Siblings Should Rank Up/Down Separately

In Anki, siblings grade independently. Say, for example, you have a recall and recognition card for the kanji 軍 (army). If I see 軍 and write “army,” I get the recall card right, and it levels up. If I see “army” later, and I write the wrong kanji, the recognition card levels down. Recall and recognition are independent: my recall card levels up regardless of me getting the recognition card wrong.

This is a much better system than the current “fail one, fail all” system employed by Kitsun.io. Recall and recognition are two different skills – why should I continuously review recall for a word when I’m only struggling with recognition, or vice versa? It needlessly slows down SRS progression for things I already know, and gums up the review queue with twice the work.

Independent sibling grading should at least be an option. That said, I’m a Kitsun.io newbie, so if this sort of behavior already exists and I just haven’t noticed it, or if I’m totally misunderstanding the system, feel free to throw something at me.

2 Likes

You are able to create cards the way you would want to: E.g. having a card with just a recall layout and a card with just a recognition layout. Or like in your current situation a card with both recall and recognition layouts, causing it to create two sibling cards.

As for community decks, those cards are made by the author, so they kinda dictate the learning style. If you don’t wish to do certain layouts, you can completely filter them out through the “Layout filtering” deck settings.

I agree with you on the thought that recall and recognition are two different aspects, and I personally split those into separate cards or filter out one of the layouts.

Another option for community decks might be to filter one layout first, and then redo the deck with just that layout unfiltered after you are done with the deck the first time, but that kinda seems like a hassle :sweat_smile: .

2 Likes

@Neicudi: reviving this thread because I’ve now seen several other people across the forums mentioning this, plus at least one other person over on the WaniKani thread.

As for community decks, those cards are made by the author, so they kinda dictate the learning style.

I don’t believe that authors should dictate learning style. Authors do us a great service creating decks, but everyone learns differently. Do we really want everyone who believes that recall and recognition are two different aspects to have to make their own multi-thousand card decks instead of using a perfectly good community deck?

I agree with you on the thought that recall and recognition are two different aspects, and I personally split those into separate cards or filter out one of the layouts.

Filtering out a layout reduces reviews, but you lose the learning you would get from completing that layout. We agree that recall and recognition are two different aspects: why deny yourself the practice on one of those critical skills?

Another option for community decks might be to filter one layout first, and then redo the deck with just that layout unfiltered after you are done with the deck the first time, but that kinda seems like a hassle.

That definitely is a hassle! :sweat_smile: For one, it doubles the amount of time it would take to “complete” a deck. God help you if you’re doing something like Core 10k with that method…you’ve effectively turned it into the Core 20k!

I believe this is one of the key features Kitsun needs to be able to go head-to-head with its competitors. If you’re dealing with multi-thousand card decks – or multiple multi-thousand card decks, as many of us are – efficiency is key, and indeed, that’s the entire appeal of SRS systems. Per Kitsun’s own Knowledge Base:

The way SRS works is that you will get reviews whenever the item you are learning is almost gone from your memory. The brain gets a reminder which strengthens the memory of the item and kind of tells them that the information you are learning is important. This has been proven to be a very efficient way of studying and retaining the learned information in the long run.

By not ranking siblings independently, the whole system is handicapped. There are now two times as many reviews in the pool (for power users, that could equate to hundreds of reviews), and as many as half of those reviews are potentially showing up long before the user “is about to forget them.”

Not only is this inefficient from a reviewing/academic perspective, it greatly increases the likelihood of user burnout. Take it from me: I burned out a few days into my free trial because of this, and only just came back for another go. I love Kitsun, but thanks to the current system, I have to go 50% slower then I did on Anki to keep my reviews to a manageable level (i.e. the same amount I had per day while using Anki). This is clearly a flaw.

Is it possible to implement independent sibling ranking as a per-deck user option at least? That way people who want the feature can use it and people who don’t aren’t forced to.

At least the level system of Kitsun wouldn’t make much sense with independent siblings.

Anki just doesn’t use siblings each is its own card.

The whole system is currently built around having a progress document per card, rather than a progress document per card, per layout. Deck authors make the decision to group certain layouts together or not. It is possible to create a card per layout, and therefor effectively a progress document per layout. But in most cases this would be overkill.

