types of people in the world: those who will be happy either way, and those who will be unhappy either way

I have only heard this once, but it has really stuck with me. The speaker’s stance was that actively seeking out interesting work and taking joy in seeking out new challenges is, in the end, much more important than which path you choose.

Perhaps the coin toss is the best option after all.

Further Reading

A collections of articles aimed at undergrads considering grad school. Mostly CS focus because… that’s probably what I’d go for.

Articles from people who went to grad school

Phil Agre, Advice for Undergraduates Considering Grad School
Ronald T. Azuma, So Long and Thanks for the PhD
Mike Gleischer, Grad School FAQ
Andrew Peterson, Should I Attend Grad School in Computer Science?
Theo Vassilakis, Why You Should Work before Grad School

Articles from people who didn’t go to grad school

TODO Find articles…

Final Result

To be determined! I look forward to revisiting this page in a few years and getting a good laugh out of it.


  1. I recently switched back to Computer Engineering, which has had no effect on my grad school oscillations. 

  2. This person was a close family friend. Having spent a lot of time with his family while growing up and overheard a lot of grown-up conversations, I can say with some confidence that he was not terribly happy about having earned a PhD and then being stuck working a job he didn’t like.