Spaced repetition can allow for infinite recall (2022)
efavdb.com
efavdb.com
Interesting fact, SuperMemo is written in Pascal.
Are there any public sources about SM being written in Pascal? TIL
Be aware that it's not peer reviewed, and some of the findings take a lot of experimentation if you want to dive deep and reproduce them.
My understanding is that his thoughts on sleep are pretty controversial within the sleep optimization community.
But the self reporting aspects of the “data” always made me suspicious. People in certain phases of sleep dep, hypoxia, or a number of other debilitating events don’t know they’re experiencing symptoms until someone else tells them they are. That’s why you’re trained to look for signs in others.
So I am sorry but I don’t trust you if you tell me you feel great on four hours of sleep a day. Let’s do a task or a test and see if you can brain or you’re just euphoric.
I haven't used Supermemo, but believe it has fairly strong sleep tracking capabilities.
I've dabbled about doing the same in AnkiDroid. We could pull data from Android Health, and do something interesting with it, but we've got hundreds of other things to do, Google will probably pull our app from the Play Store if we request health data, and it at the moment it's a little out-of-scope and would be better for an addon using our API
You have to be careful and spread your tests over a bunch of subjects to find the patterns.
I wish to see a FOSS incremental reading application in the very least, such an underrated concept.
Be the change you wish to see in the world
There's an Anki plugin[0] though I know almost nothing about Anki or incremental reading, so I have no idea who good it is.
It's a fairly opinionated space, which is surprising since there is good data. But regardless, SuperMemo, FSRS, both work, you will make progress with either. People can just pick either and get started.
https://github.com/open-spaced-repetition/free-spaced-repeti...
TL;DR: SRS is, perhaps counterintuitively, most useful for those of us seeking true mastery over something.
Counterpoint: a noggin full of facts does not an expert make. Makes a good cocktail party trick, or Jeopardy champion. Expert in a subject? Definitely not.
Except for the regex syntax, and ffmpeg pipelines, I won’t ever dedicate brain-space for those.
So I suppose I contradict myself. Huh. Nuts.
so, you might want to google it, because some part in the back of your head itches to remember something that is actually relevant and located on the rim (of the knowledge graph) of something that you do remember.
fucked up indexing. an 8 hour rabbit hole and that one of 43 links you didn't open in a new tab.
I find it also true for books, you unconsciously internalize things by reading the full book VS reading a summary. I sometimes reread books I read 10+ years ago and I often have these "ah so that's where I got this idea from". Most of the time it isn't even about the main point of the book, it can be a sentence or a dialogue that left an imprint on your psyche
I'd hate to be a simple proxy between google/llms and the real world
Silly.
The search is vague, and the result is nondeterministic, but in a lot of cases it's still the better method.
https://fchollet.substack.com/i/137628402/llms-as-program-da... I like the way this article puts it. "LLM is a continuous, interpolative kind of database"
That funny way to spell "hallucinated"
For example, neither google nor an LLM replaces the utility of building your own vocabulary so that you can express yourself in a foreign language when you're in convo with someone.
I imagine spaced repetition is not actually a cause of repeating yourself. That's just caused by being too lazy to proofread before you post.
It's akin to seeing 37*55 and knowing it won't be a 20 digit number, greatly reducing your search space. Imagine feeding someone without knowing what they like (how do you choose something?) vs. "allergic to dairy"?
[1] https://www.jsoftware.com/papers/tot.htm [2] https://alexalejandre.com/programming/forge-deeper-thought/
I had some problems with their documentation though. Following their examples was pretty straightforward, but I always had trouble e.g. what is the difference between State.Learning and State.Relearning, or what does fuzzy mean in FSRS parameters, or reps on a Card. You have to either read the algorithm explainer or the Anki source code to understand that.
I complain here, but this is open source and if I was slightly more ambitious I would submit PR with the missing docs (or at least an issue).
PS. my work is here https://github.com/runarberg/shodoku in case you are interested, but the source code bears the mark of being a weekend hobby project, so read at own risk ;)
I'd like to port my own flashcards app, Manabi from SM2 to FSRS so this is helpful
I made this companion reader app as well, btw, for learning Japanese: https://reader.manabi.io
This library also makes it easy to undo a review (one of my missing features) as long as you keep the review records (which this library also makes easy).
> If you insist on paying for this app, consider donating to UNWRA instead.
Glad to see solidarity among the Japanese learning community.
It is great to have the docs well organized like this, what is missing is just a short explanation of what all those mean. The information is available, e.g. in the README or in the source code. So honestly it is just a minor annoyance, rather than a major issue. I was able to use ts-fsrs just fine actually.
The Anki database format is still pretty opaque, despite a few efforts over the years, and it would be great to open some of this this up to allow apps to export apkgs
It doesn't help that our wiki is noindexed by GitHub: https://github.com/ankidroid/Anki-Android/wiki/Database-Stru...
I agree that our documentation is pretty bad at the moment haha. But I'm hoping we'll start cleaning things up more this year. I'd also like to potentially move away from using github's wiki system...
There are 2 big problems:
1. If you are studying seriously, you will encounter many words outside your SRS practice, so the FSRS prediction fails flat.
