Scott Young, an author, programmer, and entrepreneur, is on a journey of continued self-mastery and personal improvement. After graduating from university, he explored two questions: one, what is the ideal way to live; two, what is the best way to learn?
In this interview with The Times of India, Scott answers key questions that would help students, professionals, and learners become more effective in their learning journey.
Q1) As students, we tend to expect the same “types” of problems as we have practiced before. This expectation leads to developing a ‘pattern-matching’ mindset. How could we apply principles, techniques, and ideas learned in class to a problem we may have never seen before, and how could we use this technique to solve real-world problems as well?
SG: This is a big question. I devote a chapter to essentially this problem in my upcoming book, Ultralearning. Essentially it boils down to the problem of transfer. Transferring knowledge learned in one context to another is notoriously difficult and despite incredible effort, there don’t seem to be many reliable procedures for generating transfer beyond simply mastering the material.
The approach I try to outline in my book is directness, which basically means you keep an eye on where you’d like to apply your knowledge while you’re learning it, so you tackle those kinds of problems. This is hard for many students who are learning mostly to pass tests. However, those with a keen interest in learning beyond just school would be wise to start planning applications for ideas and working through small projects when they get a chance to apply those techniques.
But otherwise, you’re correct. I’m not sure there’s an effective procedure which will guarantee you can get knowledge that is fully flexible without significant training.
Q2) Even though we may have used the ‘Feynman technique’ to really understand a concept, at times, however, it is hard to diagnose whether we really understand it, or suffer from the “illusion of explanatory depth,” as you put it. How could we, as students, resolve this illusion? In other words, how to identify whether one truly knows a concept well enough or not?
SG: Depth here is all relative. What it means to understand something is defined by some kind of usage case. Of course, some usage cases are more involved than others, so that means you’re going to have to develop more skill, let’s say, to work out a physics problem than to simply memorize a formula for a flashcard. However, there are many situations that don’t entirely overlap, so learning always must begin with some intended use in mind.
Given this, my advice usually rests on getting feedback. In the case of doing problems and then finding similar problems on exams, the challenge is to seek out problems that are varied enough so that you can really test the breadth of your knowledge. That might also involve working on applications outside of normal textbook problems.
Knowing something “well enough” is always about asking whether you know something well enough for a particular purpose.
Q3) Incentives tend to drive us, and they also tend to drive the desire to learn. I assume this should not be the case… We should be willing to learn, irrespective of the incentive of a grade or the disincentive of Pass-Fail. How to overcome this incentive-disincentive play?
SG: I’m not against extrinsic incentives. I design my projects with them in mind. I think the problem is that external motivators can overwhelm your intrinsic ones, and sometimes that can be a problem when the two incentives aren’t aligned. As in, your extrinsic incentives tell you to cram and just pass this exam, while your intrinsic incentives, which want to learn the subject deeply are being dominated by that pressure.
Q4) In your blog posts, you talk about ‘soft deadlines’ to manage 4 exams at once. However, my question is about managing 1 midterm at a time, and studying only for that course (because of the upcoming midterm), at the cost of all other courses (due to a lack of incentive). For example, if a midterm X is on April 9, and midterm Y is on April 19, we would tend to study only for X until April 9, at the cost of falling behind in Y. Is it okay to do this “swing studying”? If not, how to avoid it?
SG: You need to do what you need to do to study. So, while I’m not saying swinging between midterms is necessarily ideal, if you’re really unprepared for that first midterm then this strategy isn’t too crazy. It’s just a deviation from the ideals of spaced practice, however, few students uphold that ideal.
In general, I try to advocate a continuous learning approach that doesn’t require much “studying” right before an exam. You should aim to learn the material, practice and review it throughout the semester, so all you need for an exam is a light refresher. This was how I did school in university, but I recognize that it can be difficult for many students to implement.
