The real problem is teaching children that "reading" is merely decoding words off a page, and "listening" is merely hearing the words someone says.
The book itself is a way to open the door to learning, not the end point. Reading subject titles first, cross-referencing, and working examples and doing projects help teach and cement ideas.
There's techniques like SQ4R [1] for reading, and hugely applicable to general technical and article reading is "How to read a paper" [2]. For lecturers, there's "How to Speak" [3], and for those attending a talk there's the Cornell Notes system. [4]
To others wondering about the discrepancy between SQ4R being mentioned in the comment text and the title of the linked source being SQ3R (https://en.wikipedia.org/wiki/SQ3R), it looks like others [1] add an additional R (marked with an asterisk below), for the following:
When I studied Physics, I could sometimes take hours to digest a single page of a textbook or my lecture notes - providing that it was a good textbook.
It’s not enough to just understand what is written. You have to understand why it is like described. Why it can’t be different - or could it? You need to build an intuition for both the physics and the math describing it.
When learning new programming concepts, I would often have to learn them 5 or 10 times before they clicked. Once they clicked, they seemed completely obvious, and I couldn't imagine what it's like to not understand it.
Two mathematicians are discussing a theorem. The first mathematician says that the theorem is trivial. He then proceeds with two hours of exposition. At the end of the explanation, the second mathematician agrees that the theorem is indeed trivial.
I remember this having been my experience with teaching myself to code as a kid. All I remember is initial endless frustration from having to re-read tutorials and documentation and still not understanding anything. Then at some point I can't remember, everything clicked into place and I can't understand why I didn't understand it before. Once I had those basics down, I haven't had that kind of struggle.
I went through a similar phase with math, struggled with basic algebra for years, then at some point it all clicked into place and I unintentionally ended up skipping all high school algebra to get to early calculus.
It makes it really hard when helping other people through those early concepts (like pointers), because "keep working and at some point it'll just click for you" isn't very convincing.
> keep working and at some point it'll just click for you
Haha, monads still haven't clicked for me despite knowing F# and Clojure pretty well. I think of them as "functors that you can flatten" (and functors are "things you can `map` over", and `map` is "a structure-preserving transformation"). I know monads syntactically, but not intuitively. Despite this I still manage to work with monads pretty well - sometimes rote knowledge is enough. Maybe if my work involved discovering new monads I'd feel like I'd know them better. (I just see the same option/list/result(ish)/generator-style monads over and over.)
Ah, so you understand monads. The trick is that someone told you they're a big important deal. They aren't. They're useful, so they get used. But there's no great insight or revelation waiting for you. You've already got it.
I’ve started to call it paying the “time tax” or “staring tax”, I have to just review tough concepts (code, concept, leetcode solution, math problem) for a long while before it eventually clicks.
That brings to mind a learning staircase that comes up in art, where you go between stages of learning to draw better and learning to observe better. Since you can't draw better until you can observe better, but you can't observe better if you can't draw well enough to match your current observation skills. So in a similar way, you have to pay some "time tax" to move forward.
I used to require this, but after having spent many years programming this stopped being the case. I've reached the point where i can see a new abstraction and usually immediatly understand it, regardless of language.
Has your career taken you across many different programming paradigms?
A C programmer with no experience in concatenative or, especially, functional languages will have little chance of understanding code written in Forth or Haskell. They will have to start at the bottom of the learning curve.
The differences between languages aren't just skin-deep. Same goes for the design patterns in those languages. Haskell's applicatives have no analog in C, for instance.
You're not wrong, but those don't crack the top-10 of in-demand languages. Enough functional paradigms have also crept into more popular languages that you wouldn't be completely lost.
Yes. This is the same for both math and programming. These skills take a lot of practice to develop. It’s like playing a musical instrument: you won’t learn how to do it if all you do is read the book.
What’s strange to me is how people seem to resist this fact. Like it’s obvious to pretty much everyone that you need to practice an instrument to be able to play music on it. But for math and physics at least they seem to think it’s somehow different? That they ought to be able to solve the problems just by picking up a pencil? And when they don’t get a decent grade on the exam they look over at the person who got 100 and say “that’s why! I can’t do it because I’m not a genius like they are!”
The feedback is a lot faster when playing an instrument, you hear if it sounds good or not. The reward is built-in. This is a bit harder for maths/physics once you're past the trivial stuff and most people don't find it rewarding to solve physics/maths problems.
>The feedback is a lot faster when playing an instrument, you hear if it sounds good or not
The feedback is only fast if you're already skilled. It's pretty common for new students to have no proper sense of what sounds good (particularly since it's familiar to them), such that they struggle to figure out what they need to improve. Similar to digital artists flipping the image often to make it easier to identify issues by making the image somewhat unfamiliar to them, forcing their brain to actually look at the image rather than just using its previous idealized mental representation of what the image looked like.
> most people don't find it rewarding to solve physics/maths problems.
Maybe that's the difference between people who are good at it and people who aren't. I found solving math problems (the ones you get in a practice course at a university) hugely rewarding. Finally figuring it out was so satisfying.
Coincidentally, my endeavors playing an instrument were never particularly fruitful.
To be fair, the "you need to practice an instrument to be able to play music on it" hides a very complex process.
Some people "practice" by trying to play the same thing over and over, and it is equivalent to "hearing" vs "listening". Learning how to play an instrument (and also a particular piece of music) is a very complex process. Just like you need to learn the skills to "learn from a book", you need to learn the skills to "practice an instrument".