2. More importantly, language learning is not about recalling words.
The algorithm I use for my app has a different set of objectives, it's <30 lines of straightforward code, and it works better (I tested it for months to be sure about it)
I'll open source it soon
It takes a text, and creates a bunch of cloze deletion cards, each presenting you with a number of lines, and asking you to recall the next line.
https://super-memory.com/help/read.htm
Disclaimer: AnkiDroid maintainer
Still got a long way to go for me to be happy, but we make great progress every year
https://forums.ankiweb.net/t/ankidroid-2024-roundup-2-20-1-c...
If you don't want to upgrade to FSRS
* Export a colpkg
* File -> Switch Profile -> New
* Import
* Deck Options -> Enable FSRS & Optimize
* FSRS Simulator (experimental)
My design principal is that I don’t need to be limited to few words, a sentence, and maybe a picture for each flash card. Instead each flashcard can be the all the relevant parts of the whole dictionary related to the kanji that you are learning.
Another design principal (where I deviate from Wankikani) is that drawing the kanji helps remembering it. So my app has two sided cards, one for drawing and one for reading. Both reading and writing practice include every word in the dictionary + example sentences that contains that kanji.
My third design principal is that I can learn vocabulary at the same time as I learn the meaning and the writing of the kanji (in fact, I like to learn a few vocab words instead of memorizing the different readings). So you can bookmark vocab words you want to learn along the kanji, and the next time the SRS system picks the kanji up for review, these words pop up at the top, to help you recall.
Note that the app is still in development, so use at your own risk, although I am actively using it for my kanji-studies. I still have a couple of features missing before I can call it done enough.
I've checked your website and HN profile but there's no email to contact you, so here are some simple suggestions: when you enter a deck, I would suggest you add 2 shortcuts or buttons: one to randomly jump from the present card to another one of the same deck, and the other to randomly jump to another deck in the review set
This is because when faced with a sequence, the brain learns the order the element in the set, which may affect recall.
Being able to randomize the set helps prevent that, and give better results.
I think you were browsing the cards and using the quick navigation to go to another card in the same deck. When you actually review the cards using the SRS system to review the cards (the big “Review” button), there is a setting (in the settings view) to enable fuzz (a small random delay; defaults to on) which will effectively scramble the cards for your next review. I actually had been using it for a couple of months without fuzz and just random noise in my retention abilities were enough to scramble the decks, at least in my experience.
But a button in the quick navigation bar when browsing the cards to jump to a random card is not a bad idea.
Yes, I was just looking at how it works because I liked how you presented your design principles.
My personal design principle is to minimize friction and allow immersion to concentrate on the task and not on the navigation.
> I actually had been using it for a couple of months without fuzz and just random noise in my retention abilities were enough to scramble the decks, at least in my experience.
I don't know about you, but when using cards or other repetition-based systems, I've found my recall is strongly increased by randomization: if you can "predict" which card you will be shown next (and I often did), you are not learning the content of the card, but their sequence. This is helpful for some tasks where order matters, but it's often an hinderance for things that don't have a natural order
> But a button in the quick navigation bar when browsing the cards to jump to a random card is not a bad idea.
A button is nice, but since your content doesn't fully fit on the screen, you should also consider keyboard shortcuts: being able to immerse in the viewing helps a lot for memorizing (cutting distractions)
After a while, I guess you must have nailed the common part (the drawing) while still benefiting from the varying part (the list of words): if you can have that on the screen more easily, without much friction (like scrolling down past the drawing), you will be able to immerse into that. Exposing the drawing area could be a setting, or more simply you could have shortcuts or buttons to jump to the word section.
Actually, maybe you should also get a button and a shortcut to randomize the words themselves: to take a quick example, https://shodoku.app/kanji/%E5%85%88 lists 104 words.
I think you know the #1 better than say the #103, because you have seen the top of the list more. If they are sorted by frequency (ie how common they are), that could be good when begging, but after a while you may want to learn the less common one, and having to scroll down may reduce your desire to do so.
I tested jumping to the bottom of the page with the End key, and I think you have a "load-on-demand", so you should be able to test my hypothesis quickly: look at the web server stats for the chunks of a given page for your own IP, and plot the frequencies: I think the bottom where the #103 word is will be seen far less often that the top where the #1 word is.
You have made a very cool project, and I like your design principles, so I hope you won't find my suggestions offensive or disrepectful.
Thanks, I appreciate it ;)
> Keyboard shortcuts
It is on my missing feature list.
Another missing feature is next or previous stroke buttons (and keyboard shortcuts). I want to be able to use this flawlessly on a device without a touch screen, where you simply draw the kanji on a peace of paper (or even your palm). Users that use it like that should be able to benefit from stroke hints as well.
On the vocab list, the purpose is to give you some example words which you can use to learn the reading. I usually bookmark 1-3 words in my first writing practice, which will then appear at the top of the vocab list for later reviews, I will than try to learn along with the kanji. Another missing feature is integration with Anki proper. I would like to quickly create a vocab card for that word and add it to my Anki deck (I would rather not use this app for vocab practice). My current ordering of the vocab words is kind of a mess, I do some JLPT word-list prioritization. Having the most basic words on top. randomizing could be a good feature here. Another could be alternate sorting for the vocab words, or even a search bar.