Q5) talks about learning for mastery, however, at times, “beating the curve” overtakes learning for mastery. Is this mindset of “beating the curve” incorrect? If so, why, and how could we help avoid beating the curve, and learn for the sake of mastery?
SG: You’ve brought up this distinction before, so I’ll repeat my views again. I don’t really see the grade/exam paradigm in schools as being *necessarily* harmful for learning. I modeled my learning path on taking MIT exams. While this may have some downsides, I think picking a rigorous standard to reach also pushed me to learn a lot of things I might have dismissed had it just been for self-education. So it is with “beating the curve” or “getting straight-As” or other academic incentives—they can be a detraction if they focus you on methods that don’t align with deeper learning, but they can also foster deeper learning by giving you encouragement to learn things you might not otherwise. It really depends on the student.
I worry more about “beating the curve” attitudes in smaller classes where the competitiveness can lead to students working isolation, failing to help each other study and envy/jealousy for people who “throw” the curve. I think this is probably a much smaller problem because when you take a large enough class, curves tend to be extremely regular so this kind of problem is less so. But in a class of fifteen, I would hate to be graded on some kind of competitive basis.
Q6) In any student’s day, one always has these 5-, 10-, or 30-minute intervals. These could be either between classes, or while walking, or even waiting. How could we make the most of these short time intervals?
SG: I like using flashcards/Anki. These aren’t useful for every subject, but there’s a lot they’re useful for and they’re super easy to squeeze in a few repetitions while you wait for things. I learned a lot of Chinese that way.
Q7) How to ensure that one has a deep understanding of the assignments instead of understanding just enough to scrape by and get full credit on them?
SG) I’m convinced understanding isn’t really something you have or lack, without reference to a specific context. The kind of understanding needed to ride a bicycle and to correctly design one are almost mutually exclusive, yet both and a bicycle engineer could be said to “understand” bicycles.
Similarly, to say you don’t understand something enough, yet scrape by on assignments, is implicitly saying that there’s some other, harder context, that you’re unable to apply your knowledge towards. In that sense, having an eye towards application can help.
Q8) Since one might forget “crucial” theoretical course content (for e.g. the specifics of row-reducing a matrix, intricacies of multivariable calculus, or tedious formulae and their derivations), what is the point of studying those in the first place? In other words, why is theory or tediousness still important, even though one might never use it? (I ask this question in the context of a computer science degree. Many claim that one should only take those classes which would be “useful” in industry. For example, few claim that one need not take the mathier version of because all that math is not actually needed in industry; rather, one should take the more programming-oriented version of ML. Which approach do you think is better, and why?
SG: I don’t think the students are wrong. If your goal were to reach professional standards as quickly as possible, doing a lot of math is probably not the fastest way to become a good programmer.
The justification for a ton of math usually comes because of the university structure, rather than any deeper reason. Math is *extremely* important for academic CS (which some might argue is really just applied ), so professors try to impart to students what they feel is important.
The other place I can see math mattering more is if you’re going to be working on the technical frontier. So maybe your average programmer who is just going to build solutions that fit within previous work doesn’t need much math. But someone who wants to push a new paradigm of machine learning would definitely benefit from the higher math that underlies why those functions work.
My feeling is that math is good, but probably overdone for many highly-ranked CS curricula.
Q9) Oftentimes, we may develop a love-hate relationship with a subject in that they might like it only after we “get it,” but we may continue disliking it unless we “get it.” How to navigate this situation of liking a subject only after one “gets it,” and what does that say about one’s passions and interests?
SG: This is true of everything and everyone. You like things you feel you’re good at. You dislike things you don’t.
Honestly, I don’t know whether our likes and dislikes have much more content besides this fact. “I don’t like X” is generally synonymous with “I’m not very good at X, and so it’s frustrating and unpleasant for me.”
One of the reasons I care so much about learning is that it’s a transmutation process of going from “I don’t like something” to “X is really cool.” If you can do that over and over, you build confidence and it expands the places you draw meaning from in your life.