With the improved algorithm in the works to be included with Anki, I think it would be a good idea to reconsider four-tier grading as the standard.
I want to be able to recommend the use of Anki with minimal modification to people as a gateway into using it, and the inclusion of FSRS is a good step to that (I consider it a major streamlining improvement that cuts down on scheduler jank and eliminates a big part of Anki’s unintuitiveness as well as just making it more effective) but four-tier grading is anything but intuitive and means you have to make up personal rules on how to grade your memory that you may or may not follow consistently and makes your reviews take more time to judge, and potentially leads to card misschedulings (one bad “Easy” on a high-interval card and you’re kinda F-ed). A good scheduler can get the idea if you’re spamming “good” on a card anyway, right?
Would it be a problem to have a more streamlined pass/fail system as the default with four-tier as opt-in for those who swear by it? I think it would really make Anki a lot more approachable and accurate by default for new users.
Thank you for reading my suggestion Sorry if this has already been suggested, I tried to search before posting at least.
I can’t rule out an option to hide the other buttons in the future, but I suspect it would somewhat hamper the scheduler’s performance, FSRS or no. There’s a reason space answers good by default - the assumption is you’ll use that most of the time, and use the other two buttons for the comparatively rare times that something was too hard or too easy, but not completely forgotten.
With present bult-in scheduler users can choose not to use Hard and Easy. Would it it hamper the scheduler’s performance?
@L-M-Sherlock
I invite you to this discussion. It is not a matter of removing Easy, Hard buttons or making them optional, but how Anki’s future FSRS scheduler will adapt to using only again, good buttons.
If you only use again and good, the optimizer has considered it. I recommend keeping your habit because the law of your review behaviors has been recorded in the review logs. Keeping self-consistency will make FSRS more adaptive to you.
If FSRS goes into Anki will there be a built-in optimizer? I know there is a discussion about it. If yes, that hampering should not be an issue.
From the point of view of ordinary user, as I am, that has little comprehension of math behind scheduler, discovering pass/fail method was revolutionary, because it helped me to decrease self-assessment time which amounted to decreased review time.
What I would expect from FSRS is to adapt to my personal memory patterns (that would be revolutionary) and let me focus more on learning.
Before pass/fail method I frequently caught myself on thinking whether to use Hard, Easy, meaning I was trying to guess, predict how my memory would work. My preference is that this should be left to the adaptive scheduler and I should focus on repetitions. Maybe I am wrong, but I assume that it is better for my final long term results if I spend time doing more repetitions than spend time on choosing which button should I press.
I am aware that other users may not find this process of taking decision whether to use Easy, Hard that troublesome or time consuming.
I suggest a middle ground. Make “easy” optional, but leave “hard” by default.
Back when I used the straight reward addon for scheduling, I decided to experiment with using it to replace the easy button. It didn’t take long for me to notice that grading reviews felt a lot simpler for me in the normal case. I found that before, every time I got a card correct quickly, I had to stop and think to myself if it was quick and confident enough for me to mark it as easy, which took a bit of time and effort and made those “quick” reviews not so quick sometimes. When I switched to straight rewards, and ignored the easy button entirely, that burden was completely removed. Now, when I got an answer correct quickly, I just clicked “good”. No further thoughts about it necessary. That made reviewing easier for me, and it actually seemed to increase the amount of reviews I was able to do each day by ~15%, at least according to the stats.
I think “hard” is a valuable rating to have because it’s useful to let the scheduler know when you got a card right, but you almost didn’t, so it can take that into consideration before you end up failing next time. This is important because failing a card is something that needs to not happen often. Also, because “hard” reviews tend to take longer and be more mentally straining than “good” reviews anyway, there’s not much extra effort in marking something as hard. Most of the time It’s obvious when something should be marked as hard.
With “easy” grades however, almost the same info could be inferred by the scheduler from you simply continuing to mark the card as good. Having the explicit option to say “this was too easy” is of course useful sometimes, but the problem is that I found myself always having to ask “was this too easy?” which slowed me down and made reviewing the easier cards more difficult than it should’ve been.
I don’t know enough about how FSRS works internally to know how important the easy button actually is to that scheduler, but my experience with “easy” when you use the default scheduler combined with straight rewards, is that the juice ain’t worth the squeeze.
And FSRS predicts that my memory stability when grading easy is almost twice of stability when grading good. And the stability will not change when grading hard.
