Iteration cadence for spaced repetition experiments

By the very nature of spaced repetition, any experiment that tries to test some aspect of it will take months to complete. This is because in order to see how much you remember something, you need to wait for long enough to have forgotten it (if not for some intervention). This slow feedback cycle presents a challenge for those wishing to significantly improve their spaced repetition routine. To see if a change improves your ability to recall something, or to test out some cool new idea, you have to wait for several months to see if it worked.

This is also one of the things that makes academic work on spaced repetition not very useful. Most of them do very short-term reviews.

expect iteration time of 3-4 months: that's how long it takes to start to really forget things you've read about, which is when you can start to actually see what you retained vs what you didn't, so that you can change how you do things. but 3-4 months is only if you come back to review things to see what you've retained. if you don't use things, it could be much longer until you realize you've forgotten things. luckily, with anki, if you put in some low-effort cards, you will be forced to review something in the 3-4 month time period, where you get to see what you retained. this is how i found out that i wasn't retaining some things in analysis, which is how i eventually figured out i needed to put in problems to solve.

people praise fast feedback loops, but one thing to realize with spaced repetition software like anki is that the feedback is spaced in a way. you can't tell if you're memorizing things well until it's like 5 months after you've added the card. or take my series test cards; i didn't realize the subtleties of requiring the sequence to be nonincreasing or nonnegative for different tests until a couple weeks after i'd added the cards -- before now, everything would sort of have been in my "fresh" mind. it's only after i've forgotten the context sufficiently that i can appreciate these subtle hypotheses in the theorem statements -- and only because i've added these as cards! makes you wonder, what subtleties am i missing by not adding something to anki and seeing in what manner i forget the details.

on the other hand, some feedback is immediate. if your card is too big for example, you will notice on that same day, or a couple days later.

I'm not sure if there is a way to speed up the feedback you get.

I'm feeling this a lot now when trying out incremental reading on SuperMemo. Even after a month or so of IR, it's very hard for me to tell how much more efficient it is. The final steps, like going from the extracts to cloze deletions or q&a cards, don't really happen until you've seen an article multiple times, which might take a few weeks if you've imported many articles. So you can't even see the full benefit of IR in the first few months of using it. Incremental/spaced things are easier to use but harder to assess.

See also

• Spaced repetition allows graceful deprecation of experiments