I actually just host this on github pages, so I don’t have access to access logs (or I don‘t know how to access them). In the near future this is going to be a progressive web app with a hard cache of all the assets (which I generate into static assets from an sqlite3 database on my own machine) so you could actually use this offline with if you previously browsed the kanji.
Sorry to nitpick: Free Spaced Repetition Scheduler
Even without the confusion of toggling a hidden detail, the or-or-off toggle button is known to be bad UX as it is unclear whether the toggler is on or off, unless clarified with a label. https://news.ycombinator.com/item?id=39342649
My next feature is going to be custom decks, which is implemented on this page. I will for sure fix this (by adding labels, and changing the icon to a checkmark for active decks).
Thanks again, this was helpful.
Spaced repetition can allow for infinite recall - https://news.ycombinator.com/item?id=32381206 - Aug 2022 (169 comments)
I'm now working on flashcard auto-reviews while reading, so that if you read words/kanji that you have cards for, it'll automatically mark those cards as reviewed rather than ask you to recall them shortly after you just had to recall during your reading.
I hope this helps people replace more SRS time with actual reading (or listening)
Either find a way to make it less boring or switch up the approach. Whatever gets you to keep a habit is what's important.
Same with any habit we want to develop. Cooking. Exercise. Language learning. Building software. Meet people.
Even then, personally I don’t find Anki that boring at all. But I make custom cards that include audio, images, and other items.
You: But I make custom cards that include audio, images, and other items.
I think you're agreeing!
This obviously isn't the case if there's an alternative approach that does work for you.
> Even then, personally I don’t find Anki that boring at all.
Right, and you're having trouble picturing how something you get for free (or take for granted) has a cost for others. Isn't it obvious?
Running 10km a day is a great way to stay in cardiovascular shape. If you like running, it might be a great daily habit for you. But would you say other people must not care about their body just because they're not willing to run 10km, but they are willing to bicycle, swim, or kickbox daily?
So the point is that while SRS might be boring at times compared to other types of learning, the benefits are worth it. Effectiveness of learning does not correlate directly with enjoyment.
To phrase this another way: yes, in general you should find learning methods that appeal to you. But spaced repetition is so effective that it’s worth implementing, even if you don’t enjoy it.
Edit: using the metaphor that you added: if say, doing jumping jacks for 10 minutes burned 10x calories than every other exercise does in 60 minutes, and your goal is to lose weight, then yeah, it would seem a little silly to me to avoid doing jumping jacks because you don’t enjoy them. The benefits are worth the hassle.
i'm not saying spaced repititon is useless - it is a great beginners tool. However you need to move on as soon as possible. once you can use the skill that is better in gereral. Doctors use spaced repition more than most because there are a lot of rare things they will never encounter but mixed in is one rare thing they need to know instantly and they cannot know in advance what that thing is. Almost nothing else is both like that and doesn't allow time to look it up.
note that reading/doing something is itself spaced repititon. It isn't algorithemic but it is still randon spacing of things you need to know and the more you need it the more often it repeats
No, they are not more effective, they are less effective. The fact that they are less effective is the only reason that people do spaced repetition. It takes less time to learn a fact if you optimize for the way memory works. Reading, listening, and doing the thing you are interested in builds memory just like spaced repetition does, just in a haphazard, random way.
> note that reading/doing something is itself spaced repititon.
But worse. There are plenty of things that simply can't be learned by spaced repetition because nobody has come up with a good way to do them, and/or they are physical. But some facts simply have to be memorized, and when doing the thing you will have to recall them. You can either look them up every time you have to use them, and a) forget completely if there's too large a gap between incidents of having to look a thing up, or b) spend all day studying something that you'll completely forget in a week. Spaced repetition is about using a timing trick to make those facts reflexive and long-lasting.
If you know nothing, spaced repetition can get you over that initial hump well. However I maintain that you quickly reach a point where you are beyond that and then space repetition is less useful than other courses of study in general. Use well it can still be a supplement to other study methods, but those others should be the priority.
I built and ran the academic side of a language school for four years, worked on two language learning startups and also learned two languages in different language families from my native language (and from each other) to a pretty fluent level.
In all those years I’ve never met a single really successful learner who made SRS a major part of their studies, though some used it as a small supplementary practice.
The thing is that words aren’t usually discrete pieces of information in the way that names of capitals or things a med student has to memorize often are (unless your goal is just to play scrabble with them, in which case SRS is great). Meanings don’t map one to one across languages, collocations are important , etc, etc. Putting sentence cards into Anki is better than isolation words with translations but even then, you won’t get as much cultural information or even raw quantity of input as you would from extensive reading and this is a topic L2 Acquisition researchers have covered in depth.
I think part of the problem is that SRS sounds really compelling to engineers and it’s generally easy to build into an app, so that’s been the focus of most language learning apps for the past 20 years.
There are some better ones that exist mostly to help learners handle native text and audio, though. LingQ, Language Reactor and language-specific apps that do similar things are great.
