Another type of card is a "situation/response" card, where the front is some situation you might be in and the back is the response that you want to remember to execute. For example:
I have collected a lot of these type of heuristics, if others are interested.
It feels dishonest to share them because I don't use them myself. I lost a lot of my enthusiasm for life when I got stuck communicating with my computer using voice recognition.
I know you wrote this comment nine years ago, but if you're still willing to share, I'd love to hear more of these heuristics! I'm hoping to build a deck for doing cognitive-behavioral therapy on myself or something similar, and what you're talking about has a lot of overlap with that. So please share.
Can we design an experiment to find out whether these types of cards produces decent results?
Piotr Wozniak (the father of SRS) has some important thoughts on these things on his site. Check out especially the articles (for instance) and the FAQ.
There's also a list of other posts on this on the wiki.
For another idea, see my post on importing books whole. This is also useful for other things worthy of deep understanding, like certain blog posts ;)
It seems to me that the spaced repetition approach (and other "flashcard" style methods) is especially effective for memorizing large amounts of information, where the links between question and answer (or for language learning, written - spoken - translation) are somewhat arbitrary.
When it comes to more general kinds of issues, such as the examples you provide above (critique the logic, car purchase decision making) it feels like a more ideal learning process is not so much about learning the association between somewhat arbitrary labels, question/answer pairings, but rather learning to generalize to novel situations. Further, preparing the optimal answer for a case like car-buying decision making is not nearly as straightforward as the known answer to a question (or a word and its definition, or translation equivalents, etc).
Studying algorithms is somewhere in between - learning the pairing between an algorithm and its name is definitely the sort of task in which Anki and other such methods can be very useful. But I wonder how much your understanding and generalization of the algorithms themselves is actually benefited by the spaced repetition elements of your study program. Of course, your approach, in which you elaborate the flashcard based contents (e.g., considering how algorithm Y may apply to a current problem you are working on) can indeed provide greater understanding - I'm just not sure whether that is dependent on spaced repetition.
Seems to me like it would work; but to learn more general ways of thinking instead of specific facts, you'd need to have a class of questions on a card, instead of a single question. Essentially, we'd need to change the anki software to allow a type of metacard, with fairly flexible rules for generating question-answer pairs.
My experience has been that thinking through specific algorithms, especially in graph theory, has caused graphs in general to be easier to work with -- which is to say that I expect rationality exercises like these would probably work. I will try to brainstorm ways to test this in the spirit of TrE's comment.
Another way to accomplish this goal is to edit the card manually every time you see it, but then again, that takes quite a long time for each card.
Teaching procedural knowledge seems possible, from my personal experience (beware of generalizing, the usual disclaimers apply). Learning for my A-level, I have done well doing such things. In chemistry, I learned many reaction mechanisms, but when I looked at the cards, I didn't think of the specific solution, but rather looked at what special functional groups there were and how they would react. I felt like I had to rediscover the solution over and over again each time I looked at the card. Of course, at a certain point, I simply remembered the cards and their solutions, so having multiple cards for teaching the same skill seems kind of sensible.
As I said, this worked quite well for me, I felt the exams were quite easy. I'm quite sure flashcards are about as good at teaching skills as they are at teaching facts, if done correctly. Whether or not procedural knowledge is more important than factual knowledge is a different question, but I'd assume so.
Recently I have started using Anki in a new and complimentary way. I am curious if any of you find it similarly useful and/or have other anki tips :)
The basic idea is that instead of putting down a challenge/response pair for facts, we put down a challenge/response pair for ways of thinking. A train of thought. Ideally, this is something akin to a lumosity.com game except tailored to your area of focus.
A simple example from algebra:
The fact based approach would be to make a card titled "what is the quadratic formula?" with the answer of "x == (-b +- Sqrt(b^2 - 4ac)) / 2a"
The way I am recommending is to make a card titled "derive the answer for ax^2 + bx + c == 0" with an answer that shows the steps. When shown the card, you would then either solve it in your head, or using a pad of paper. I assume that sub minute tasks are ideal.
The specific area I have been using this in is the study of algorithms, with challenges like "Hopcraft-Karp algorithm for bipartite matching", and it has so far proved very helpful at getting fluent with the names, deepening my understanding of the algorithms themselves, and with seeing new places to apply them in my coding.
This might be overstepping, but something like this seems like it might be appropriate for The Center for Modern Rationality. Something like "critique the logic of the following three sentences", or "Sue is about to buy a car. How should she go about making a decision".
This is my first real post to LessWrong, so if you have style corrections those are solicited alongside any comments on the post itself. Thanks!