I agree with your point but I think playing a musical instrument is a bad analogy because of one fundamental difference:
In most intellectual fields, you can take your sweet time to remember what you learned so long as you remember it eventually. Music requires practice to recall everything in real time.
Not everything is hard work. Some are smarter (or better equipped) than others. You can spend an eternity teaching me math and still won't be able to accomplish anything. It's almost like different people exist and they have different skill sets and interests.
Unless you have a specific learning disability, I refuse to believe that you are unteachable. Maybe you have zero interest (in which case, of course you won't learn). That's fine. Maybe abstract reasoning doesn't come as natural to you as it does to others, that's possible. But we know how plastic the brain is. Given enough time and dedication, you should be able to learn things - certainly not PhD level stuff, but the basics.
> Maybe you have zero interest (in which case, of course you won't learn).
Quoted sentence does a lot of heavy lifting in your reasoning. Being intrinsically interested in something is integral part of learning. And, yes, obviously for people who have zero interest in subject it will act as "specific learning disability". I know this because I tend to be interested in programming-related subjects and, consecutively, capable of learning them, but I could never muster any interest trigonometry or calculus homework. I've also known multiple people who felt the same way about debugging their software development homework - they weren't interested and hated every second of it.
I don't think you can count "zero interest" as being a learning disability. I hated drawing and still do (and other things too), but I could probably still learn the basics of it if, say, my life depended on it.
I guess you mean well but this sort of reasoning is borderline offensive. I said math is hard to me and you immediately jumped to "abstract reasoning" as if that means math and nothing else.
I apologise if I have caused any offense, because that wasn't my intent.
I feel you missed the main point of my comment, though. Even if abstract reasoning, or certain forms of abstract reasoning, were not natural to you (something that I can impossibly judge), it wouldn't mean, as you initially claimed, that it would be impossible to teach you mathematics.
Oh, understood the main point just fine. You are not wrong.
Just got a little upset about the rest. I wouldn't have reacted as I did had you phrased it a bit differently, like in the post I am currently replying to.
I don't think your offense was warranted. You've got a chip on your shoulder. Other people may be able to tell but either haven't confronted you about it, because it wasn't worth it to them, or you haven't responded well to feedback.
Yeah I experienced this firsthand. My roommate in college was a physics genius. He would grok concepts without really studying very intensely— often after a single reading or lecture.
He would intuit solutions to problems based on extrapolating from the textbook. The undergraduate stuff all clicked right away for him.
Even when we scored the same on an exam, the difference in effort expended was dramatic
You are confusing the difference between a necessary condition and a sufficient one. The argument was that practicing was necessary to gain a deep understanding of math, not that it was sufficient.
Ironically, this is one of the first things they teach you in college level math (or at least they did in my case).
I’ve tutored many students in math and it’s really common for people to feel the way you do. They’ve had bad experiences at some point or another, whether it’s with teachers or classmates.
There are a lot of different reasons why students feel the way you do and (unfortunately) give up on math. It’s a shame because I believe most of those reasons are specific to the high school setting and from what I’ve seen, students have a much better time if they come back to math later in life and study without that environment.
Interest also plays a role, of course. And not everyone will have the interest to go back and study again, later in life.
I did come back to it as an adult. Spent a while doing 2d and 3d graphics. It was much easier as an adult but rendering involves too much of this math stuff and I gave up. Definitely not for equipped to shine as graphics programmer.
Which brings us to my original point - not everyone can do it (well).
This is the "problem" today that I see in the classroom (I used to teach CS until recently)... learners today do not want to dedicate the time it takes to truly understand. Not all of them, but a LOT of them have the opinion that they just need to get through this damn algorithms class, get the damn piece of paper (diploma) so they can go out and start a job doing "real" programming.
Those students, they do not get FAANG jobs, or make big money. The few that want to understand are much more like the articles description, and they end up five years on far ahead of their peers at work.
I wish they'd always explain 'the initial seed' first. What was the initial thought or idea, that, when
pursued and expanded on, resulted in the concept that im trying to learn right now?
That sort of initial seed would give me so much information...
I find it beneficial to just keep asking why. Sure, the answer is sometimes "because universe", but asking why (not just how) over and over again helps me build my mental model.
On that topic... Does anyone have any clue how and why magnetism works? I have a A Level Physics test next week :)
I simply can’t read the text book, at least without context. Doing the problem sets over and over again works better for me. Practice is the key, it might make sense to unsuccessfully attempt the problems before ready the chapter, since you would have more context for readying.
I never really got the difference between the dot and cross products until many years after graduation. And then I only got it because I managed to stumble upon a good video with somewhat decent animations.
Like, 2-D, sure, seems easy enough. But once you go to 4-D, things get really hard to grok.
Passionate science educators online have really filled in a lot of my education.
math requires something different than memory, it requires understanding and fluency in its variety of symbolic languages, which can only built with intellectually engaged work.
math requires an absurd amount of memorized techniques/patterns ... maybe that's what you mean by fluency?
of course these are rarely static (after all one can learn proofs of theorems without being able to solve any problem), hence the need for the "muscle memory" of how to apply these techniques to the particular patterns.
Books work well and have been the standard for learning
many topics for hundreds of years.
They have few technical problems,
Easy to sell, borrow, give away.
Yes, you can go about reading a book in different ways
depending on what and why you are reading and what works
for you.
