LESSWRONG
LW

Comment Permalink

If only there were hundreds of studies listed in http://www.gwern.net/Spaced%20repetition#literature-review which demonstrate that there are many ways of spacing learning which don't exploit feedback like the SM algorithms demand...

ChristianKl12y00

If someone takes a weekly class that's teaching dribbling every monday, then you could call that "spacing learning". I don't think it's valuable to put that way of learning in the same mental category as Anki and Supermemo.

One of your studies says:

After 2 days, initial testing produced better retention than restudying (68% vs. 54%), and an advantage of testing over restudying was also observed after 1 week (56% vs. 42%).

If you are looking at driblling or throwing free throws, I don't see a clearly distinguished testing from restudying.

See in context

17 Want to help me test my Anki deck creation skills?

by atucker

22nd Jan 2013

1 min read

17

I'm interested in trying to make better Anki decks for the LessWrong community, but I want to see how well I can actually do this first. There's a lot of knowledge out there about how to format and create decks, but it's still a decent amount of work, and there are lots of people who would benefit from Anki decks, but who wouldn't make them themselves.

In order to test my deck-creation skills, I'd be willing to do a summary + deck of a chapter or two of a book, then release them to the community for feedback.

I have two questions:

Who's interested in/willing to evaluate the decks? If there are enough volunteers, I'd also be willing to try different deck-making approaches.
What book/chapters would people like to see covered? I'm currently thinking of trying to do Eat That Frog or some similar book with a lot of recommendations and useful details. I don't really think that a math-heavy book would be well-suited to this, at least now.

These links are fairly useful/relevant.

http://www.gwern.net/Spaced%20repetition
http://www.supermemo.com/articles/20rules.htm

As a side note: I still do intend to do this, but have been fairly busy with the start of this semester. If there's no progress by March 1st, then you should consider this to be on indefinite pause.

Spaced Repetition

Personal Blog

17

New Comment

33 comments, sorted by

top scoring

Click to highlight new comments since: Today at 1:58 AM

[-]PECOS-912y110

59 seconds has many experimentally-supported recommendations and useful details which are mostly distinct from each other, which seems ideal for spaced repetition cards. I much preferred it over Eat That Frog, although I only skimmed the latter.

[-]Dorikka12y40

Here is the link to a 59 Seconds deck that I created a while ago. I think it's missing some stuff on parenting and attraction (and probably more), though -- just though that I'd post it in case it was useful.

I haven't used SRS in a while, so I haven't upgraded to Anki 2 -- that's why the deck is in my Dropbox instead of shared through Anki. If it doesn't work, let me know.

[-]ChristianKl12y00

I second the recommendation of experimentally-supported self help. Everyone would profit from knowing more about it.

If the deck is well made and the topic is interesting I will integrate it into my main Anki deck and give feedback to everything I consider suboptimal.

(I'm using Anki for ~1.5 years with 8500 cards in my main deck)

[-]jsalvatier12y30

I'm curious and willing to help you/encourage you.

Finding a good book or chapter is important I think.

Also this is Eat That Frog!

[-]ChristianKl12y20

I don't have an issue with reading self help authors like for motivating yourself. When it comes to learning material via Anki I would however prefer books with back up their claims with scientific evidence.

[-]Jolly12y20

Ping - did this ever happen? I'm looking for good examples of anki decks for book summaries before I start creating my own.

[-]Pablo12y20

[-]Dorikka12y20

At one point, I was trying to make a deck with emotionally significant examples of different biases/failures of reasoning. I was also planning to include some miscellaneous rationality-related stuff -- Milgram experiment to highlight strength of conformity effects, etc. There are loads of these in the Sequences.

I was working on this a year or so ago, but I lost the relevant file in a computer crash and most of my motivation with it. :(

[-][anonymous]12y20

A World at Arms is one of the non-technical recommendations from the list of best textbooks. History seems a good candidate for SR. Presumably the global approach of the book means it wouldn't be too redundant with prior knowledge of WWII history.

A more useful book would be Starting Strength which one would benefit from memorizing to maintain good form in weightlifting. But as other concrete recommendations it probably won't attract the largest audience for your experiment.

[-]ChristianKl12y00

History seems a good candidate for SR.

Why? Understanding historical events isn't primarily about remembering dates. SR also always you to learn information that more complex than just dates. There's nothing that makes one field like history special.

Presumably the global approach of the book means it wouldn't be too redundant with prior knowledge of WWII history.

What is that argument supposed to mean? Reducance isn't an issue. If you already know something you just hit a few times "Very Easy".

[-][anonymous]12y00

I believe that, for a student, history is more about acquiring knowledge than skill. And SRS works better for the former. Where am I wrong?

Are all fields equally approachable with SRS? I am not questioning the spacing effect, but the adequateness of the software.

I agree with your second point, the inconvenience is trivial. I was generalizing from my own preferences.

[-]ChristianKl12y00

I believe that, for a student, history is more about acquiring knowledge than skill. And SRS works better for the former. Where am I wrong?

That's true. You can't learn to play basketball by using Anki. On the other hand in most scientific fields knowledge is important.

[-]gwern12y00

You can't learn to play basketball by using Anki.

Literally speaking, yes. You could learn basketball history, statistics, official rules, and maybe things like classifications of tactics or something via Anki, but not basketball itself.

