I’m confused about why the SRS system handles my mistakes the way it does. The WK Knowledge Guide says:
Instead of skipping past a review that might be difficult to remember, we’d rather you spend a bit of extra time and try to answer the question. We know it can be frustrating, but did you know that the act of struggling to recall something can actually strengthen your memory?
Okay, I get this. It makes sense to me. However if I get it wrong more than twice, my SRS level goes down even further. If it’s good for me to struggle to remember, why would evidence of that struggle — multiple attempts in a single review — lead to a need for more frequent SRS checks in the future over my just looking up the answer immediately?
I’m glad I searched for the algorithm though. Now at least I’ll allow myself a second attempt before looking it up.
If I’ve forgotten it, I’ve forgotten it, and I need to go back. No arguments there. My question is only about by how much. Multiple attempts in the same review don’t show that I forgot it multiple times!
If I just look at the answer immediately after I make a mistake, although the Guide implies that’s not as good for my learning, I only go back 1 or 2 SRS.
If I struggle and try it again a couple of times, the Guide tells me that’s good for my learning. But I get sent back double, so 2 or 4 SRS.
Why would I need to go back further having engaged in good and recommended learning behaviour? Isn’t that counterintuitive?
I think their reasoning is that reviews gauge your mastery of the entry.
If I’m faced with the entry and I don’t know it, it’s demoted. But if I see it again in the same review session, having already been marked wrong and I still don’t recognise the kanji, or still don’t know the meaning, WK judges my retention of that word to be very shaky indeed. One demotion isn’t enough - they judge I need to see that word a bunch more, so it’s demoted even further.
There is “oh right, that’s what it was” forgetting, and “uh… I learned that in a haze and remember nothing about it” forgetting, and WK tries to respond to that.
I’m not saying it’s a perfect system or anything, just that that seems to be the WK reasoning.
Some SRS algorithms function the way you expect, for instance the newer FSRS scheduling of Anki only takes the first review of the day into account for a given card, further mistakes don’t reschedule the card.
Other schedulers on the other hand go in the opposite direction and reset the card to the lowest SRS tier in the first mistake, erasing your progress.
It’s fairly arbitrary overall. WaniKani’s implementation is rather blunt given that it has no concept of “ease” and similar per-card difficulty attributes, just fixed intervals that never change. It does the job though.
I’m not sure that the guide is implying this. It says that spending a little while trying to recall something is more helpful than giving up immediately if it’s not instant recall. It doesn’t say anything about what you should do once you’ve tried to answer once and got it wrong.
Personally I think that if you answer and get it wrong you should look at what the right answer is, to refresh your memory. Guessing twice seems unnecessary and counterproductive. (I don’t use WK; the two SRS systems I do use, Anki and jpdb, both automatically show you the full right answer for each review, so you don’t even need to go looking for it.)