The goal of SRS is to make sure that you remember all the things by heart and bring those things up for revision which are on the verge of getting forget, but since every element (radical, kanji and vocabulary) has fixed number of times it will come up with fixed interval, it makes it painfully slow and lot of us lose the motivation.
Right now the time intervals are as follows:
Apprentice 1 → 4 hours → Apprentice 2
Apprentice 2 → 8 hours → Apprentice 3
Apprentice 3 → 1 day → Apprentice 4
Apprentice 4 → 2 days → Guru 1
Guru 1 → 1 week → Guru 2
Guru 2 → 2 weeks → Master
Master → 1 month → Enlightened
Enlightened → 4 months → Burned
Now let’s say I took a new lesson which had 10 items and after that I got busy for a week so couldn’t do any reviews and lessons. When I come after that and take the reviews, I get 5 items correct and 5 wrong. I think it is not wrong to say that I learned those 5 correct items to the level of Guru, so instead of upping the level only up to Apprentice2, those items should become Guru2. Similarly, if I come back after a month and can get many items correct, those should become Enlightened.
It will speed up the progress by a lot as some items which we have truly learnt will not block the queue. It will also make WK worthwhile for those who already have N3/N2/N1 or just know a lot of kanji beforehand, and also those who had to reset WK because they left it for a long time.
3 Likes
I mean, it’s a moot point as it won’t be implemented, but in the interest of philosophical discussion, wouldn’t this incentivize doing reviews less often in order to skip levels?
4 Likes
Skipping reviews will only be effective if you have really learnt the items, otherwise you will keep getting it wrong and move at intended pace.
2 Likes
This is how jpdb handles it iirc.
It makes sense with how the SRS algorithm is supposed to work, but I agree there’s slim chance wk actually implements it.
7 Likes
I agree that speeding WaniKani is highly desirable. Not enough people complete WK, and too may seem to get snowed under with reviews and give up. Why wouldn’t you want to reduce the WK effort if it leads to the same learning outcomes? Whether this particular suggestion is the right way to go ahead, I don’t know.
I’d like to see WaniKani Experimental, a separate program people can optionally volunteer for, which serially implements a few of the speed-up/effort-reduction methods that get suggested by the community. If, after say a year or eighteen months of testing, a method shows that it provides better learning than Wanikani standard, the innovation gets transferred the standard program and the next innovation gets tested. Better means more kanji are being burned in the same time frame, fewer people (a lower proportion) dropping out, and more people reaching their learning goals.
For example, I’d be fascinated to know if the multiple SRS-level-drop when you are wrong post-Apprentice is really a good idea. I’m not even sure if you should ever drop below Apprentice 4 once you’ve reached Guru. I’d like to know how much worse the learning if participants answered both meaning and reading as part of the same review question (a souped up version of back-to-back); I’m convinced it would speed reviews. People are crying out for more control over reviews, but there seems to be a precious complex over the current SRS process that ensures change never comes. Look at the first two reactions on this thread: no matter how good the idea, they said, it will never be implemented. If innovations are limited to those which increase cognitive load or just workload (such as kana vocabulary) in the main stream, some problems with WK will get worse not better.
I applaud the effort to continuously improve the data sets, with regular updates coming out on Community. An experimental strand would also allow process improvement ideas to be trialled with a reasonable volunteer sample, while ensuring that less good ideas remain isolated from traditional, less adventurous WK learners. It might encourage change even in the traditional, less adventurous WK administrators, for whom the current SRS parameters appear sacrosanct.
4 Likes