How does it translate into user behavior? Does your data indicates that it is advisable for you to use Easy? The better memory stability means the better we remember, correct? So, it is not advisable for you to use Good for cards that are Easy because memory stability will decrease? I don’t think it is the correct conclusion.
Rather your higher memory stability when grading easy means that your subjective feeling that you know the content of the card is more correct than your subjective feeling that you know the content of the card when you grade ‘good’.
In other words more often when you grade yourself ‘easy’ you are more times correct in this self-assessment than when you grade yourself ‘good’.
But this doesn’t lead to the conclusion that using ‘easy’ improves memory. It only tells us that your self-assessment was correct.
It is neither a voice pro two key Anki nor against it. It is still a user preference where they want to spend more time: on self-assessment what key to hit or perhaps more repetitions when not using ‘easy’.
My conclusion is that if future bult-in Anki FSRS will have included optimizer, using two keys will not hamper the scheduler.
Using hard and easy could help FSRS schedule your cards more accurately. For example, If I were to stop using easy and used good where I used easy in the past from now on, FSRS will give a shorter interval for those cards. For these cards, my real retention will be higher than my desired retention. Similarly, If I use good where I used hard in the past, FSRS will give a longer interval. My retention of these cards will be lower than my desired retention.
[quote=“L.M.Sherlock, post:7, topic:34147, full:true”]
Using hard and easy could help FSRS schedule your cards more accurately. [/quote]
Yes, that is the trade-off but on condition that user do self-assessment correctly, which is not always the case.
Yes, using ‘good’ instead of ‘easy’ means more work, at least in theory. The good news that real retention can’t get over 100%. Just joking.
That is important. That is why I use Again instead of Hard.
It ultimately comes down to the fact whether users who want to use two key Anki understand how Anki works and how correct and honest they are in their self-assessment.
So, using two key Anki may affect the amount of work but it should not affect the scheduler, because scheduler is a math formula after all.
My whole deal is that when you are a new user running the defaults, four-key is less intuitive, more likely to be used in an inconsistent manner and not worth the marginal accuracy improvements and workload reduction (at least from my uneducated intermediate user perspective). FSRS already represents a much more extreme optimization in those departments from the default scheduler and a ton of current Anki users (including those that use 2-key grading) either haven’t opted to start using FSRS or are unaware it exists/how drastic of an improvement it is. You could do a lot worse than FSRS + pass/fail grading, and a lot of people are.
When you are a new user, you are still getting used to the software and not in the position to immediately develop a preference (or even an awareness that there is a preference to be had) yourself.
As a first-time user switching from rote memorization and quiz apps I was entirely unfamiliar with the concept of self-grading at all. If I were less technically inclined I would’ve quit right there just because of how foreign and intimidating this one part of its design is. Suddenly being asked to grade your memory on a scale from 1-4 when you haven’t ever studied like this raises a lot of questions and doubts about the efficacy of the software (“is it really as robust and effective as others say it is if I have to literally tell it how well I remembered something?”) and puts pressure on you not only to grade whether your mental recollection of the back was correct, but “how correct” it was. A lot of people just don’t think like this.
Part of what makes Anki so intimidating and gives it the reputation of being unintuitive is the impression that it is necessary to become familiar with the culture of add-ons and scheduler tweaking out the gate before you have even gotten yourself to the point where you are a consistent user and seeing benefits.
When I “pitch” Anki to prospective new users, neither grading workflow (having to go through the chore of grading your memory on a 4-tier scale, or go add-on shopping/just ignore the Easy and Hard buttons, working against what is visibly the “intended” design) gives a good first impression. This kind of person generally wants to focus more on learning and less on memory grading systems.
I don’t think people who aren’t power users will really care that much if they see easy cards a little early sometimes. Those reviews by their nature take up a very small amount of review time anyway.
I like the default 4-keys, and just want to give a different opinion.
There may be someone who thinks like this: “I give it more information thus it should give me more accurate intervals, and that’s good”.
I have never used quiz or other 2-keys apps, so I feel it natural to use 4-keys. Maybe this is the reason why you feel inconvenient at the beginning.
You don’t need any add-ons or tweaking to use Anki at the beginning. The default configuration of Anki may not be so good but it works, and the advantage of Anki is that it is customizable. You can install more add-ons as you have been knowing Anki better for a longer period, or you may just have dropped Anki after a try .
I’d like to quote this here: “Just using Anki instead of the traditional cramming ways is already a huge advance for beginners and you don’t need to be an expert to use Anki.”