I did, but very differently from how most people do it.
Premade decks with single words sucks, yes. But creating your own flashcards with lots of audio and mined sentences from your learning materials accelerates learning better than anything I tried before.
And yes, it requires a lot of immersion / exposure and other things. But when done properly it works really really well.
Or if you only learned closely related languages, what aspect of it was "really successful"? Accent? Writing ability? Something else?
It is a lot effort, I find Spaced Repetition actually very demanding – but effective, that's the point.
If you take Paul Nation's math (the famous https://files.eric.ed.gov/fulltext/EJ1044345.pdf) and instead of doing that you mine the hardest sentences you find to an SRS (basically, separating grain from chaff) you save yourself a lot of time, because you will never find a piece of text long enough where you know exactly 98% of the words.
In my case, I use flashcards primarily for reading and listening comprehension, but lately I am using it a bit more for pronunciation, and it can also work well for spelling.
This is important for three reasons. First is polysemy, which he mentioned. This is words with multiple meanings like fair (reasonable/just / light or unblemished / a type of public event with entertainment and vendors).
The second and in my opinion even larger issue is gaining and understanding of how the words are used and their scope. E.g., there's no word in Chinese that quite matches "nose" in English. The closest is 鼻子, but that can refer to an elephant trunk or a pig's snout, which "nose" can't. In some languages, there's a word that can be used for humans and pigs, but not elephants. In others, the same word also encompasses bird's beaks. The only way you'll learn this is from encountering the word in a lot of different contexts, not from drilling it repeatedly in the same context. Furthermore, there are a lot of words that tend to be used with or near each other but near others (collocations). In English, it's normal to say you're doing "pretty good" or "absolutely fantastic", but saying you're doing "absolutely good" would be very strange.
Finally, there are a lot of shared cultural stories each language community has. In English, these would be from Christianity, from classical Greco-Roman figures like Aesop, from German storytellers like the Brothers Grimm, other European storytellers like Hans Christian Anderson, etc. In Chinese (or Korean, Japanese, Vietnamese or other nearby countries) people are going to have a lot of shared stories from ancient China—philosophers from the Spring and Autumn era, historical stories and dramatizations of the interesting times in history such as the Three Kingdoms era, and many, many folk tales. If you don't understand at least the core of the cultural cannon, you'll regularly be confused by things in TV shows like soap operas or variety shows, even if you understand every word in transcript. Reading fiction will be even harder.
The more distant the language is from the one you speak natively, the more the second and especially third points will impact you unless you take in an enormous amount of input.
Your essay exposes many unconsidered ideas I can USE
Such as sorting spoken language from written.
*written* has vast Man-Page (Library!) For all languages. From parts of speech to detailed standards for each sort of publication, intended audience.
The only Man-Page (comments) I could find for spoken dialog was "Miranda Act", I better see what we have NOW.
Thanks
Yeah true, but an obvious argument is that this is where discipline comes in. If you are one of the people Anki works for, then you have to find the level of discipline required to stick with it.
I think there are diminishing returns as well by having to many daily new cards, as there is only so much you can commit into memory every day. So you will probably end up having terrible recall as well (which further adds to your reps).
I’m using Anki to learn Japanese. I actually go a step further in taking it slow as I’m still on my N5 vocab (first 1000 words) after 9 months of learning. I think it will be another 9 months before I can read stuff I‘m interested in in Japanese (which will probably be Go books).
If you don't use that, then you have to use your time dedicated to learning, where Anki is probably more efficient than your learning method anyway.
For language, 10-20 mins / day is great, and if you are putting enough effort, 6-8 reps is enough to learn a card with 2 new concepts at ~15secs / rep, which really makes a difference long term.
I made a tool[1] with that approach in mind and tailor-made for language acquisition patterns. The key part is that creating cards (audio included) must be super-easy and fast, but they need to come from you nevertheless.
You only have to "catch up" ASAP if you want to minimize the risk of forgetting any card at all, and then having to memorize it again. Otherwise, just working through the backlog at your preferred pace is the best strategy; you'll still be recalling most of the cards and the system will push repetitions for those that you do recall further out in time.
Perhaps the Anki devs should add some kind of special workflow in the UX that shows up when the user has a huge backlog to go through due to missed reviews, and tells them this. Because you're right that just seeing a bare number (863 cards to go through!) can be quite scary.
As a product though I struggle to make it work. I assume people mostly afraid of their notes and cards stored in the cloud vs local storage, which is understandable.
The first time I published a site, I was surprised by how much traffic came from mobile devices, even though my page was intended for desktop users. I really shouldn’t have been surprised, but fortunately I had some basic analytics and saw fairly quickly how bad my bounce rate was on mobile and was able to work on it a bit.
Update: made it look a little bit better.
This is the 1000th time you’ll have read the same thing. The hivemind naturally regurgitates the same information at a predictable cadence. We’re due for a few Rust posts on HN, for example.
If you stay tapped in, you will always be learning from this organic body.
Repeated exposure (e.g. re-reading a book) does something, but active recall is much more efficient at solidifying memories.
Not really. Depends on the definition of spaced repetition, but you typically want increasing intervals (normally exponentially) for it to be effective.