I think it is highly unfortunate that universities these days,
at least in science, then to force students to buy a new edition
every year, or at least every couple of years.
There are just about no scientific discoveries in math that requires
undergraduate and few matters that require an entirely new edition
of graduate books.
I have also heard but never experienced that some books apparently
have an online component that is valid only for one person or one
semester or possibly one student per semester.
If true this is even worse.
I strongly prefer books to videos.
A lecture is good to prepare you or to wet your appetite.
And to clarify known difficult spots in the book.
> have been the standard for learning many topics for hundreds of years
That means very little, history is littered with awful standards used for centuries, there is no magic that translates longevity into greatness
> They have few technical problems
Hard to bookmark (and find what those are for later), hard to search, hard to annotate (and search those later), hard to edit, hard to embed rich/dynamic content, hard to track understanding, hard to adjust content based on that, hard to tailor to a specific reader in general, hard to update layout, hard to collaborate, ...
Well, books continue to work for a lot of people (including myself). Every time I read a book about a topic, I come away with deeper and more contextual knowledge than if I just piece together scraps of information from badly written blog posts and overly technical documentation - and that's only for programming stuff, for history etc. books tend to be even more valuable.
> Hard to bookmark (and find what those are for later), hard to search, hard to annotate (and search those later), hard to edit, hard to embed rich/dynamic content, hard to track understanding, hard to adjust content based on that, hard to tailor to a specific reader in general, hard to update layout, hard to collaborate, ...
Not all of these are even true of physical books - people have been bookmarking them and scribbling in the margins for ages. And also, e-books and PDFs exist, for the people that prefer them.
As for the "tailoring to a specific reader" and "tracking understanding", that's a problem for any learning resource, as people are incredibly varied. Maybe in certain situations, learning resources exist that are extremely adaptable, but I've rarely seen that. In general, for a given topic one should check out different books until one finds one that is well-suited, skip sections that are obvious, consult other sources where one remains confused, do exercises if the book has them, etc.
Books being hard to edit and/or collaborate on could maybe (if it's even true) be a problem for authors, but why does it matter for readers?
The one thing that I do consider true is that you can't do extremely fancy visualisations in a book. But there's no problem with supplementing a book with e.g. a website where you can find interactive visualisations, etc. (also: simply not every topic needs complex interactive visualisations).
> Not all of these are even true of physical books - people have been bookmarking them and scribbling in the margins for ages.
All of that is true for physical books, if you're limited to tiny margins and can't easily find your margin scribbles later, then it's a format fail, you just continue to use the same flawed logic "done for ages = good"
> And also, e-books and PDFs exist, for the people that prefer them.
Which mostly repeat the paper medium, so fail to solve most of those issues
> As for the "tailoring to a specific reader" and "tracking understanding", that's a problem for any learning resource, as people are incredibly varied.
Yet it's especially a problem with books
> In general, for a given topic one should check out different books
In general one should not limit oneself to books, even for just a starting point
> Books being hard to edit and/or collaborate on could maybe (if it's even true) be a problem for authors, but why does it matter for readers?
For example, to remove "sections that are obvious", add "exercises", find reference to "other sources", see explanations from other readers etc.
> there's no problem with supplementing a book with e.g. a website where you can find interactive visualisations, etc.
The problem is it's a big limitation of the format, and you can't integrate it well if you have to switch back and forth
> also: simply not every topic needs complex interactive visualisations
> Well, books also continue to fail a lot of people
Yes, because learning is hard. A book is not a guarantee that you'll learn something - but neither is any other sort of resource.
The thesis of this discussion is that "books don't work" and that's blatantly false. Maybe there's certain people for whom they really "don't work", but as a general statement, it's false.
> Which topic do you think can't benefit from one?
I don't need fancy data visualisation if I read up on the history of the Roman Empire, for example. Or if I want to read Plato. Or to understand axiomatic set theory.
---
I'll just put it differently: if you prefer other kinds of resources and they work for you - great. But to say "books don't work" just means that you artificially limit the things you learn. Writing books is comparatively easy (if you're an expert in the area). Doing fancy interactive experiences etc. requires much more time, skill, knowledge etc. - there's probably 3 to 4 orders of magnitude more stuff that has been written in books than has been made available through other means, so saying "books suck" means just locking yourself out of a lot of stuff.
You're wrong re the thesis: this discussion is about the limitations of the format the op claimed were "few", and those limitations are what help fail people, you can't ignore it with a "learning is hard" mantra, that's not relevant when the argument is that it's made hardER by the limitations
And of course just dumping your expert knowledge on the page is easier than creating an effective learning experience
Boomkarking a physical book is super easy and a lot more intuitive and easy to use than an eBook version.
Some books come with one or two strings built in that can be used.
Mostly to track where you are in the book right now,
and the other one for something you find important.
Then there are thousands of different bookmarks you can buy
as many as you need.
You also have adhesive-colored tabs you can use to mark whatever you like.
You can also bend the top or bottom corner of a page,
Or use scissors and remove the corner entirely.
Bookmarking technology for real life books is mature and easy to use.
Some books make it easy to write down here you are externally.
Page Numbers, or like the Bible with several numeric markings
that makes finding a quote fast and easy.
Books can come with indexes that make finding things easier.
To browse through index entries can also provide interesting
insight.
It is super easy to annotate.
You can use a pencil and things can be erased if you desire.
You can use pens
You can use highlighters,
You can use stickers,
You can write the margins
You can attach post it notes
You can attach documents (pressed between the existing pages.