What I mean by “quiz apps” are apps that test your knowledge by having you input your answers or judge you in some other way instead of self-grading and are designed specifically for the material they cover.
I agree that Anki is already a huge improvement and that you don’t need any add-ons to use it from the start but that doesn’t mean first impressions aren’t important or that streamlining wouldn’t do any good.
Pass/Fail grading is way easier to start using effectively immediately after you install the software so much so that many community-maintained Anki guides for different subjects advise switching to it as one of the first steps after installing it.
If it is that popular to defy such a default maybe it is worth considering whether it is worth being the default.
I think 4-key has its place for people who prefer it. 2-key is something anyone can use effectively. And it is much less intimidating to be greeted with as a first-time user just trying to memorize Seinfeld trivia or something.
I don’t remember being intimidated by those 4 buttons when I started using the program, on the contrary, everything was pretty intuitive right from the beginning: Again, Hard, Good and Easy.
I also never felt that I needed to install a ton of plugins or learning how to tweak the settings before using the program, I just started using it with default settings just right from the beginning.
I agree that a pass/fail system would be very useful for new (not power-) users. Maybe 4 button answers are not “intimidating”, but it can be confusing: should I press Hard if I feel like I “almost” remembered it? was this really easy or not? etc.
I recommend my students using Anki, but the few of them who know of it almost always tell me it’s too difficult, and probably the self assessment is one of the difficult points, especially if set on a 4 level grade.
I try to simplify the process as much as I can, give them premade decks and prepare them in a way that will explain everything, but still I often fail to motivate them enough to use it o keep using it. That’s probably my fault, but anything that would reduce this load would be welcome, and 2 button grading would definitely help. Telling a prospective user to use only Again and Good when they see 4 buttons (that is, telling them to ignore an integral part of the interface because it will be easier for them) can be counterintuitive (why is it better to use 2 buttons if there are 4? why offer 4 in the first place? how is Again different from Hard? etc.).
I don’t know if it should be the default, but I do think at least Anki could offer the option with a toggle to use 2 (pass/fail) or 4 buttons (maybe as deck options or as a global setting, I don’t know).
Kind of. RMSE and log-loss look good for two-button users (even better than for 4 button users), but the problem is that our current definition of D isn’t very well suited for two-button users. If we were to ditch Hard and Easy, we would have to re-define D to make it more suitable for 2 buttons. Though, again, according to data FSRS performs well anyway.
It’s a weird situation, because theoretically the current D should work poorly for 2 buttons, but the data says “Nope, FSRS works just fine”.
The only explanation that I can think of is that D does, indeed, work worse when only 2 buttons are used, but this is offset by the fact that a lot of users use 4 buttons inconsistently, so on average it looks like using 2 buttons results in better RMSE/log loss. In other words, the benefit of more consistent grading offsets the downside of D being designed for 4 buttons.
From what it is said on this thread it doesn’t look like it hampers the scheduler’s performance. And in fact, before FSRS many guides precisely instruct new users to use only Again and Good, to avoid ease hell.
Also, as far as I know, mobile users don’t have anything equivalent to the spacebar (without going through the options and tweaking them, which a newbie won’t usually do), so users probably feel all of them are equally important.
In Anki Desktop an addon can be installed to remove Easy and Hard (although add-ons are not too accessible to complete beginners/not techy people), but even that can’t be done on mobile and Android developers said they wouldn’t implement 2-buttons unless Anki Desktop does it as well, since they want to be more in line with the main app.
So in my view there is no downside to implementing this. Whether it’s a default or not, and how to implement said option (normal option, first use prompt, etc.) is another discussion, of course.
Assuming this is the referenced issue: Add a "Pass/Fail Mode" · Issue #15700 · ankidroid/Anki-Android · GitHub @David Is there any particular reason for this? Or is it just a matter of being in sync with Anki on desktop? I have been running without the Easy&Hard buttons (essentially Pass/Fail) for a while now on both desktop and mobile. Personally haven’t noticed anything weird… I could create a PR for this in the Ankidroid repo if you want, it should just be a matter of adding the appropriate preference setting.
That is not correct. In AnkiMobile, users can tap on the right side of the card to reveal the answer, or answer Good. Users don’t need to reach for buttons if they only use Again/Good.
I read every post made here, and such an option may come at one point.
Sorry if this has already been suggested, I tried to search before posting at least.
.