I have been using SRS for everything in my life for the last 2.5 years and I have no doubt I will be doing ANKI flash cards on my death-bed someday (as depressing as that sounds). It's the single most important productivity hack I have ever discovered (when used effectively). I successfully used ANKI to learn a 2nd (and now 3rd language) and retain a lot of knowledge I acquire from books:
I think what really puts me off is the effort of building those cards. It’s not quite a productivity hack if I’m having to sit and spend hours making those cards (vs. someone making those cards - and effectively - for me in exchange for $$). I like memrise a lot for some languages though, and they’ve long been advocates for this technique by building an entire business around it. Not much good material for math/CS on it last I checked.
There's a few decks I've seen for math/CS, but I haven't had much luck myself, and have resorted to making my own. I've been trying to figure out the best way to do it.
It's pretty easy to do with Supermemo. But, it's still not as easy/simple as making cards for vocab.
For a specific math/CS topic it's not too hard, just in general.
Exactly, making the card is half of the trick to remembering. If you can’t commit to making your own materials you might as well forget about using things like this. Yes it’s harder, but no there is no shortcut for learning in life. Just like exercise you can’t fake doing the work.
I find it’s less the act of making the cards and more the lack of tools. Its possible to make Anki cards on the iPad app but many features are missing. Perhaps I just need to try AnkiWeb.
This gets used a lot in language learning, especially in the Japanese learning community where people are trying to memorize thousands of Kanji characters. Like the article says, the idea is that recalling a fact when you are on the verge of forgetting it causes it to be reinforced, increasing the interval before you will next forget it. Using the Japanese language SRS tool WaniKani (https://www.wanikani.com) I’ve memorized around 900 vocab/characters in the past three months (which is actually kind of slow compared to most people).
This has been a fun experience for me. I’ve always felt like my memory was poor compared to most people, but now I have proof that with the right strategy I can remember anything I want.
> Some researchers also believe that semantic priming is a factor.
When learning a new Kanji in WaniKani you are given a mnuemonic to associate with it. I find that my retention is much better when the mneumonic feels strongly associated with the shape/meaning/reading of a character.
Another thing that I’ve found helps retention outside of SRS is using/encountering an item outside of a study session. The surest way for me to remember a new word is to hear it in a show or song where I fully understand the context it’s being used in. Also if the thing you’re memorizing more of an abstract idea, then explaining it to someone else is a great way to reinforce it for yourself.
When I looked at WaniKani some time ago, I got the impression (based on their description, I didn't try it) that they first teach you radicals, then kanji made out of those radicals, then words made out of those kanji. Is that correct?
I didn't try it because the approach seemed to lack context to me, and also because I already know Chinese, so learning Kanji wasn't a priority (I wanted to focus more on pronunciation, vocabulary and grammar). Instead, I built my own sentence-level SRS by taking examples from the Tatoeba corpus, segmenting them with Mecab, adding audio using Open JTalk and then scheduling reviews based on the probability of not knowing one of the words in a sentence. I've been dog-fooding it for three months now, and in theory I should be able to understand 14310 sentences from the corpus.
Yeah, it does build up from radicals->kanji->vocab over 60 levels. I agree, for someone who knows Chinese already WaniKani is probably not the most effective way to study. But for people who are unfamiliar with the Chinese/Kanji character set learning radicals first really helps with mentally parsing characters with high stroke count. However as I’m sure you know Kanji readings can be a lot different from the Chinese characters they came from, so there is still some memorization work to be done.
Also WaniKani doesn’t touch on grammar at all, so it works as only a part of an overall study routine. I do wish there was a similar SRS service for grammar/reading. Building up your own decks is such a huge amount of work, and for someone who is just learning it’s nice to have a curated deck that is verified for correctness. Sounds like you have an awesome system that’s working for you though.
I did a big surge of learning this way for a post-graduate exam in 2010. It was very effective, much better than any of the 'usual' ways of learning.