See 9.1: https://pcl.sitehost.iu.edu/rgoldsto/courses/dunloskyimprovi... [PDF]
Not affiliated but 3 months in and I feel my math foundations are finally solidifying into something that will allow me to build into areas previously out of reach
But I'm at about 70% of Foundations III and 15% of Stats and Probability and the amount of formulas is starting to be overwhelming.
I basically only put two things in there: "Prove X" and my ordinary problems. Then I'd just do that. Those cards usually took at least a minute apiece, but I got really good at abstract algebra, at least when compared to similarly skilled peers. Mind you, I have always been good (perfect SAT score) but not great (never scouted for Putnam) at math.
Contrary to what you might think, memorizing definitions is less useful (though not totally useless) than memorizing the proofs of theorems which rely upon those definitions.
I would almost certainly go back to doing this if I ever want for a master's degree with a math class again. The idea crosses my mind every now and then to do it with linear algebra, which counterintuitively isn't actually a prerequisite for or all that similar to abstract algebra.
Spaced repetition always seemed like those schemes to get you fit or slim in 30 days that never work. There is exactly one way to get physically healthy, and it's super-unfun: diet and exercise. Same thing with your mind, you have to exercise it and feed it appropriately for months and years. Spend the time you would spend on spaced repetition reading books or watching lectures and doing exercises instead.
And spaced repetition is a very efficient way of doing that.
Using cards for spaced repetition has always been common between top students. I would not be surprised if brilliant people used it in one way or another. Modern spaced repetition apps are just the paperless equivalent.
It's a trade-off though, I now read less and tinker less. Do I regret that, you bet. But still Anki/SRS works for me, especially because I often do it at times when I wouldn't be able to effectively read/tinker (perhaps tired, or getting kids to sleep). That's a long way of saying, do what's effective for you, but there's no point of being so dismissive of what others are doing.
Yes. Assuming the content is around your level - contains words you do not know, but not too much. Obviously, if you know every single word in the movie, you wont learn new ones. And if you understand nothing of it, you wont learn either.
> If by consuming content you meant taking a specific movie/episode and breaking it down, studying it relentlessly, and moving on to the next after you've seen it 20 times and practically have it memorized, then maybe that would've helped?
That sounds like self torture rather then anything else. Why would anyone do that? What would be the point of watching the same thing till you memorize that thing?
The value of consuming content comes from seeing many different sentences in many different contexts. So you build more connections in your brain and you eventually learn to combine own sentences. Movies specifically have value of showing something more similar to "normal speech" in speed so that your brain wont have time to translate.
Memorizing segments goes against everything it is supposed to provide.
> I can state with high certainty that simply watching a lot of movies or shows in the language I was learning did absolutely nothing to help my vocabulary compared to using anki.
For me, each time I tried anki, it seems to work short term. And it completely fails to teach me new words. Flashcards seem to be the least effective way to actual learn words that are truly new. I keep forgetting words I have learned from anki much much faster then words I learned from elsewhere.
And going over pieces of said content in some sort of structured scheduled form isn't "consuming"?
Indeed, but I think that being relevant assumes that people using these techniques believe it will somehow make them brilliant/geniuses.
> Spaced repetition always seemed like those schemes to get you fit or slim in 30 days that never work.
I'm not arguing for spaced repetition, but the whole point is you do some amount every day (or as needed) for the rest of your life (within reason). So it's not equivalent to get fit/slim quick at all, it's more about disciplined improvement of yourself. If that isn't effective for you don't do it, but based on your comment it seems like you're coming at it from the wrong angle.
Fair enough, I haven't looked into spaced repetition deeply.
I just always got this weird vibe from people talking about it, like they think memorizing all the world capitals was a way to improve your intelligence. It's not.
Same thing with, like, "memory palace" stuff: being able to memorize a deck of cards is probably fun, and it's a nice parlor trick, but it will not help you solve real-world problems, analyze situations, make discoveries or be a better engineer. The way to do that is to just read books, study, work hard in your chosen field and keep up with the latest developments. That is the only way anyone has ever gotten good at anything.
Otherwise I know the fact will be written in the sand, it won't be there for me to use at the time when it would be useful. That's terminology from a book on memory I read a while back, which ironically I've now forgotten name of because I never put it in Anki.
Also should say I used to be much more scatter gun with what I put in Anki, but these days I combine it with Obsidian which I think is more managable.
Anyway not trying to sell Anki to anyone, if you don't need it don't use it for sure. I just know (think?) it's working for me, at least compared to the alternatives.
It definitely is. I don't understand the disdain for facts. If, when talking about a city, I know that it's a capital city, I am smarter than I would be if I did not know, and I have a better ability to participate in the conversation. If I know a bunch of constants (in whatever trade I work in) offhand, I can quickly do calculations in my head. Does knowing how many centimeters are in an inch make me smarter? Yes, it does. There was a reason why you memorized every letter in the alphabet.
> The way to do that is to just read books, study, work hard in your chosen field
Spaced repetition is literally just a way to study. You make the cards as you read books.
> The way to do that is to just read books, study, work hard in your chosen field and keep up with the latest developments.
These are the same things, just a different form.
Repetition!
If you're learning a language and trying to learn vocabulary, a new alphabet, etc, anything that involves lots of recall being necessary, spaced repetition is an excellent aid because without the basic shit you can't do any of the rest.
Plus it's supposed to be consistent over a (very long) time, not a "quick fix" so it's actually really the opposite of the "get slim quick" type schemes.
It's probably not especially useful when you're not cramming stuff/trying to build up a base to stand on, which lets you actually get on with the actual meat of the learning.
Repetition techniques simply ignore associations.
Learning new vocabulary works up until 200-300 words, then it's useless.
And if you're learning an alphabet where before you know 1000-2000 unique characters, you can't reasonably read anything, it's still very useful!
Plus who says that repetition techniques can't include context (like sentences for instance?), Bunpro being a pretty good example of this for learning Japanese grammar
Obviously don't just use SRS all day, consume media, read books, etc.
Spaced repetition is not meant for conceptual things or skills. It's meant for only the simplest of facts.
It's meant particularly for foreign language vocab, names and quantities in medical school, names of cases in law school, that sort of thing.
It has little to no relevance in math, physics, and engineering.
Just because it's not meant for all things doesn't mean it doesn't excel at some things.
> It has little to no relevance in math, physics, and engineering.
That's one bit I disagree with. Engineering is full of facts/concepts, things you often need to know inherently to be able to apply them, or even to know to google them at the right time. So I think SRS can apply there too.
You'd be surprised how big of a help this actually is for college exams. Not only for cramming before exams but also to retain the facts after for follow up classes (I see other students constantly struggle with having to relearn whole subjects while I don't because Anki implicitly makes me relearn them every week).
This is more a lack of imagination on your part than necessarily an issue with SRS.
- “reading books” (coming across the same ideas in different books, tying new ideas to old books)
- “discuss and debate” (remembering and being reminded of previously learned things)
- “study their subjects” (?) (presumably restating the previous two? - Maybe original research, which is finding new facts or reinforcing old ones)
- “work hard” (??)
I also wonder if there’s something about the mental process of people who seem to “just get it” - is their mind subconsciously turning over stuff they’ve learned in the same way an external spaced repetition regimen would? They may be getting the benefit of spaced repetition, but from the outside it doesn’t look like they follow a study plan.
Also some people just find it fun to go through their Anki deck instead of doomscrolling while on the subway or waiting in line. Whether there's any real benefit for that person is debatable. It's “fun“ in the same way going to the gym, or drinking kale smoothie is fun.
The key for improvement is deliberate practice, and one component of that is "working really hard" which you've pointed. But this is still too vague to be helpful. For those interested in the science of expertise, I highly recommend Peak by Andres Ericsson.
I'm probably one of those people, but commuting is one of those examples where you have a small (hopefully) amount of relatively low value time, time that is somewhat interrupted. What else of value would you do in it? Maybe listen to a podcast, catch up on blogs. All fine, reasonable choices, but doing a bit of Anki is a reasonable alternative.
Only time I feel like I've wasted those periods is when I end up wasting it (just scrolling through social media or random videos). Anything else is I think a reasonable choice.
This misconception is a common mistake by people that think Anki / SRS are equivalent to the spacing effect itself. That is not the case. The spacing effects exists independently of any particular software implementation.
I don’t think any one claims that memorizing a bunch of facts will make you a genius; rather that if you are aiming for optimal learning and memory, SRS is pretty much the gold standard and is backed up by tons of scientific research.
I don’t see spaced repetition as one of those get-fit-fast schemes. I see it as going to the gym as opposed to generally living an active lifestyle. Some gifted people can get very, very fit without going to the gym, just by playing sports and other physical activities. Gym isn’t that fun compared to playing soccer, and it takes a lot of time and consistency to be worth it.
Also, you can fit Anki into parts of your day where "proper" learning is more difficult. I used to do it on the bus to and from school, where it was far to loud - and I was often too tired - to learn through other means.
The alternative was scrolling through HN... I did that a lot, too.
Specifically, when do you create cards, how do you make sure your cards are good quality, what are the scheduling settings, how is it organized (sub decks?). I feel like Anki itself is overwhelming...
I primarily added Anki cards from Obsidian. I'd have a heirarchal note structure for the subject (e.g. physics -> paper 3 -> astrophysics -> stars -> star type categories). Then I'd have my plainly-formatted notes, and below it I'd have some flashcards. The Obsidian4Anki extension let me create cards from the markdown.
The cards would strictly be on the topic in the note, generally be written around the same time or a few days later, and would be Q-and-A or clozes.
So as an example, "What colour is an O class star::Blue" would be in my notes and turned into a matching card, or "A {1:type 1A supernova} always has an {2:absolute magnitude} of {3:-19.3}" would create three cloze cards.
At the top of each note I'd use a note property to specify the notes' place in the Anki heirarchy, and I'd use the first three parts of the Obsidian heirarchy. So I'd save it as Physics -> Paper 3 -> Astrophysics -> [many, many cards]. I always duplicated notes to make new ones, so I rarely touched this.
Generally I'd practice all the physics cards at once, but categorising them this way let me drop the paper 1 cards once I'd done it and focus on paper 2 and 3 etc.
Honestly, though, I first started this when I was fourteen. My notes and my Anki decks were a complete mess compared to when I stopped when I left school four years later. You'll find what works for you - and clean things up - as you go along.
You aren’t Einstein, I presume. Neither am I. So first of all, you can’t say that with any certainty.
But more importantly, the vanishingly small top end of the bell curve does not hold useful lessons for the rest of us. Things that mere mortals have to expend effort on are not even a second thought for them, so they have nothing to teach someone who is working to learn things they never even tried to learn.
This is why great coaches are usually middling athletes. This is why great teachers are rarely great in the field they teach. This is why lessons from those who are great are often frustrating and filled with statements like “just…you know…do X.”
You are 100% right but have absolutely no point.
This then ends up as something analogous to a table of weak pointers. I keep a sort of abstract digest of the information in the mental database and I use it to go look up hardcopy when it becomes necessary to actually reference.
The result is something akin to a vector database. I can look at a problem and apply many different approaches simultaneously. The mental machinery autonomously eliminates the worst paths through my knowledge graph and I'm left with a few pretty good options.
Actually memorizing all the information my mental database links to would be explicitly detrimental to the process. This pushes the signal/noise ratio way too low, and more importantly wastes time I could spend indexing a larger amount of information.
Perfect recall just isn't necessary for the way I solve problems. Reference books exist for a reason. every machinist ever has a copy of the machinist's handbook for a very good reason.
The process is all about knowing what is possible and drawing paths through the graph of possibilities. The details aren't relevant until the implementation stage when you'd be accessing reference materials anyway.
I find that spaced repetition does increase recall. I don't think that's in dispute. But better recall does not make you a better engineer. I don't think it's something even worthwhile in the general case, unless you're specifically trying to build a deep specialization.
People in general really dislike this notion. People want to believe that hard work and dedication makes an Einstein. The idea that it's just how some brains are is, I dunno, unfair?
But it's a tradeoff. There's a huge psychological burden involved. I won't get into it, but it's something Einstein spoke about. I'm also no use on my own. My abilities have led me to specialize deeply in generalism; I have no real specialization in any one field. If I had to operate on my own, it'd take me ten times longer than with a couple of specialists behind me. But with a good team, the amount of work we can put out is scary.
No, not everyone can learn to think and operate this way, sorry. My brain is wired different and you probably don't want it, it really, truly sucks. But that's a good thing. A healthy mixture of perspectives and abilities is what makes a successful team/company/project. You can't really get by with just a generalist or just a specialist. You need both perspectives.
Well put. But in that case, couldn’t you utilize SRS to memorizing high level concepts (e.g., coarse characteristics of various algorithms as opposed to how they are actually implemented, etc)
In general, I don't find techniques like this helpful. My brain is just really good at holding the overall concept of what I read. Once I've digested the information, the abstract is logged in the database forever. But that's just the autistic nature of my brain, I think.
I've met a ton of brilliant people at Olympiads who use Anki, and you'll find heavy usage from people who rank #1 in highly competitive exams (JEE Advanced, Cambridge Medicine etc...). I graduated top of my class, and I attribute this to consistent Anki usage
But, of course: there's a subset of truly incredible people who don't need it. I've met some people at 20+ languages, no Anki
----
Disclaimer: AnkiDroid maintainer. Did a bit of work on a new website today, so stats are top of mind.
There are a number of anti-Anki users (and to be honest, I don't blame them: there's 30k+ cards that they should learn in the 'meta' decks, and our onboarding is lacking). They've got an abusive workload.
The longer I'm US med-adjacent, the more jaded I become. There's continuing score creep, and this is due to the bar being raised[1], in order to maintain a failure rate.
I believe there's systematic academic misconduct due to the pressure to publish. The AVERAGE successfully matched neurosurgery resident has 37.4 research items (papers/presentations/abstracts)[2].
[0] https://www.cureus.com/articles/70371-an-analysis-of-anki-us...
[1] https://www.youtube.com/watch?v=1gKKAZ5aO8E [mean score in 2022-3 was 248. This was a 99th percentile score in 1993-4. Mean student in 1993-4 would no longer meet the passing score, exams are not getting easier]
[2] https://medschoolinsiders.com/medical-student/how-many-publi...
Commenters who argue with you, suggesting that discussion is repetition, miss one thing: associativity of memory.
Discussing material, reading something tangential or in another aspect of the subject, all that makes more associations, not just simply repeat the same fact.
I had tried that 20 years ago learning a new language: after ~300 words, my brain could recall every word in the cards, but not in real text. At some point I looked a word in a dictionary, then vaguely remembered that I had looked it up earlier. And a minute later recalled that I had a similar word in the cards -- I browsed through them and found it, and guess what, repeating by the cards, I named it very easily (I guess, brain learns any cues, like cursive writing features, and cuts every corner). But this knowledge never transferred to the real language practice (at the time I was starting to write and say something sensible).