The way you do bookmark, highlight, and scribble can be of tremendous
value for the next reader.
(Or it can be flat out annoying).
It is also easy these days to take a photo of a page with your cellphone
and do whatever you want. then you can if you so wish print and glue
it in, or insert a page or whatever into the book.
A book can have very rich illustrations and graphics.
If you do need to reference rich media you can print a QR code in the
relevant part of the book so people can quickly pull up the video
or whatever there is a need for.
You're describing a bunch of variations is a primitive unsearchable design, there is no value in that sticky richness related to the issue I've described
How do you group all the bookmarks that are related to a given topic? Use a specific bookmark color? How do you then change the whole group? Or, physically go to every single bookmark and change it to a different color? How would you list all paragraphs you've bookmarked together (let's say you want to track a story of a given character interaction with another one)? How do you add longer custom text to the bookmarks? Physical ones don't have enough space for your questions that you'd later like to go back to to find answers to
I mean even in the most trivial case- using a string to track where you are: why is this even needed when a computer can do the tracking automatically? Also, that string only tracks 2 pages, not the exact point in text
Indexes are also not great, they're slow to use, usually incomplete (so any primitive full-text search would beat them), and uneditable by the reader, so don't reflect
"Take a picture" is a laughable suggestion, why would I also need to waste time to OCR "to do whatever I need" when I could start with a better format in the first place?
> I think it is highly unfortunate that universities these days, at least in science, then to force students to buy a new edition every year, or at least every couple of years.
This seems to be a specifically American (or anglo? not sure about the UK/Australia/etc.) thing. I've never heard about this happening in Germany where there are standard textbooks in about every field that people have been using for decades - and in many cases, people just simply go with the lecture and the associated notes instead of a textbook.
I studied CS in Germany and pretty quickly learned not to buy textbooks too fast: lecture was announced with book X as necessary to have.
First lecture the professor holds up book X. "I have taught using this book for years. A few days ago I decided to use book Y (holds it up) going forward."
TIL, but still: the lecturer changing their book preference at the last minute still seems different than "you have to use edition 15 of this book (which came out 3 months ago) which is exactly the same as all the previous editions, but the exercises are rearranged, so you'll get confused if you use an older edition", which - if I understand it correctly - is what seems to happen sometimes in the US.
Do you have any evidence specifically about lecturers being bribed? Because there could be a whole other range of other explanations for why this practice continues.
working for two uk universities. and this not the only stuff these corrupt people got up to - one had the cheek to photocopy his own (crap) book using free to him university copiers, and then make his students pay for the copies. nothing was done about it, of course.
That's of course deplorable. But it just feels to me that if it happens so regularly, there must be some systemic causes and not just instances of invidual educators being bribed
what other reason can there be for very, very bad authors have new editions of books come out every year and get enforced on students by tenured clowns?
> it happens in the uk as well as the us - university lecturers are bribed by the publishers.
I really, really, really wish people would quit repeating this without evidence.
Maybe ... maybe ... this is true for some huge name who can force the adoption of some huge number of textbooks. Maybe.
However, for any random lecturer who teaches an upper level technical class, the maximum level of revenue any seller can expect from that class is probably about $200 x 50 students = $10,000.
Where the hell is the bribe money going to come out of that? By the way, last I checked, the publishers wouldn't even let us keep the sample copy they sent to us. Yeah, they're that cheap.
And do you really think that your technical lecturer/professor or anybody in the publisher is going to risk going taking potentially illegal behavior for $1000?
If your lecturer/professor is that money driven, they would be better off simply not taking an underpaid, overworked position teaching your sorry ass and instead teach the same technical material to professionals who will pay more than your semester tuition for a one-week program.
Now, if you want to talk about the publishers bribing government bureaucrats over high school textbooks, that is an entirely different and completely documented thing.
The university textbook issue was big when I was a student in the mid-late 90s.
In fact, one of my CS professors went on a rant about it, how it is simply a way to invalidate the used book market... he then told us to buy books from an online storefront with this funny name (that he assured us had nothing to do with the rainforest) that sold them for a sizable discount instead our own university bookstore.
A top down description: start with what something is, why it's important, and how it fits in with other things, then an outline of the details, then the finer details and so on. This is almost impossible to find in textbooks or lectures, which for the most part are bottom up.
Trying to figure out the thing without any help, or very little help, at each step of the above. This sounds pointless, but it actually works really well. If I try my best to figure out how something works without knowing anything I end up "priming" my brain for the answer, and end up getting a (relatively) instant understanding. It's like you make a hole that is the rough shape of the answer in your mind and the answer fills the void. Put another way you motivate your brain to wanting to know what it is. The best bit is how far you can get by yourself and the positive feeling that gives you.
This might just be me, I've never met anyone who feels the same way about learning.
I don't think it's just you, I have a similar need for a holistic understanding of the process/needs before I feel that I'm learning/contributing in an organized way. I left the biotech startup world a while ago because I found that in the organizations I was working with, there was an extreme siloing happening from top down management, and I chafed under the unspoken assumption that I was expected to operate as a black box transforming data from input to output without knowing the whole process
This isn't helped by how the biotech world stratifies itself into PhDs and non-PhDs
Towards your first point, I love interacting with LLMs (ignoring all the issues about confabulation etc.) because I can ask it for top down, high level outlines and drill down to the bottom however I please, and ask for multiple paths down as well.
I agree, that's a bright spot which I've just started to get into. It's a real boon for personalised learning. Even mistakes by LLMs are helpful: you know that they're unreliable and you have to stay on your toes, so what they're telling you makes sense. That only benefits actual understanding, rather than rote learning.
> Trying to figure out the thing without any help, or very little help, at each step of the above. This sounds pointless, but it actually works really well
I found it to be very effective though time consuming. Even if you don't figure out the solution, when you read other solution later, everything just clicks.
I like to think about it when I'm doing something else that doesn't require thinking. You also have to let it percolate. It kind of reflects the focused/diffused thinking modes of actual learning.
I've used spaced repetition to memorize nearly 5,000 words over the last 15 years, but I struggle to apply spaced repetition outside of key/value pair memorization.
It seems that SRS is best used for "know what" rather than "know how".
I am curious if anyone has applied spaced repetition to memorization of things other than terminology and how they did it.
I built an Anki deck[0] to prepare for and pass on my first attempt the Certificated Flight Instructor (CFI) initial oral exam, considered by some to be the toughest checkride — although pass rates have been rising in recent years[1].
The problem with SRS in my opinion is the inability to do many-to-one mappings. That is, a single question that builds upon your knowledge of already-learned simpler concepts by combining them together into a compound problem.
In the context of language learning this could be a sentence which tests multiple individual vocabulary items and grammatical patterns together. In the context of maths it could be using multiple techniques in sequence to solve a more complex equation.
I think this would bridge the gap between the atomic knowledge that flashcards give you and the more generalized knowledge that applied practice gives you.
AFAIK, no existing SRS is capable of this. I started working on a prototype web app that could do it but I haven't finished it yet. I plan to return to it at some point.
From a language learning perspective, I think this is accomplished via memorization of complete sentences. I am building an SRS as a side project where you memorize complete sentences instead of single vocabulary words and your response is graded by an LLM so that you don’t lose points when your answer is not quite correct or differs slightly in prompt meaning. is that what you meant? I have had good luck using this technique for memorization of grammar rather than memorization of vocabulary.
Dude, nice job! I've been using SuperMemo daily since January 2006! I started using it only for Japanese vocabulary words, but after a few years I tried to branch out and start using it for other stuff. It took a while to really wrap my head around using SRS to learn bigger concepts outside of learning languages. To make a long story short, it all comes down to breaking down any concept or idea into single sentence bites that are added to your flashcard collection. Very often complicated concepts require many flashcards, and (to me) THIS is where learning truly happens. Looking at an idea and saying "what makes this idea work?" and making sure those ideas are covered adequately in your flashcard collection. Now as I listen or read something, I'm constantly asking myself "Is this idea covered in my SuperMemo collection?" And if not, I write down single sentences in my Notes app and add them the next morning after I finish my repetitions. (When do you do your reps? First thing in the morning or throughout the day?) I was annoyed when I was going down this path because there were little to no resources for using SRS to learn complex information (Outside of the SuperMemo web site), so a lot of this I had to learn by trial and error. Let me know if you want to chat, so cool to find others that are dedicated enough to go 10+ years with SRS!
> Like lectures, books have no carefully-considered cognitive model at their foundation, but the medium does have an implicit model. And like lectures, that model is transmissionism. Sequences of words in sequences of lines in sequences of pages, the form of a book suggests people absorb knowledge by reading sentences. In caricature: “The author describes an idea in words on the page; the reader reads the words; then the reader understands the idea. When the reader reaches the last page, they’ve finished the book.” Of course, most authors don’t believe that people learn things this way, but because the medium makes the assumption invisible, it’s hard to question.
> Readers must learn specific reflective strategies. “What questions should I be asking? How should I summarize what I’m reading?” Readers must run their own feedback loops. “Did I understand that? Should I re-read it? Consult another text?” Readers must understand their own cognition. “What does it feel like to understand something? Where are my blind spots?”
> These skills fall into a bucket which learning science calls “metacognition.” The experimental evidence suggests that it’s challenging to learn these types of skills, and that many adults lack them.Baker, L. (1989). Metacognition, comprehension monitoring, and the adult reader. Educational Psychology Review, 1(1), 3–38. Worse, even if readers know how to do all these things, the process is quite taxing. Readers must juggle both the content of the book and also all these meta-questions. People particularly struggle to multitask like this when the content is unfamiliar.
> My collaborator Michael Nielsen and I made an initial attempt with Quantum Country, a “book” on quantum computation. But reading this “book” doesn’t look like reading any other book. The explanatory text is tightly woven with brief interactive review sessions, meant to exploit the ideas we just introduced. Reading Quantum Country means reading a few minutes of text, then quickly testing your memory about everything you’ve just read, then reading for a few more minutes, or perhaps scrolling back to reread certain details, and so on.
A far more practical idea is to internalize the practice of 'metacognition' so you can apply it to the vast corpus of books that already exist.
It is also interesting to consider previous attempts to wrap a cognitive model around the physical medium of the book, like the monastic 'lectio divina' of the middle ages. https://en.wikipedia.org/wiki/Lectio_Divina
Books don't work? I beg to differ. Books are how I've learned 80% of what I know. They work better for me than any other method I've experienced. However, I'd agree that textbooks don't work. At least, they don't work for me.
I like books, but I basically agree with the idea that the traditional front-to-back text narrative format is hugely inefficient. I think a lot of people interpret technological developments as some of natural process, and therefore the omnipresence and long existence of books implies that they are a great learning method.
