[llamaz]

I have tried this, and tt's a good pedagogical method for learning-how-to-learn, but it's inconvenient in practice

I think it's worth learning how to use Anki instead. The key challenge there is selecting the right level of coarseness, so that it doesn't take forever. Even if you don't end up reviewing your cards, the act of synthesising information helps you learn it. Down the line you're able to encode important information in finer detail.

Contrary to what you might think, this method is especially useful for really difficult concepts (e.g. in math and physics). Sometimes you sit through a lecture and have no idea what's going on until you start doing the exercises. In these cases you might have to use memory as a crutch, to try to memorize the main ideas (or useful facts) to create the scaffold which you fill in later.

[SpaceManNabs]

I used Anki for my ML grad (wish I did for undergrad). Formed a card for each subproblem of a problem in a pset, theorem, lemma, general concept, and definition---and for each of these things, using the 20 rules of learning from supermemo (? i think). I also used it for easily grokked things that I wanted to make a habit. Made a deck for each class, and also made a deck for talks and seminars I visited, grouped by semester.

Super useful. I feel like I don't really forget the majority of the vibes like I did for my astro undergrad. I remember GR a lot less than convex opt even though I spent many more hours of rigorous study and even though I revisited my GR material more. Revisiting my astro notes is still effective for those little details, but getting all the context into my working memory is much more exhausting.

[ravenstine]

They tried to get us to take Cornell notes back in high school. I think they can work for some people, but I found that it created too much mental overhead. Seemed to turn me into more of a recording device than a mind comprehending a lecture. The note-taking style seems to assume a certain format of lecture, and if a teacher deviates from that then it can be more confusing than simply taking notes without the extra columns and cells.

That said, I think it's a good idea, but might not be for everyone. Kind of like how there are different apps like Notion, Trello, Quip, OneNote, etc., and while Notion works for a lot of people, some people benefit more from Trello's boards and cards.

[jerf]

In grad school for computer science, after some struggle, I eventually found that the best thing for me was just to not to take notes at all. Just engage with the class maximally, and make sure to do the homework relatively soon after that (maximizing the time between when you cover something in class and practice it with homework, aka, "waiting until the last minute", is a bad idea for all sorts of reasons).

Obviously, my plan doesn't work for everyone, or every class (this is a bad idea for fact-dump classes for me). But certainly after this I tended to look at everyone's list of "things you have to do to succeed in class" more as a menu than a proscription. Still do look at a lot of things that way; you can see it in our industry too, where you can find people telling you just have to use this type of type system or that type of database... yes, thank you for adding to my menu, but I'll examine that for myself, thanks.

[Bartweiss]

I've had this problem with every form of structured note-taking I've ever tried. My focus is redirected to writing everything down and ensuring I'm using the appropriate format like a stenographer, so badly that when the professor asks "Any questions? Does everyone understand this?" I don't know whether I understand - I haven't engaged my brain yet.

Even worse, I have to start wrestling with the format as soon as anything unusual happens. If I miss something, or the professor makes a mistake, or someone asks an important question clarifying that bit 10 lines ago, or the structure of the information isn't linear, I want to be able to bounce around my notes connecting and fixing things in a way that I'll understand later.

Structured systems like Cornell notes work well enough in a slow, perfectly sequential lecture, but everything works there. And I imagine they might probably be good for memorizing completely synthetic content like the rules of an unfamiliar game, where there's not much thought to apply. But for practical notetaking, I think their best use is after the lecture, as a way to convert "get the content down" notes into a study-friendly format.

[sukilot]

Thorough note-taking from a live stream is only appropriate when it's impossible to record.

In school, thorough note-taking during lecture is a terrible idea. Only jot down your own questions to follow up on later; don't copy the material. Record the material and replay it later to review.

Note-talking "offline", at your own pace, from a book or a recording, is a great way to build your memory of the material.

[Bartweiss]

It seems like most things taught as note-taking methods are actually study methods. Obviously, the hope is to combine the two and automatically create a study tool, but I think that's usually a mistake.

In practice, the combination seems both difficult and ineffective. Most lectures are not perfectly clear and sequential - they have mistakes, backtracking, and questions. And much lecture content is not linear, especially in the sciences, so you get diagrams that don't naturally break into "read cue, reveal content". Trying to force the lecture into the note format distracts from comprehension, and produces notes that are good for retention but not for learning content you didn't understand. Even if you get past all that, there's a fundamental problem with trying to organize your study material before you know what's on the next slide, and trying to organize questions after writing down the answers.

Anki or any other after-the-fact study format lets you cut past all that. You can take notes fluidly, aiming at low-distraction completeness. And then you can rearrange the content for retention, not lecture-compatibility. Charts and non-linear content can be presented smoothly. Best of all, you can actually study with cues and content in a many-to-many relationship, linking dense info like a ternary diagram to many different prompts.

[kd5bjo]

In many ways, it comes down to the standard writing adage of "know your audience". In the case of notes, that's generally a future version of yourself; knowing how you plan to use the notes in the future will inform their structure now.

My system is constantly evolving as I discover what works for me and not, but it's in a pretty good place right now. Information gets processed in several distinct stages. I start with giving the learning material my undivided attention, whether it's a book, lecture, video, or museum. Shortly afterwards, I record my thoughts and observations in a journal as unstructured prose. This is usually about 2/3 recording things that were presented and 1/3 random connections that came to mind. The important thing is only to record what I actually understand -- if there was a lot of stuff that went over my head, it's a sign that I need to revisit the source material later after I have a better grasp of the fundamentals.

During a weekly review session, I read through all of the journal entries and index them by subject in a physical card file. Each card is headed with the subject, obviously, and the card body has a reference to the journal entry, the source material, and a 1-2 sentences summary about how the journal entry relates to the subject.

At the same time, I also make flash cards to add to my Leitner box (which serves the same purpose as Anki). My goal with these isn't to remember everything, but to keep enough of the subject fresh in my mind that I'll be able to think of the correct subject headings when I want to look information up in the future.

I use physical cards for these because I feel like they're a better serendipity engine than anything electronic I've tried. Whenever you're interacting with the system, you'll end up glancing at a bunch of arbitrary cards unrelated to your current task, and occasionally that'll be the trigger you need for a new idea.

One form of notes that I want to add to this is the long-term reminder. The idea comes from Chris Hadfield¹ and presumably others in the astronaut corps: once you've done the work figuring out how to reduce theory to practice, make a task-focused reference for it that gets filed away if you ever need to do that task again. The goal here is to directly record conclusions: if you are annotating a diagram of a machine, for example, indicator lights should be labelled with the corrective action to take in addition to the parameters that cause them to illuminate.

¹ Masterclass.com; Lesson #17

[_jal]

I tend to think that, so long as one isn't doing something really odd, most of the value comes from practicing any system at all.

The important thing is the repetition/correction cycle.

[ar_lan]

It sounds like you are implying spaced repetition and the Cornell Note-Taking System are mutually exclusive.

I personally don't like the flow of Cornell's style, so I didn't use it as a student, but SRS software (Anki is still my preferred) didn't replace me needing to take notes and review those after class.

If anything, I would expect your Cornell notes to be a good basis for how you create some of your cards.