However, once the exam was over, there was no way I could maintain the 1 hour plus of card review each day. Eight years later I remember very little of what I learned.
Admittedly I was studying quite hard, ending up with nearly 4000 cards after 6 weeks, perhaps there is more luck with less intense schedules. The material was also complex and esoteric.
4000 cards in six weeks is completely doable long term as long as you configure ANKI correctly. As an example, I have 14000 cards in my Italian deck, which I have been making for 2.5 years, and 6K in my french deck, and I do my reviews in < 30 minutes a day. Everyone can find 30 minutes
Dr Ali Abdaal runs a YouTube channel (about studying medicine at Cambridge University) and one of his videos is on "Evidence-based revision tips", with citations for the studies he's working from - here: https://www.youtube.com/watch?v=ukLnPbIffxE it's about 20 minutes.
He says that spaced repetition is effective, but basic repetition of re-reading, re-watching, re-listening is not effective.
Spaced repetition with "active recall" comes out significantly more effective - instead of exposing yourself to the same material over and over, challenge yourself to recall the material at the time when you're on the edge of forgetting it; the active mental effort of doing that appears to fix information in memory much more effectively than reading or hearing it again.
A consequence of that is his suggestion that notes and review material should not be facts you want to remember, but questions that will prompt you to think and recall what you want to remember. "Writing questions for yourself makes you engage in cognitive effort, and the more brainpower it takes to recall a fact, the better strengthened that connection seems to get, according to the evidence at least".
Hm, I'll probably develop mobile before Firefox so FF would be long way out, but not out of the question! You can follow @ForgetMeNotBot on Twitter for (very infrequent) updates.
Useful review. One caution I have while reading posts like this, is that the author does not distinguish between the testing effect, and the spacing effect.
While the value of testing (and especially spaced testing) is well established [1], the value of spaced restudy (e.g. you see both the question and answer together) is less established (but see this Bjork paper [2], which argues there is some benefit to spaced restudy).
There was a popular discussion on this topic somewhat recently on HN[1] when a Wired article from 2008 entitled To Remember Everything You Learn, Surrender to This Algorithm made the front page.
This sounds a lot like learning an instrument. You should be practicing consistently in short bursts every day. You don't have to practice for more than an hour or two. The consistency is what's key. You'll develop better technique faster if you practice 30 minutes a day as opposed to 3.5 hours every Sunday.
What I'm curious about is how can I use something like this for books, or podcasts? Re-reading entire books or re-listening to 1hr+ long podcasts is obviously not realistic.
Should I be re-reading/re-listening to certain chapters? Keep notes and refer back to them often?
There is something called "Mindmaps", learn about it, it is very useful.
With mindmaps and using the "pause" button in a podcast you can take notes WHILE you listen or read that are minimal. The trick is that you do not need more than a few words-images-sounds in context to reconstruct all the podcast or video or book.
What is context? Imagine you have a screw and a piece with a hole on it. Where the screw goes? There is only one possibility. You do not need to store the info "where it goes". You can reconstruct it on real time.
With multiple elements interacting, reduction is enormous,in fact neurons work extracting connections of key elements. Take japanese "r, l sound", because the language does not make the distinction,a monolingual japanese could not differentiate between a "Ra" sound and "la" sound. The additional info is of no use so reural connections prune it.
You hear the additional info, but discharge it.
With context you could reduce information to remember more than two orders of magnitude(100-1000 times less).
What you do normally is forget most of what you hear. I believe re reading entire books or podcast is completely realistic if it is a good book that is important to you. There are not so many.
If you have read lots of books or podcast you know which ones are essential.
For me for example it could be "The C programming language" or "Structure and interpretation of computer programs". I have created a C and LISP compiler so those books are sacred. For Einstein, it was Newton's or Faraday's work.
I read books that I could reread 5 times and I learn something important and new every time. As I get deeper, I become a master that could understand concepts and ideas that myself as a novice could not grasp.