I think Kitsun already offers the options most people need in order to manage their lessons/reviews, by giving them options such as filtering siblings, delaying siblings by X hours and the abovementioned creation tools (for personal decks).

That does not mean that I don’t understand where you are coming from, but I also don’t think I’d neccesarily count it as a flaw of Kitsun, rather than a flaw of a deck.

Sadly it’s really not as easy as to just adjust the progress system as it is the basis of how Kitsun works. There’s a ton of things that would break and I’d have no good solution for.

I’d have to think about it further, but right now I don’t see a good solution without effectively splitting all layouts of a card into multiple cards.

Just to give an example:
That would turn the 10k deck from a 20k card set into a 50k card set (5 layouts), along with progress documents getting split into 5 different progress documents and so on… This would then also cause new problems such as having to manage the content of 5 different cards which actually use the same values. Those are card bound, so there would have to be a new system where card values are separated from the cards, and put into their own collection.

Of course there are other (perhaps smarter) ways to do this as well, but as far as I can think of, they all have major drawbacks/problems that’d need to be solved.

1 Like

Reviews don’t necessarily go up with the siblings delay. You just do them at different times. To me I see it as twice the exposure for the same workload. Sure, visually it might look overwhelming (100 looks much nicer than 200), but you can also do them in half the time. If anything, 100 reviews with delay activated feel really nice to do. In less than 10 minutes and they’re done.

At least that’s my experience with this feature :slight_smile:

1 Like

Along the lines of delaying sibling cards, how long of a delay have people found to be reasonable / useful? (jprspereira, I’m mostly asking you… :sweat_smile:) I’m very early on here, but so far with delay set to 24h, it seems like it’s taking FOREVER to level anything up.

As an SRS user through Wanikani and Bunpro, I also worry about recall and recognition being too closely linked, although it sounds like they may be very difficult to separate. Let’s say you get the RECOGNITION layout first (RECALL second) 4 times in a row, and each time your brain only has to be able to do recall it from however long ago your sibling delay is set to (several hours). The next time it comes up for review (may be months later), let’s say RECALL comes up first, and now you have to recall it from several months ago rather than several hours ago. Doesn’t that at least partially mess up the effect of the SRS system?

2 Likes

My personal opinion on this is that it’s not that big of a deal. When learning a language and actually using it (e.g. consuming media) you are bound to come across words that might be in your SRS queue in one way or another. With the same reasoning you could say that this also influences your SRS, but I believe it’s what’s supposed to happen, if anything it reinforces the word already in your mind.

After using SRS for a few years you come to realize that SRS is not the holy grail, but it is the most effective way to build up your knowledge until you naturally start acquiring more vocabulary through context.

Mixing SRS with actual consumption of media is the most effective way of learning for me, and because of that I really would not sweat the details that much. But that’s just my opinion on this subject and I’m sure others might disagree with me on this :smile:

As a disclaimer: I personally don’t use the sibling delay feature.

That’s my experience too, @takanuva. I’ve been heavily testing this feature for a couple of months myself, and even though theoretically it doesn’t seem perfectly designed (yet), my recall of Japanese words overall has improved tremendously. It’s the best solution that I’ve found so far when it comes learning recall and recognition together.

I’ve only tested with 23h because by postponing it to the following day, I know that it won’t influence today’s load of reviews. This way, I can more or less predict how much time I’ll dedicate to SRS the following day. So in a way, it’s a very practical timing.

Nevertheless, I talked with @Neicudi to allow us to choose different times for different SRS levels (instead of just a fixed one), which will pretty much solve all the possible imperfections that this feature might have. He seemed to approve that idea :slight_smile:

Different delay times for different SRS levels is a very interesting idea… but it is a bit scary to think about: let’s say the delay is set to 4months for burn reviews. If you have all layouts active on the Core 10K, that’s 4 different layouts, each 4 months apart (16 months :flushed:), and if you miss any of them, you’re basically back to square one…

It wouldn’t work like that though. The intervals between siblings would be different (and much lower) than the main intervals of the SRS :slight_smile: