Registering my continued dislike for dedicating effort on this feature (I voiced similar complaints when this was originally floated). My reasons:
retaining knowledge via spaced repetition is a poor fit for most of the posts on LW
the general advice is that using cards written by others is much less useful than cards you write yourself
nudges LW towards encouraging posts that are more amenable to space repetition, which I expect will marginally decrease other kinds of content
implicit in this is an argument that LW would be marginally worse if it spent more time on teaching things that you can learn by remembering them and less time on other things, like working towards reducing confusion about things we are currently confused about
When I originally floated the idea on Facebook, I was considering implementing it as an LW feature directly, which would’ve been a lot of effort. Then it turned out ThoughtSaver was already building it, with cross-site integration in mind. The actual implementation on LessWrong works the same way as the Metaculus and Elicit integrations we already had, and took less than an hour on our end. (Much more than an hour of effort on ThoughtSaver’s end, obvciously, but amortized across more applications than just LessWrong integration, and that part of the effort didn’t come from the LessWrong team.)
Fair enough, though I disagree with these points. For one thing, this feature already works, so I don’t expect it will take any significant amount of time from the LessWrong dev team.
But to respond to your more specific points, from my perspective:
(1) there is a common misconception that flashcards and/or spaced repetition is mainly useful for facts and definitions. It is also very useful for concepts (“a generative adversarial network is...”), takeaways (e.g., “the most important three points this post makes are...”), connections between ideas (“X and Y are thought to be different but they are connected through Z...”), strategies (“Here’s a process you can use...”), and so on. I certainly don’t think all LessWrong posts are suitable for this, but I think you’re underestimating how many are. My guess is it is a good fit for 5%-20% of posts but that’s a super rough estimate. I personally use flashcards for complex concepts all the time, including plenty I’ve created for myself from LessWrong and Slatestar Codex posts.
(2) indeed it is a standard belief that if you pick up someone’s random flashcard deck that they wrote for themself you are unlikely to get much out of it (I agree with this too). That is different from a deck written specifically by the author for an essay (almost nobody has experience with this, other than through the experiments I link to in the bottom of the post) so the “I can’t pick up my friend’s deck” argument doesn’t have much weight in this case in my view.
(3) I would argue that even posts that e.g., “work towards reducing confusion about things we are currently confused about” have takeaways that are worth remembering (e..g, “what should you think about differently having read the post?”, “what was the approach of reducing the confusion?”, “what is a useful analogy for thinking about this topic?”, etc.)
These are all quite reasonable, and I’m pretty open to the idea that I’m mistaken and anchoring too much on the fact that I didn’t find flash cards for spaced repetition useful, which might ultimately bias how heavily I weight things or assess the likelihood that flashcards would be helpful.
The process of making a piece of knowledge suitable for space repetition is a lot about making the piece of knowledge as clear as possible.
When it comes to my own writing, I don’t think it discourages me to write about ideas I believe are worth writing about that this new feature exists. I might find it interesting to experiment with a new way of writing posts and use the feature, but I don’t believe that this is zero-sum.
Asking “Is this likely to be useful?” is the wrong question. I think it’s better to ask: “Is this an experiment that has the potential to produce something useful for the art of rationality?”
Trying out new ways to represent knowledge about the art of rationality has the potential to lead to progress and is thus worth doing. We shouldn’t try to focus on one approach of gathering knowledge about rationality but try many.
The general advice is that using cards written by others is much less useful than cards you write yourself
As someone who has used SRS on a regular basis for multiple years, I disagree with this. Sure, there are times when my best option is to hand-craft cards for myself, but this is usually a function of there not already existing high quality cards that test me on the things I want to learn. Creating cards is useful for allowing me to focus on exactly the facts that I want to learn; the other times where this is true is when I have a specific model and vocabulary for understanding a field that isn’t standard; there, creating cards allows me to target the specific vocabulary and ontology that is unique to me.
However, often times I want to learn something that doesn’t meet these criteria: things that I can easily think about using a standard ontology and vocabulary, and are common enough subjects that someone has already taken the effort to make a sufficiently high-quality deck. In this case, making my own cards is just a waste of my time (and the time that is wasted isn’t negligible; I’ve spent up to an hour on multiple occasions creating cards for domains that could just as easily been downloaded if I knew of a high quality deck that already existed). I’m using premade decks for learning pronunciation and meanings of Japanese Kanji and for mapping flags to the countries they represent. There are many subjects where I have cards I made myself that I would be just as well off using equally high quality cards made by another person, and others would be well-served using the same cards I made: learning the pronunciation and meanings of Russian words, mapping coordinates to the cities they locate, memorizing poetry, mapping astronomical symbols to the bodies they represent. Even in the case of mathematical definitions and associations, while the cards I use are tailored to me, similar cards could certainly be made that would be suited to a much wider audience, that would be nearly just as useful as the cards I actually use.
I’d estimate about 2/3rds of the cards in my Anki deck are too specific to me to have been mass produced, but the other 1/3rd would be very well justified in having been made with the intention of being able to be used by others, and that ratio could be even higher (maybe up to 2/3rds mass produceable) without causing problems.