In reality, though, the format of a book itself is probably just more "evolutionarily fit" compared to alternatives. If a device that could record and replay audio was somehow invented prior to the printing press, it seems highly likely to me that we'd live in a audio-based world, not a text one. After all, societies spent thousands of years using only spoken language, with writing as a fairly recent phenomenon.
A more efficient book system, IMO, would incorporate spaced repetition more directly, whether by highlighting the key points at the bottom of each page, or even by deliberately including "reminder" paragraphs throughout the book.
I typically have a low opinion on rote memorization, I remember years ago at university my roommate swore by his note cards (physical index cards at the time) while I strongly maintained the need to build up intuition around the math and physics I was doing. He was in med school and I was in engineering, but I felt that the memorization while successful at least for him, was limiting in terms of potential innovation.
I was listening to a podcast with someone who went from BS/Mechanical Engineering => MD. He had to give up his "first principles" style of engineering-thinking and just accept that rote memorization was the way to go, at least for the majority of his courses.
I disagree that rote memorization limits innovation in the medical field. Biology has complex problems, and "For every complex problem there is an answer that is clear, simple, and wrong". There's a reason why so many clinical trials fail, and it's not for a lack of technical innovation.
I feel like the medical _system_ as a whole needs systematic/organizational disruption, not technological disruption. For example, CGMs are a technology that have been around for decades and only now are we seeing usage in non-diabetic populations. The system (they require a prescription) is preventing the innovation.
Yeah med students seem to be unique in the amount of stuff they need to remember, at least in terms of users Ive worked with. The amount of things medical students need to memorize to pass their USMLE exams is insane. I designed and built a learning tool at a startup for med students using picture mnenomics and spaced repetition quizzing that seriously helped those students and lead to the startup being acquired.
I'm not the one you're asking, but I surprised you don't think that mnemonic isn't used in other domains.
It typically doesn't look like a fully fleshed out pizza shop, but I distinctly remember explaining an cell scaffold that I had synthesized like a bowl of spaghetti and meatballs to some of the undergraduates I was mentoring..
I never felt that other domains didn't address the natural inclination to use abstraction as a way to convey concepts/processes.
but I'm _surprised_ that complex metaphorical images like the one in the parent haven't spread to other domains. As you said, we often make verbal metaphors when teaching, but explicit visual metaphors are _relatively_ rare. I'm talking about literal political cartoon as mnemonic device. How often do you see visual metaphors in textbooks? They're usually more interested in correctness.
Perhaps AI will make generating these images easier, if the difficulty in creating the graphics was the sticking point. Or maybe the problem is in sharing them, which is a problem that I'm working on.
I know everyone always just responds to the title, but do read/listen to this one. I think it's Dwarkesh's best episode ever. Dwarkesh always asks great questions, and Matuschak has this amazing knack of interpreting those questions in the most interesting ways, while constantly challenging their assumptions.
I'm a huge fan of Andy's work and have been for years. Absolutely loved this episode.
If I may suggest another one of Dwarkesh's episodes that I thoroughly enjoyed: his 2 episode interview of Carl Shulman. It is an absolute treat. Completely different subjects, of course, but equally engaging.
Since you’re here Is there anywhere to get the software you are using in the video for reading/making flashcards? It looks amazing. Please tell me you have it in a GitHub repo somewhere.
I had the opposite experience. If something's easy to remember, I don't need Anki. If something's hard to remember (I really struggle with birthdays, for example), even Anki doesn't help! The cards all end up as "leech" and Anki stops showing them to me. (I found this behaviour amusing, since the whole point of using Anki is to help me learn things that are difficult to remember!)
They say if a card is ending up as a leech, it's a "problem" card and you should somehow improve it. But I'm not sure how to improve this: Side A: name. Side B: birthday. (Cards are also shown reversed.)
Arguably I could add some kind of mnemonics, or learn the Mnemonic Major System for remembering numbers. (On that note, does anyone have specific advice for remembering birthdays? The Dutch among you will say "hang a birthday calendar in your toilet", which was probably the correct solution all along...)
Give yourself more opportunities to latch onto the card.
Instead of Name -> Birthday, try "X's birthday is in {month}", or "Y's birthday is the {N}th of January" or "Z's birthday is just before {some notable date, like Christmas}" or "A's birthday is just after {B ... someone whos birthday you remember}". e.g. I remember my best friend's wife's birthday because it occurs shortly after mine. I remember a girl who I used to live across from 25 years ago because her birthday was just before mine.
This way, you have multiple different ways of relating to the information. If the related key date is coming up, you can think "oh...Z's birthday is around now"...which should be enough for you to functionally act on it.
Try to avoid the mindset of single cards. If you _really_ want the information, break it down in different ways.
Wrestling with the information in this way is a good way to find patterns and relations to it in the first place (put the information IN your brain well)...then use spaced repetition to practice getting the information OUT of your brain well.
> If something's easy to remember, I don't need Anki.
That probably depends on what and how much you are trying to memorize with Anki. It is easy to learn a few words in a foreign language, but try to learn three to five new words per day for a year straight and suddenly you realize how far spaced repetition can bring you.
For leeches, what helps is to create more cards with the same concept (and do not suspend the leech, you are right that is not a good idea). If I struggle learning a word, I create more sentences using that word, or connect it with synonyms and antonyms. For birthdays, maybe try to add some memories related to that birthday: some particular gift, a happy moment during that day, the weather or location where you celebrated... Just see what sticks :)
Why would you need to learn birthdays (except maybe closest family)? It's maybe the canonical thing to put into a calendar.
"Because I want to" is obviously fine, but it still reminds me of the infamous "capitals of the world" flashcards that spaced repetition practitioners warn against.
It wasn't until I discovered Anki a few years ago that I finally was able to remember my parents' and sisters' birthdays (despite calendar reminders every year).
Why remember something that can be looked up? Latency; it's super useful to have things appear in your brain so you can immediately do something useful with them instead of slowing down and looking something up and potentially losing context.
So, Anki did help me learn a few of them, while a few would just never stick for some reason. Amusingly, the few birthdays I did learn didn't really help me because even though I knew the birthdays, they wouldn't come to mind until days or weeks after the birthday passed!
So what I'd actually need is "what are all the birthdays that occur this month", and then have a ritual habit installed that at the first day of the month I would spend a few minutes pondering this list in my mind.
(Again, the birthday wall calendar would probably work better here...)
For me specifically the motivation was that my mother, who is naturally very good with dates and birthdays, would keep telling me "It's your {uncle, grandma, cousin}'s birthday, don't forget!" and I would think to myself, "what kind of bad person am I, that I don't even know my own family's birthdays?"
Similarly, I struggle with understanding/remembering the structure of my family tree, at least for the family members that I never met (I am often told stories about ancestors, and am shamed for my inability/unwillingness to naturally and automatically memorize them). So I broke the family tree up into individual relations and made Anki cards of them.
That worked a little better than the birthdays, but still a struggle, because it is "meaningless information" as far as my own brain is concerned. I might consciously think it's important to have knowledge of your ancestry, but my neurology apparently disagrees... Alas!
Don't bother. Do you want your distant relatives to toil away trying to remember where you fit in the family tree? Life is too short for that. Forgive yourself and move on.
I'm bad at it too and my wife is not, so she can figure out my own family tree better than I can; it makes for good party banter.
Why are spaced repetition "practitioners" warning against that?
I have a "Roman Emperors" Anki deck. It serves absolutely no purpose except making me feel good about it, but it works well (except maybe for the final few emperors, those tend to be leeches).
If you're truly interested in Roman Emperors, that's fine, of course. The warning is more about "I just started spaced repetition, I need cards quick, so let's do capitals, US states, and other random trivia".
I admit birthdays of acquaintances don't quite fit under "random trivia", but I'll still maintain that learning random birthdays seems useless, when the calendar works so well.
The difference here is important. I see the birthday calendar (well, my Dutch housemate's one) every time I sit down on the toilet. As for non-wall calendars (both digital and physical), stuff goes into them but it never comes out!
(Though the digital one does become useful due to the automatic reminders.)
I did consider this, especially since the few that I was able to memorize didn't help me remember the birthday on time. So perhaps a more useful association would be:
Pimsleur is my favorite system for learning foreign languages, based on spaced repetition.
I'm a casual tourist, who likes to pick up some rudimentary language skills before travel. There are many of us; we understand the importance of smiling at our hosts and getting the words out while more formal learners are stalled in self-induced aphasia.
For us, the design of spaced-repetition systems misses a crucial point: We don't care about measurable process while stockpiling some language skills before a trip, and we certainly don't have time to sit around studying once we're on a trip. The goal is to ingest as much content as we can, in some form that sticks, to recall more flexibly as we struggle to speak in-country.
For this purpose, memorizing a radio play as if one is an actor is more effective than spaced repetition. For example, just the Pimsleur dialogs, edited from audio files one owns. The words are there, slow to access but quicker than opening up a phrase book. Within a week traveling, one's mind reorganizes access to the language one actually uses.
The dig on learning a radio play is one doesn't know the words out of sequence. Yeah, that's why one can learn them so efficiently. For those of us of a certain age, it's the "next song on the album" effect. What did Knuth say about premature optimization? Learn the words one actually needs out of sequence, on the ground, in use.
I was wondering about the usefulness of spaced repetition for language (both natural and programming). I thought of analyzing a language for how frequently tokens appear in it (or vocabulary for a human language) and then prioritizing those cards first, and making them appear more frequently... then I realized, reality already does that for you, automatically!
(On the other hand, Pimsleur is great, and taking a systematic approach to "100%ing" the "core" of a language is probably a great idea.)
Depends on the kind of book and subject matter. Dwarkesh is right about books you need to read all or nearly all of to get the benefit. But to answer "Do people care about understanding?" for a lot of books: No. The reader cares about their problems and solving them.
Based on reader feedback, only a special kind of reader reads a programming book cover to cover. 90%+ people who read substantial parts of a programming book, dip into the parts that interest them after, perhaps, reading the first few chapters to figure out how to navigate the book.
Online, you have the freedom to do better and to publish similar material in multiple modes, from passive and linear, to self-paced interactive to human-guided. There isn't one answer, which the post touches on in several places including regarding spaced repetition.
LLMs are likely to have an impact on the value of pedagogical techniques like spaced repetition because spaced repetition is a tool for overcoming human impediments to acquiring and navigating huge, difficult, knowledge bases. Then using that knowledge requires enormous effort and aptitude. That's where humans competing with LLMs are like pick-and-shovel miners competing with a giant dragline at mining coal.
After watching the studying video I'm surprised there is no direct GPT-4 integration with his tool. There's a clear opportunity for dynamic question generation such that you have notes/cards which can't be memorized but instead ask novel questions around the same point.
