I No Longer Tell My Friends about Anki/SuperMemo
masterhowtolearn.com
masterhowtolearn.com
First of all, if you think you've found the world's ultimate fighting technique, don't try to show it to random people for approval. Use it. Anki yourself in secret until you're a super-saiyan, and then unleash your strength on the quivering masses. Hear the lamentation of their women, until they are begging you for mercy and also to teach them about spaced repetition.
That brings us to the second problem... spaced repetition doesn't work. Or rather, it does work, but it does the wrong thing. If you want to be a walking encyclopedia, maybe SR is a good approach. But if you want to understand the world, be able to do things, be funny and gregarious, know the answer but also know when to play dumb... SR can't help you. Even if you want to learn a language, if you can't learn it by using it, do you even need it?
I was excited about SuperMemo, Anki, etc. too at one point. Eventually I decided that they just aren't that good. Now here's the thing: if you really think SR is the best thing ever, don't argue with people - prove us wrong.
Not everything can be solved with a hammer, but sometimes you have a bunch of nails that need to be hammered-in.
Not having to look up syntax and what not is a far better productivity improvement than editor hacks for example.
There’s a reason why this is popular in the “rationalist” community: they’re mostly college kids. I have never seen a coworker or other industry professional use any of these tools.
Why, because doing isn’t the same as remembering. Over time you remember by doing. That’s it.
I agree with you that for knowledge you need everyday it is better to just learn on the job. The relevant knowledge will stick/you can note it down in a cheat sheet if needed. There's just not enough time to learn the full feature set of a programming language, you won't need most of it anyway and you will be a better programmer by reading existing code and writing code yourself, which incidentally is similar to how best to learn a natural language (listening and speaking)
Prove you wrong how? By learning things with it? Plenty of people already have used SR to pass tests, aid in learning a language, etc. There are scientific papers that show SR to be an effective learning tool.
I really don't get what you're asking for. It's not like SR learning is going to let you "unleash your strength on the quivering masses".
You also have to propose an alternative to it. In your comment, you seem to have critiqued it with "if you want to be an encyclopedia, maybe SR is a good approach. But if you want to understand the world .... SR can't help you".
The fact that you say "If you want to be an encyclopedia, maybe SR is a good approach" indicates to me that you think SR does actually work, and so there's nothing to prove here.
Yes, obviously if your strawman is "SR compared to moving to the culture to learn a language", then SR will fall short. But no one is saying that, everyone's saying "Instead of looking up a word in a dictionary each time you forget it, use SR, and instead of just doing immersion, do immersion + SR"
The author of the article is specifically struggling with his evangelism. He says he thinks the world would be a better place if everyone used spaced repetition. I don't know (can't imagine honestly) what sort of utopia he envisions, but he needs to show instead of tell. Just like I need to.
It's so insanely hard to lead people. I think the only thing that works is to lead by example. Any sort of talking is just fluff. You have to show people how to accomplish or attain something they care about.
As for SR itself... for decades we've had some of the smartest people (and perhaps even some of the best educators) like Richard Feynman decry rote memorization in the classroom. SR takes the worst 8th grade history class you ever had and doubles down on it. You can choose what to memorize, but it's still just memorization. Why not use a personal knowledge base (== your Anki deck) that you can refer to, and let your memory naturally cache the things you commonly use?
If you have a problem and you're considering SR as a way to solve it, my advice is to really drill down into what the problem is. Why do you think you need to fill your brain with snippets of data? (If you're just trying to pass a test, then sure, use Anki.)
Also, you seem to underestimate the importance of having encyclopedic knowledge. I agree that it is not useful to have random facts in your mind, but passive knowledge is extremely important - like even on topics you are not well-informed about, having learnt something about it in the past will put you in a much more open-minded position on it. If people were to remember some really basic biology from high school, they would be much less open to bullshit like “alternative medicine”,anti-vaccines, etc.
At the same time, learning in most areas requires some memorization. Even if, in terms of knowledge skills, SR only helps directly with the memorization component, there's considerable empirical evidence that it does make that more time efficient. Using it to structure that part of learning allows you, all other things being equal, to reduce the proportion of learning time devoted to memorization.
> SR takes the worst 8th grade history class you ever had and doubles down on it.
Well, SRS maybe. Having first learned about SR in ballroom dance teacher training as a tool for structuring both curriculum for in-person instruction and personal practice, I disagree.
The common usage pattern for spaced repetition software may do that, but spaced repetition is more than just that.
Because you can combine instantly different pieces of your knowledge to attain new insights.
If it's not in your brain you might not even think of Googling it.
It is also extremely labour intensive compared to spaced repetition without noting - using direct material of similar or higher level, preferably different modality - passive, creative, vision, sound, kinesthetic.
How do you note a movement? Or sound?
You'll waste some time reading material you already know, but that is the essence of spaced repetition; the use of different modality makes it much more potent.
This is what the SR tools don't get, or where you need to put in huge amounts of time building decks. Most of them pose passive problems. Even the simplest learning books have exercises which are supposed to engage active modality, on top of testing.
This standard isn't tenable for some things.
For example, I think the world would be a better place if we gave everyone free healthcare. What's that, the standard for this has to be me literally doing the thing I'm advocating for?
In a similar vein, all the author's really saying is "if people cared about learning, and implemented strategies to learn, we'd live in a better world" (a more informed, thoughtful society perhaps). Whether SR actually accomplishes any of that is questionable, but the idea that people continuing to practice to learn will improve the world does not seem like it merits a "show me and I'll believe you" answer.
Just like it's unreasonable to say "well, implement free healthcare, until then I won't believe that it's worth doing", it's unreasonable to say "convince everyone to use SR, until then I won't use it or be convinced". It's an impossible bar.
It's also impossible to lead by example for aggregate behaviors. If you think that a park would be better if everyone picked up dog poop, and you lead by example by picking up all your dog's poop, that doesn't mean you're wrong if other people still don't pick up their own dog's poop.
I'm not really into flashcards, but I'm seeing a lot of comments like this, that flashcards are only good for memorization. I think they're more useful than that. You can use flashcards similar to how you would train a machine learning model, large iterations of input and feedback. People who are serious about language learning with SRS don't just study words but whole sentences per card. You could probably use flashcard to train yourself in many things that would be considered skills, not just knowledge. For example, you could put chess problems with answers on flashcards. If the resolution were high enough you could probably train yourself to recognize forged paintings. Things like that.
Knowledge isn’t that crap you get out of books. It’s whatever you can know that empowers you.
Yes, merely memorizing a lot of facts is not very useful. But unfortunately, burning a lookup table to your head is often a crucial step for actually using many types of knowledge, and it's a real headache for beginners. I find it's entirely possible to fully understand a result, yet unable to clearly remember it. When you need to use the result, having to pause and think (even for a few seconds), creates a significant cognitive overhead. Being able to recall some results mechanically is often advantageous.
For example, a long time ago, it was lifechanging after I had discovered "man 7 units" and eventually remembered that 1 kilo = 1/(1 milli) = 1e3, 1 mega = 1/(1 micro) = 1e6, 1 giga = 1/(1 nano) = 1e9, 1 tera = 1/(1 pico) = 1e12, I was able to calculate much faster, and to say "The CPU is clocked at 1 GHz, which means it ticks every 1 nanosecond" without thinking. Previously, I fully understood what a SI prefix is, but I simply couldn't build a lookup table in my brain to map words to the number of zeros, and always needed to pause a moment to count. The man page allowed me to instantly check it while making calculation without leaving the commandline, effectively it was spaced repetition without software, and I finally remembered them...
This is when spaced repetition is useful - burning some lookup tables to your brain. It cannot help you to understand things, but it has its place. Just be aware of its limitation.
Now 20 years later I run all my infrastructure from my basement, no cloud dependency for me, my OS does not spy on me and all tools I need for day to day life are Free... Freedom!!! See I proved them wrong, they should have gotten into it... Right? Wrong! Nobody cares!
Until they're obviously affected by the implications (e.g. of depencies & privacy) :)
Same experience.
But now I found having a cynical mentality is actually liberating, it allowed me to mostly ignore arguments and holy wars among people regarding to the platforms I'm using. I can only say, "It's simply my personal preference, and I fully accept any good or bad consequences of my choices. Talk to me only if you are genuinely interested." Now my main computer is not even running a standard distro anymore, but QubesOS, an experimental security operating system. When I first installed QubesOS, I found I was unable to work on my personal projects because some features I needed was missing. Instead of writing an angry blogpost (it sucks!) or defending my choice (this view is unfair for a project with limited resources!), I immediately cloned the git repository and spent a week hacking the source code to make it work for me without saying anything. Objectively, is it a good user experience? Not at all, but I'm personally okay with that, and hacking code is much more enjoyable than advocating the platform in an argument.
I guess this mentality is not good for encouraging improvements, and I fully respect those who choose to actively engage, but for me, I need to have a rest.
> A commonly used phrase by software developers to indicate that the bug reported by a user is not repeatable on their machine, and will therefore receive no more attention. Usually connotes a dismissive approach, where anything that is not visible immediately to the developer is "someone else’s problem" and is therefore not worth fixing.
[0] https://events.ccc.de/2016/11/22/hello-this-is-33c3-works-fo...
You're right, just like the OP is right. The people you're talking to you don't care about the specific things that you care about (privacy in your case, improved learning/memory in the OP's case).
That's understandable - different people care about different things.
I'm often left with the impression that discussion of biases for many ironically becomes a way to reinforce biases. I've even seen people argue as much - "I'm aware of my biases, and you're not, which is why I'm correct, and you're wrong."
I think maybe you conflate complaining with hostility. Humans, even rational ones, complain when things go counter to their expectations. It's normal human behaviour, and everyone, everyone does it. It's so universal and basic, an AI would probably not be complete without it.
I can't think how I would use this in a software development sense. Coding katas come close to this and a useful but I'm not sure flash cards would do the same.
Maybe it would be handy to memorize a libraries API but then if I use something enough that memorization would be useful I'll learn it by way of using it.
musical notation
geography
history
chemistry
metallurgy
physics
healthcare
...
This is one of those 'what have the Romans done for us' comments. Intelligence + facts = smarts. Facts are to thinking like fuel is to an engine, without facts your thinking will be either useless or wrong. So regardless of how smart you are some degree of access to facts is a requirement and if you can do this without having to look things up all the time it will go faster and with access to more facts the quality of your output will go up.
I think you'll find you learn new things quicker with judicious use of spaced repetition. As a bonus, you'll never forget those things as long as you maintain the habit. And we're all always learning, so that shouldn't be hard.
I do think the hardest part of spaced repetition use outside of the classroom is choosing what to memorize. In school, those decisions are made for you. I find it easy to bite off too much and start memorizing minutiae of a technology instead of the essential items. But I think it's just another skill to master.
Hardest parts of physics and chemistry in understanding. Again, you can use flash card memorization here and there, but the first of it is different.
Once you listen to more detailed lecture or read more detailed book, they become clearly distinct due to many small differences and stories. So misplaced event suddenly "dont makes sense", because you can explain why it could not happen then.
Exams are often in a very well-defined scope where you can rote learn most of the definitions, and the few exercise variants that appear in preparation also can appear on the exam.
I don't think that this is the best way to learn many things, but exams are partially tangential to true learning and it works well enough in my experience.
Pure memorization of facts is also really helpful to creating new understanding: Have you ever read a textbook that stacked definition upon definition and at some point you can't really keep up anymore because you're going back to the previous paragraphs all the time? At that point, pure memorization of definitions (even without understanding them really) already helps massively to reduce cognitive overload and makes forming new understanding form the rest of the text even possible.
Spaced repetition systems are not just about “rote” memorization. When used right it is a very personalized practice tool that can help with quite high-level ideas.
I wouldn’t fx. make flashcards for an entire API — instead I would focus on the concepts and features that are important for what I am doing. You can make it about “exactly” what you want to practice and remember.
Think of it more as a collection of personal notes with build in recall practice, rather than remembering scores of “random” facts.
anki itself is not super popular since there are apps like zyzzyva that are specialised for scrabble word study, but the technique is.
You can program yourself with flashcards[1]. EG I have a card that says "What should you do when someone is explaining something and you don't understand?" (Ask questions until you understand). When my colleague is explaining something and I just don't get it, this flashcard triggers me to ask questions, instead of pretending to understand.
0: https://en.wikipedia.org/wiki/Emotive_conjugation
1: https://www.lesswrong.com/posts/iocMeRcSKf3gAxsTh/self-progr...
My shop teacher would tell us to write down when we needed to borrow a tool. If we had to borrow a tool twice, it was time to go and buy it. I think the same rule is useful for spaced repetition. Even if you'd learn it eventually by regular use, once you have it in Anki you'll never forget it, even if you spend a couple years away from that particular technology. When you come back, you'll still have those configuration flags at the tip of your fingers.
Congratulations, I guess?
Generally something is worth optimizing it this much when you're going to use it more than 10 times. Insert obligatory xkcd chart here: https://xkcd.com/1205/
There are things that are this useful, but they're generally in category of mathematics, potentially physics, chemistry or linguistics.
Learning parameters of a shell command, especially all of them, is a waste. Learn useful things first.
This is the main problem SuperMemo and Anki sets face, the ordering is by difficulty only and not by product of difficulty times usefulness.
To get widespread SRS adoption, two things need to happen:
1. The primary software needs to be simplified and prettified. The aesthetics of Anki are unappealing to pretty much everyone. And just using it has a learning curve far greater than basically every other popular app.
2. The marketing message needs to be about benefits, not features. No one cares about the forgetting curve or the intricacies of available card types. Everyone cares about saving time and money, learning cool stuff, and having fun.
I use Anki a lot, so I’m not in the best position to judge its first-hand impression. But, what would you say should change? Both in terms of visuals and interaction?
Genuinely asking.
The success of WaniKani for example is notable because you can effectively make the exact same thing for free using Anki (and the unofficial decks are available) but it is still quite popular to pay for the managed experience.
From the workflow of downloading Anki and adding to the default deck, how do the settings and knobs factor into it?
I’m not here to just argue. I am legitimately interested. I just don’t want to assume too much.
https://apps.apple.com/app/ankimobile-flashcards/id373493387
It looks just as dodgy as “doctors hate this one weird trick”, and costs a ton up front. Why would I believe something like that?
I use it daily and I just find it functional, but I’m also very used to it.
- The UI is an ugly light gray with too much text. I'd make the design far more minimal.
- Too much of a focus on text. Video, audio and GIFs should be more integrated into cards, especially for stuff like foreign languages.
- Way too many options: note types, sort fields, tags, on and on. I've been using Anki for years and I don't even know what half of these things do.
- The entire process of adding cards is confusing and not immediately obvious. There's not even a tutorial.
- Premade decks have zero quality control and you have to go looking for them online.
- Above all, Anki just looks and feels like a chore. When I show it to people (like OP), the overwhelming reaction is always, "Meh." There's nothing exciting, beautiful or futuristic about it.
I can't imagine why anyone would want to use the original desktop Anki, which chains you to a PC or laptop. Though in the pre-mobile-computing era, such a program would have been attractive.
This technique begs to be mobile, so you can do reviews in any occasion when you would be otherwise idly waiting for something, like standing in line somewhere. Buses, trains, planes, waiting rooms, washroom stalls ...
You can micro-review. If you have just a few seconds to spare, you can whip out your phone and just do one review and put it away. That's one less to do later.
I sometimes do AnkiDroid reviews lying down in bed, before sleeping. You can turn of screen rotation and work sideways.
I've woken up in the middle of the night and completed the last few reviews of the day. That's possible from having the phone near the hand; no way would I get up, and sit at a desktop PC to do that, or even laptop.
The desktop Anki is completely pointless except for (1) serving as the reference implementation and (2) certain complicated maintenance tasks on large decks.
It really isn't even a comparison...the mobile app is only useful if I'm on the go and have a minute to do some cards. An hour on the desktop version will let you do hundreds, even thousands of cards, while mobile is maybe a hundred, tops, assuming you want to stare at your phone that long.
(For languages with an ideographic writing system, there is a benefit to writing by hand, but that's a whole separate story.)
It sounds as if your flashcards are in a format where the foreign word is the answer, or embedded in the answer via a cloze. That's not the only way Anki is used. If the flashcards are in a foreign-word -> definition format, the typing would have to be definition-side, to which there is little benefit. All the more obviously so for definitions that are in your native language. Definitions are too verbose to type out, and difficult to machine-grade due to variations in wording.
I understand the efficiency benefit that keyboarding can bring, but it's not realized in Anki.
I made a very good application for kanji memorization, where the keyboard is the key to efficiency. The reason is that in this program, all of the characters you are tested on appear on the same screen simultaneously. You type answers into blanks next to each character, and use Tab to move from one to the next. The machine grades your answers when you're done; you don't have to engage in time-wasting self-checking. This is extremely fast because you can operate in pipeline mode: your eyes scan ahead several characters while your fingers type without stopping. Also, if you're having trouble recalling, you can just leave the answer blank and then easily return to it before submitting the test for grading. Often, your mind will come up with the answer in the background, while you are filling other answers.
The main source of inefficiency in Anki is the stop-and-go blocking: being shown one card at a time, and not being able to go to the next card until passing the current card. Anki is best used for the cards for which that is the most efficient approach: cards that have a simple front, and a complex back that requires self-evaluation. For cards that have a one or two word answer that can be machine-checked, I would find (or develop) something much more efficient than Anki.
1: https://quizlet.com/blog/spaced-repetition-for-all-cognitive...