> Should I be re-reading/re-listening to certain chapters? Keep notes and refer back to them often?
No - simply re-reading your notes or the original text is not optimal and can give you a false sense of progress.
It's important that your repetition involves active recall - that is, you must close the textbook/notebook and try to recall the key definitions and ideas. Only then should you open your notebook and compare your current knowledge with the original information. If there are large gaps in your knowledge, schedule the next review soon otherwise leave it longer.
I've found it's quite painful to sit and force my mind to grasp onto ideas which are just out of reach, especially when the information is just a click away, but it leads to much better retention of important knowledge.
Read other materials on the same subject. That has the benefit of giving you similar things, differently. Maybe another author is more illuminating. But both will introduce the same vocabulary, so there will be overlap. Plus you'll have to think about what you agree with.
You can with Supermemo's incremental reading feature. There is a plugin for YouTube if you can find your podcast there. I'd recommend converting your ebook to html first though.
The best introduction to spaced repetition, Michael Nielsen's Augmenting Long-term Memory (http://augmentingcognition.com/ltm.html), includes a detailed explanation of how to use Anki for studying books.
I recently built a Chrome extension that encourages you to add anything you don't want to forget (say, insights from a podcast), then quizzes you in your new browser tabs on those chunks of knowledge using spaced repetition.
Interesting. This is basically something I learned in college and am relearning through necessity with my son (11 years old), who has ADHD. Long study and homework sessions are impossible and, for him, tremendously discouraging. We've learned to take things in 15-30 minute chunks with breaks in between and we spread things over several days and the results are positive. Getting homework and projects done isn't such a chore and he feels better about himself since he's better prepared for tests.
I used Anki back in 2008 to help me learn Japanese. The best result was memorizing 10,000 vocab inside a 12 month period during my second year of study. That did phenomenal things for my Japanese in combination with decent amounts of conversation.
The power of SRS if you are someone who can take a little bit of pain and who is good at devising learning strategies is off the charts.
I guess the best way I can put it is there are some things for which the traditional model of learning would only take you so far but that through a cleverly devised SRS strategy you could take yourself to places that would be traditionally not attainable for you.
Maybe that's a bad explanation, I don't really think I'm doing it justice.
Something I've settled on attempting in 2019 through SRS is learning psychology. Now SRS is only ever PART of a strategy, it's never the whole thing. But my strategy to get me to the next level that just reading pop psychology books isn't getting me to is that memorize all the replicated and highly cited studies in psychology across as many branches of it as possible. This takes the format of remembering the names of the study/paper, the year it was published, who the authors were, what journal it was published in. The second part is memorizing and being able to verbalize a summary of the the study and it's conclusions.
Think what you will about what that will do for my ability to understand psychology but I can tell you this for sure: rote memorization of all the facts then frees your mind up to get to the next level then connecting all the dots between the facts. People often criticize rote learning because they think learning is more about connecting the dots and that's what we should focus on. It's exactly backwards. First focus on getting all the dots into your head and then inferences start to come thick and fast naturally.
The key to devising an SRS strategy is think in terms of "pieces of information" and to find good sets of "pieces of information" that have extremely good power to weight ratio.
There's two things I'll say there... You COULD use SRS to memorize an entire introductory psych textbook. You COULD. But it would be the exact wrong way to go about it. If you memorize the relevant category of information you can use that to Intuit all the explanations .
The other thing to say is the "pieces of information" should have the same shape. They should be in the same category. They shouldn't be random, but highly cohesive. For medicine this would be like you start with anatomy first and you do all of that before moving on. You go in layers.
The next key is then your search stategy on how to locate the information.
The next key is defining different phases of your strategy which boils down to figuring out which set of pieces of information to after in which order to achieve which specific goals.
Nicky Case recently made this interactive explanation of how it works, which is itself an SRS system: https://ncase.me/remember/