Books work—good books anyway—, but most of us aren’t taught how to read well.
The next time you want to get something from a book do some new things with it.
1. Skim it first.
2. Find some interesting entries in the index if it has one and jump there to check it out.
3. Ask questions while you read. Write them down the book.
4. Reread the book (spaced repetition). It will go faster because you e built a structure with your notes the first time.
5. Don’t waste time on bad books.
A good analogy is pitching. The pitcher is the writer, the ball is the knowledge and you are the catcher. Nothing works if the pitcher can’t throw or if the catcher can’t catch. Likewise if there’s no ball to throw.
So it is you’ll get nothing out of the book if the writer can’t deliver or has nothing to say. But a well crafted book stuffed with knowledge won’t to any good if you receive it.
In math textbooks, the exercises contain much of the exposition. Some is in the answers. And some of facts you are expected to learn only appear in your own workings.
I wrote about a more relaxed version of this, some years ago. See Secrets of a Buccaneer-Scholar.
I wrote it to explain the techniques I use to compete as a technical guy who dropped out of high school.
It’s actually not very hard to compete if you see education as an ongoing process of personal construction, rather than some plug of knowledge that an expert gave you (along with a certificate).
I collect books and papers which I ignore until I have a relevant problem, at which point I go into a process of intensive inquiry similar to that describe in the post.
The context you appear to be missing is actually the content of the post. It's often the way with articles: they give a title, which is then expanded upon in the main text.
You're wrong. [video] is only an (unnecessary) warning for people who don't like videos. If most of the page is text there is no reason at all to "warn" anyone.
The HN guidelines don't say anything about transcripts: "If you submit a video or pdf, please warn us by appending [video] or [pdf] to the title." https://news.ycombinator.com/newsguidelines.html
Shameless plug: I'm building https://python.cards to learn Python with spaced repetition. I've uploaded some videos on Youtube of the process of distilling the information from the Python docs into flashcards, and have plans on trying out LLMs to help in this process.
As someone also building an SRS product, why did you decide on Python as your initial market? I genuinely feel like programming is best learned by _doing_, and thankfully it's pretty easy to practice writing Python. I feel like a better way to learn Python is to introduce a concept/word, and give a playground where the student can play with it.
Many reasons, most of them hypothesis. For one, the book "The programmer's brain" says that spaced repetition is the best way to learn the syntax, idioms, caveats of a programming language, but it's hard to find a product that offers this.
I don't suggest you should learn Python _only_ with python.cards. I assume every python.cards user will be doing some online course, or working with Python, and my decks will just boost their knowledge of the language.
I think a lot of people program in Python but programming in Python is not their main occupation. These people struggle to accumulate knowledge, because they may encounter concepts or APIs only from time to time, and by that time they already forgot.
Also, I want to cover some "unorthodox" topics for which spaced repetition may be specially well suited. For example, I'm building a deck called "A tour of the standard library", where you can learn all the modules that are available, just so you know what is out there, without going deep into it.
Another idea is to have decks to learn all the built-in exceptions by heart, or the nomenclature. I think this helps build a mental model of the width and depth of the language so you can better integrate all the knowledge you find while coding, searching online, or reading code.
> the book "The programmer's brain" says that spaced repetition is the best way to learn the syntax, idioms, caveats of a programming language
Unfortunately, hard disagree. The best way to learn a Lisp is to write Lisp. You can look at code all you want, but you won't get used to the syntax/parens until you start writing it. Same applies to non-Lisps.
> I think a lot of people program in Python but programming in Python is not their main occupation. These people struggle to accumulate knowledge, because they may encounter concepts or APIs only from time to time, and by that time they already forgot.
This is a great point. Static typing is great for API discovery, which Python sadly lacks. However, I wonder how much LLMs are eating this "low-code" market. Also, if Python is non-critical to their jobs, I suspect that they won't feel motivated to explicitly study Python over the long term. Doing spaced repetition is famously hard, and non-motivated people will likely give up and just do "JIT learning".
> Also, I want to cover some "unorthodox" topics for which spaced repetition may be specially well suited. For example, I'm building a deck called "A tour of the standard library", where you can learn all the modules that are available, just so you know what is out there, without going deep into it.
FWIW I'm learning Rust, and to "learn what's out there" I've just been binging random Rust videos on Youtube at 2x over lunch. Granted I'm an experienced dev and I have an intuition as to what "should" be out there, so perhaps this story is of limited use.
Sadly, hard disagree. I programmed Clojure professionally for ~2 years, and its REPL is significantly better than Python's. Despite that, nothing compares to "dot" autocomplete and being certain that a method can take a certain object type. Like, I could type dir/help and then the token in question then parse the output with my eyes, remove the help/dir and type in what I wanted... or I can hit dot and scroll the available methods/properties. It's not even close.
The book itself is a way to open the door to learning, not the end point. Reading subject titles first, cross-referencing, and working examples and doing projects help teach and cement ideas.
There's techniques like SQ4R [1] for reading, and hugely applicable to general technical and article reading is "How to read a paper" [2]. For lecturers, there's "How to Speak" [3], and for those attending a talk there's the Cornell Notes system. [4]
[1]: https://en.wikipedia.org/wiki/SQ3R
[2]: https://web.stanford.edu/class/ee384m/Handouts/HowtoReadPape...
[3]: https://www.youtube.com/watch?v=Unzc731iCUY
[4]: https://en.wikipedia.org/wiki/Cornell_Notes