2: https://forum.duolingo.com/comment/36567535/Quizlet-is-shutt...
If you research that question - “how to study?”- a bit, you will immediately be inundated with dozens of resources swearing by spaced repetition. Once you use it for a few weeks, it’s clear that the practice itself is helpful for retention. However, flashcards alone are simply not going to teach you a full language. You need reading exercises, speaking exercises, listening exercises, and writing exercises. These are rarely, if ever, all found in one place, and I quickly found myself following a few different programs, each teaching part of the full puzzle.
People don’t want a tool. They want a silver bullet. Anki is very much just a tool, and like other tools, it requires you to do work with the tool to get a desired output. Making good cards is a time consuming nightmare if you want images and sounds, and getting those cards to sync across devices is non-trivial; you’re lucky when a mobile app supports importing all your media in your decks, or lets you study reversed cards in a sensible way. Proponents will say, “You can just import some great premade decks!” This might be useful for an intermediate, but as a beginner, having a deck full of vocabulary that deviates from the words I’m being taught in my other lessons is nearly useless.
I can see how it’s tempting to tell the world - “You could learn so much faster with this one weird trick!” And the author knows that sale works; it’s easy to get someone to try Anki. Those users quickly realize that there are barriers to reaching the point where they have a net productivity increase, so why would they change what was working just fine for them already?
2-3 months ago I was about to take the Goethe B1 exam (German language). I would had probably passed it anyway, but wanted to be sure. I downloaded the official Goethe B1 wordlist (2.5k words), and one week before the exam I started learning it in Anki.
I spent on avg. 2.5h per day, and out of those 2.5k words, I didn't know maybe 800. I've learnt them. Many of those words appeared later during the exam.
I told about it my German teacher, somewhat excited, and she was sort-a 'meh, whatever, you can also learn in other ways'. I tried to explain, that "800 words, and in a week, and so fast", but as much as she loved my good exam results, she had no intention to recommend this way of learning to others.
In any case I use many of those words actively today (e.g. during my German classes).
E.g. yesterday, I tried to use word 'horizontal' in German, as it was on the list, and couldn't recall it quickly. My teacher gave me the word ('waagrecht'), and I should have obviously know it, and would know the meaning if somebody used it in German, but couldn't recall it easily.
Solution: don’t talk too much about it. Casually drop a sentence or two and then let them figure out the rest.
Your goal should be that someone hears of your important topic at all. Because it’s often just random chance whether or not we stumble over something that is interesting or good for us.
That’s why the randomness of HN is so appealing. 95% is useless for one specific person, but once in a while that person stumbles upon a gem that is exactly right for them.
And sometimes, it is the right info but not the right time and they might come back to you later to ask more, although this rarely happens.
Make a website for people that are already interested in the topic. The rest of your effort should be spend just mentioning the sheer existence of your topic. Don’t lecture anyone. That is enough.
OP talks for 10 minutes about SRS. Yes, nobody gives a shit. Instead say: I use Anki, because it helps me to learn XYZ a lot faster. Then change the topic. Not mentioning it at all is bad, because it robs the other person of the chance that this is the gem they actually need right now.
"Oh you skateboard? I was thinking of getting into that" is now met with a meager acknowledgement and nothing more. "Can you teach me?" is now met with "No", because people just don't really give a damn and they have other stuff going on that they might not realize in the moment. Replace skateboarding with programming or anything else and you have feigned interest, so it's best to turn it around and ask them what they're into.
Bringing Anki back into it for a second, I did not have any success with Anki software or really any software. It just did not help me beyond giving me access to long lists of words. Instead, SR with written flash cards and questions seems like night and day.
- Only very few have the temperament and external motivation to practice with SRS.
- Learning by reading texts reinforces vocabulary at exactly the right strength: you learn the more common words sooner.
However... think about it like this:
If just one (1) of the dozens of people the author mentioned Anki to took it up and learned a whole new skill, their life would be changed. Thats pretty cool! And reward enough.
It’s not remotely useful. Changing hearts and minds is a skill in itself. It’s why people spend 3-5 years in a university to pick up skills on how to do it. Yet here we have some anecdotal story about someone who fails to convince strangers about some method of learning that is described as “99% of people never learn learning, you can be the 1% by doing this...”, and lets be honest, those people aren’t the authors friends, because actual friends don’t just shrug off your interests. I feel like I’ve just read an article on a subreddit about cringe or niceguys/gals about someone socially awkward who don’t understand why people don’t think they are the greatest.
Maybe it’s just me, but I sure wish I hadn’t ever read this.