Registering my continued dislike for dedicating effort on this feature (I voiced similar complaints when this was originally floated). My reasons:
retaining knowledge via spaced repetition is a poor fit for most of the posts on LW
the general advice is that using cards written by others is much less useful than cards you write yourself
nudges LW towards encouraging posts that are more amenable to space repetition, which I expect will marginally decrease other kinds of content
implicit in this is an argument that LW would be marginally worse if it spent more time on teaching things that you can learn by remembering them and less time on other things, like working towards reducing confusion about things we are currently confused about
When I originally floated the idea on Facebook, I was considering implementing it as an LW feature directly, which would’ve been a lot of effort. Then it turned out ThoughtSaver was already building it, with cross-site integration in mind. The actual implementation on LessWrong works the same way as the Metaculus and Elicit integrations we already had, and took less than an hour on our end. (Much more than an hour of effort on ThoughtSaver’s end, obvciously, but amortized across more applications than just LessWrong integration, and that part of the effort didn’t come from the LessWrong team.)
Fair enough, though I disagree with these points. For one thing, this feature already works, so I don’t expect it will take any significant amount of time from the LessWrong dev team.
But to respond to your more specific points, from my perspective:
(1) there is a common misconception that flashcards and/or spaced repetition is mainly useful for facts and definitions. It is also very useful for concepts (“a generative adversarial network is...”), takeaways (e.g., “the most important three points this post makes are...”), connections between ideas (“X and Y are thought to be different but they are connected through Z...”), strategies (“Here’s a process you can use...”), and so on. I certainly don’t think all LessWrong posts are suitable for this, but I think you’re underestimating how many are. My guess is it is a good fit for 5%-20% of posts but that’s a super rough estimate. I personally use flashcards for complex concepts all the time, including plenty I’ve created for myself from LessWrong and Slatestar Codex posts.
(2) indeed it is a standard belief that if you pick up someone’s random flashcard deck that they wrote for themself you are unlikely to get much out of it (I agree with this too). That is different from a deck written specifically by the author for an essay (almost nobody has experience with this, other than through the experiments I link to in the bottom of the post) so the “I can’t pick up my friend’s deck” argument doesn’t have much weight in this case in my view.
(3) I would argue that even posts that e.g., “work towards reducing confusion about things we are currently confused about” have takeaways that are worth remembering (e..g, “what should you think about differently having read the post?”, “what was the approach of reducing the confusion?”, “what is a useful analogy for thinking about this topic?”, etc.)
These are all quite reasonable, and I’m pretty open to the idea that I’m mistaken and anchoring too much on the fact that I didn’t find flash cards for spaced repetition useful, which might ultimately bias how heavily I weight things or assess the likelihood that flashcards would be helpful.
The process of making a piece of knowledge suitable for space repetition is a lot about making the piece of knowledge as clear as possible.
When it comes to my own writing, I don’t think it discourages me to write about ideas I believe are worth writing about that this new feature exists. I might find it interesting to experiment with a new way of writing posts and use the feature, but I don’t believe that this is zero-sum.
Asking “Is this likely to be useful?” is the wrong question. I think it’s better to ask: “Is this an experiment that has the potential to produce something useful for the art of rationality?”
Trying out new ways to represent knowledge about the art of rationality has the potential to lead to progress and is thus worth doing. We shouldn’t try to focus on one approach of gathering knowledge about rationality but try many.
As someone who has used SRS on a regular basis for multiple years, I disagree with this. Sure, there are times when my best option is to hand-craft cards for myself, but this is usually a function of there not already existing high quality cards that test me on the things I want to learn. Creating cards is useful for allowing me to focus on exactly the facts that I want to learn; the other times where this is true is when I have a specific model and vocabulary for understanding a field that isn’t standard; there, creating cards allows me to target the specific vocabulary and ontology that is unique to me.
However, often times I want to learn something that doesn’t meet these criteria: things that I can easily think about using a standard ontology and vocabulary, and are common enough subjects that someone has already taken the effort to make a sufficiently high-quality deck. In this case, making my own cards is just a waste of my time (and the time that is wasted isn’t negligible; I’ve spent up to an hour on multiple occasions creating cards for domains that could just as easily been downloaded if I knew of a high quality deck that already existed). I’m using premade decks for learning pronunciation and meanings of Japanese Kanji and for mapping flags to the countries they represent. There are many subjects where I have cards I made myself that I would be just as well off using equally high quality cards made by another person, and others would be well-served using the same cards I made: learning the pronunciation and meanings of Russian words, mapping coordinates to the cities they locate, memorizing poetry, mapping astronomical symbols to the bodies they represent. Even in the case of mathematical definitions and associations, while the cards I use are tailored to me, similar cards could certainly be made that would be suited to a much wider audience, that would be nearly just as useful as the cards I actually use.
I’d estimate about 2/3rds of the cards in my Anki deck are too specific to me to have been mass produced, but the other 1/3rd would be very well justified in having been made with the intention of being able to be used by others, and that ratio could be even higher (maybe up to 2/3rds mass produceable) without causing problems.
Seems like using cards by others is still better than not using any cards at all?
There’s an opportunity cost. If you spend a lot of time memorizing cards that don’t really fit into your map of the world that costs a lot of time.