More generically speaking, maybe. A few days ago I looked into the spacing effect on motor skills question : http://www.gwern.net/Spaced%20repetition#motor-skills I don't have fulltext for all the citations yet, but it looks like there's something there (even if it's not as strong as spacing effect on declarative memory or language).

[-]ChristianKl12y00

You can play basketball well if you repeat a dozen core motor skills with high precision 10,000s of times.

It interesting to ask how you best spread out those 10,000 repetitions but I don't think that Anki helps with that goal.

There might be other motor skills were a high number of repetitions aren't central where SRS is more applicable.

[-]gwern12y00

You can play basketball well if you repeat a dozen core motor skills with high precision 10,000s of times. It interesting to ask how you best spread out those 10,000 repetitions but I don't think that Anki helps with that goal.

If you want to spread out your practices, you're going to have to start somewhere. The Supermemo algorithms are as good a starting point as any unless you're willing to hit the stacks and compare the musty motor skill studies head to head.

[-]ChristianKl12y00

Spreading out 20 times of practice is a whole different problem then spreading out 10,000. In addition it's not clear what counts as "correct" answering and forgetting something.

[-]gwern12y00

In addition it's not clear what counts as "correct" answering and forgetting something.

Not inherent to the effect; you can get the spacing or testing effects without providing the right answer or measuring the response.

[-]ChristianKl12y00

Not inherent to the effect; you can get the spacing or testing effects without providing the right answer or measuring the response.

I think it's very inherent to the supermemo algorithm. Otherwise how does learning without spacing looks like?

[-]jsalvatier12y00

I think this got derailed because gwern is trying to distinguish between the spacing effect the testing effect, and Supermemo/Anki. See gwern's literature review for clarification. In a domain where what counts as 'testing' isn't clear, perhaps basketball, then you might just try to used spaced repetition, which is what gwern is suggesting I think.

Space and testing effects are different, but work well together, and Anki/Supermemo are software that try to take advantage of these effects.

[-]gwern12y00

[-]ChristianKl12y00

One of your studies says:

After 2 days, initial testing produced better retention than restudying (68% vs. 54%), and an advantage of testing over restudying was also observed after 1 week (56% vs. 42%).

If you are looking at driblling or throwing free throws, I don't see a clearly distinguished testing from restudying.

[-]gwern12y00

If someone takes a weekly class that's teaching dribbling every monday, then you could call that "spacing learning". I don't think it's valuable to put that way of learning in the same mental category as Anki and Supermemo.

Why not, when they're very similar and may be exploiting the same underlying neurological effects (we don't really know the basis)?

[-]ChristianKl12y00

when they're very similar and may be exploiting the same underlying neurological effects (we don't really know the basis)?

Begging the question? There no possible way to learn a skill that requires 10,000s of repetitions without spreading the practice over time. Out of the many ways that people train basketball you don't gain additional information when you get to know that someone learns basketball by learning on multiple days.

If someone tell you that he's learning vocabulary via SRS with my more narrow definition that tells you a lot about the way he learn it.

[-]gwern12y00

There no possible way to learn a skill that requires 10,000s of repetitions without spreading the practice over time.

What on earth are you talking about? You can vary the spread over time - you can cram with 4 hours of practice on one day a week, or 1 hour a day for 4 days a week, and you can distribute a skill differently within sessions too in small random blocks or again cram all of one practice task into one time period.

I get the feeling you're not even trying to understand here.

[-]ChristianKl12y00

You can vary the spread over time - you can cram with 4 hours of practice on one day a week, or 1 hour a day for 4 days a week, and you can distribute a skill differently within sessions too in small random blocks or again cram all of one practice task into one time period.

As far as I understand you both of those methods would be in the same category of spaced repetition.

[-]gwern12y00

Begging the question? There no possible way to learn a skill that requires 10,000s of repetitions without spreading the practice over time.

As far as I understand you both of those methods would be in the same category of spaced repetition.

I just gave you 2 different ways, on different time-scales, that spacing could be applied to motor skills and to getting the most out of the 10,000s of repetition. If you visualize the spacing effect as being based on the forgetting curve, it should come as no surprise at all that you can demonstrate spacing on many time-scales (and my spaced repetition page includes citations dealing with intervals ranging from seconds to years).

[-]jsalvatier12y00

Above studies suggest it is valuable.

[-]ChristianKl12y00

Above studies suggest it is valuable.

Do judge whether that's true we need to find an agreement about what we mean with it.

[-]marchdown12y10

Aye. If you need another nudge, I'd like to say that it's a great idea, and yes, I would help you test resulting decks.

[-]thelomen12y00

Same here, I'd gladly test it. Also read the two recommendations 59 seconds and Eat that Frog after reading this thread, having fallen head first into the Anki rabbit hole about a month back.

[-]John_Maxwell12y10

I dunno about a book of concrete recommendations... it only makes sense to study them, IMO, if you've precommitted to following them. And you might even find that some of them don't work for you.

[-]aelephant12y10

I'm not sure how much time I will have to donate, but I build my own decks & so I feel I have a decent grasp on what makes a good card vs. a bad one. I might be willing to look at some of your cards & give you some concise feedback.

[-]JMiller12y00

I am not familiar with spaced repetition, but I keep hearing about it. I'd be happy to help if you need a complete newb!

Moderation Log

Curated and popular this week

33Comments