I remember taking driving exam, the theoretical test. You're given 20 questions with pictures and a question whether or not you can do a particular maneuver. Some groupmates just rote-repeated the entire 800-question set of last year's test multiple times (and learned the pictures rather than their meaning).
I instead did the homework: analyzed the entire rulebook, and where rules were irregular and chaotic, organized them in tables to bring any kind of system to this. And then I had no troubles remembering those single rules. Also, I made myself a checklist for the picture questions, what to look at -- are there lights (which have priority over signs), is there a tram (has priority, and changes passing order), and so on.
The pictures in the real test were re-drawn in different style, and sure some folks couldn't load the right answer from their brain and failed the test. One of them told me "You're so smart" (probably implying that I simply have much more RAM in the brain, which isn't true). That was pleasure to ears, but I didn't try memorizing the material at all, and knew very little of it beforehand.
It basically works by optimizing parameters of the "memory model" against your review history, thus fitting against your particular memory instead of using a one-size-fits-all approach.
It is described in (much) more detail here: https://github.com/open-spaced-repetition/fsrs4anki/wiki/The...
I tried anki multiple times, ended up disliking it each time. FSRF did not made it better.
But in practice long intervals are shortened when you answer wrong once or more, or when you mark the answer as "hard" when it actually takes time for you to recall a card.
For me FSRF really work well, sometimes I see cards that I feel are really showing quite often and when I check the card history it usually shows that I had some "again"/"hard" answers in some recent past, and because of that these intervals and not increasing that much.
Plus you can also always increase "desired retention" from default 90%. You will have shorter intervals, but it will of course increase number of the reviews too.
I just concluded I am better off without algorithm that assumes once I have seen a word on flashcard once, it is reasonable to see it next time in months.
* Model the forgetting curve as p = 2 ^ (−∆/h) (p = probability of recall, ∆ = time since last review, h = forgetting half-life for a particular card and user) (see https://research.duolingo.com/papers/settles.acl16.pdf)
* Numerically fit this curve to the last ~10 recent reviews of the card by the user to give an estimate for h (the half-life but effectively the intrinsic "difficulty" of the card for that user)
* Schedule the next review of the card at the time when a particular recall probability is reached. This probability should be configurable by the user: >90% for an "easy" session, <70% for a "hard" session.
* Use the "due date" to queue up cards for review. If a user reviews a card _before_ the due date, ensure that the half-life is never reduced after a succesful review. Provide some feedback to the user to suggest they stop the session when there are no pending cards left ("10 more cards due today, 100 cards due tomorrow, 1000 cards due next week")
* Don't schedule the same card twice in a row as this is annoying for the user. If it has just been reviewed and is due next, move it back to second place in the queue.
I found this works well once there are 2 or more reviews of a particular card for a particular user. For scheduling the _second_ review of the card I had to invent a heuristic like: if first review failed, schedule again in 1 minute, if first review passed schedule again tomorrow. However in apps where many users are reviewing the same cards, you could probably build a good model of the difficulty of a card for a user before 2 reviews by considering the performance of other users on that card.
www.mochi.cards
I make cards on a linux desktop and sync to ankidroid.
Really one of my all time favorite pieces of software. I would have given anything to have had this 30 years ago in school.
I found that spaced repetition really works as advertised, with software like Anki. However, it is not really suited at learning large amount of facts and concepts. If it took an hour to review your deck, it's not a good use of your time unless the cards are really high value.
For example, if I want to learn electronics, my time would be better spent on fiddling with circuits and learning through mistakes of implementation. The nice thing is that concepts are interrelated to each other, not discrete and decontextualized that usually happened with utilizing spaced repetition. You get much better value from learning through experimentation and retain it better.
Don't get me wrong, spaced repetition still apply. However, if you can learn something that will last a lifetime, you don't have to repeat it for a long time if ever.
for really rare words that I'll use 5 times in my career, sure, I'll explain it to chat gpt and it'll get it in one go.
Personally, I just went through word lists. When I had them memorised, I'd put them into Anki deck. Because I am going to forget them over time.
You first understand and only afterwards encode your understanding in cards. It's in reviewing the cards that you put the understanding into your long-term memory (or possibly reveal gaps in your knowledge, in which case you revise your understanding).
direct translation: what has this with the matter to do?
meaning: what does this have to do with it?
Where does understanding come in? I can imagine what understanding mean for mathematics, science, etc. but for language, I'm having difficulty figuring it out.
so the understanding took far less time than remembering. I could read a whole grammar book 10 times, but if I don't remember any of the words, then I'll never be able to write/say something.
having all the words on the tip of my tongue is no good for communication.
But it's possible that we have conflicting ideas of what recall/understanding means.
It is not useless. It tells you essentially two things. First, it shows the limits of your model. If you have proven that sailboats can go infinitely fast, which is obviously not the case in reality, then it is an indication that your model is incomplete and you have to refine it until it stops giving out nonsense, or establish its limits. Second, it shows that what you have taken into account in your model is not what limits the speed of sailboats or memory recall.
The premises are obviously wrong, like assuming an infinitely long lived person, which means the answers arrived at will have no correlation with anything factual. You can't fix that with any sort of refinement. To be refinable would require it to be coarsely correct to begin with.