That seems like a pretty niche edge case. I don’t think that most users misuse Hard because of this.
Honestly, I think we should just give up. Dozens of people in this topic have proposed dozens of ideas and Dae has rejected every single one of them. Well, less of a “rejected” and more of a “didn’t even pay attention to them and didn’t respond”.
P.S. Dae, if you are reading this, please don’t use the “I’m busy with other matters” excuse. This topic exists since September, it’s been 3+ months.
Honestly, I would respect you more if you just said “F##k users who misuse Hard, they can go f##k themselves”. That would be kinda based, actually.
It could still nontheless be helpful to change the numerical value to a string, just to make it more obvious even for people that don’t know the specifics (I mean it is the same as the shortcut defined in settings, but still).
In my opinion we need to recruit and a bit support volunteer developers for Anki for desktop. AnkiDroid has a donation system but not on desktop so in my view there is almost no rational reason to develop for desktop.
e.g. Luc Mcgrady(A_Blokee) is developing graphs enthusiastically on another thread for now, he is developing an add-on for statistical graphs so is a rare volunteer with good technical skills.
His Github has a buy me a coffee URL so it should be possible the community can help him a bit by creating a reddit post like LM Sherlock. Even if there are no donations at all the increased visibility of the contribution is beneficial to the developers and motivates development.
In my opinion this Hard misuse problem seems to me to be fatal for FSRS and learning apps.
This misuse of the Hard occurs in about 10% of users of FSRS.
Anki has about 3 million active users, doubling in a few years (according to AnkiDroid data).
If FSRS is set as the default algorithm, the problem of misuse of Hard may occur to about hundreds of thousands of new users in a few years.
It is almost already decided that FSRS will be the default algorithm in the near future.
If this misuse problem occurs in the FSRS algorithm, the intervals will become too large and forgetting will occur, thus the learner will fail their exam.
Default SM2 algorithm also has many problems but the problem with SM2 is that the intervals become too short, there is an increase in workload but no forgetting occurs, thus the learner can memorize the cards by the time of their exam.
So if a problem occurs SM2 can almost certainly memorize the cards, FSRS cannot be used for learning if a problem occurs.
It is more beneficial to the average user to have an ordinary car that anyone can drive than a sports car that 10% of users cannot drive. The learner’s goal is to arrive at the destination, and driving at high speed is a secondary goal.
Expertium suggests changing the UI but in my opinion it is not enough. I think it would be best to change the default UI to two buttons, or force change to SM2 if misuse is detected, and make it almost 100% sure that it will not occur.
Edit: So far there seems to be no way to detect misuse.
That’s the past me and I know at least one AnkiDroid maintainer who can also relate to this.
Not everyone reads the docs and that’s fine. We are all busy doing too many things so we need to prioritise. Life is too short to just RTFM. In fact, there is a reaserch paper with the same name that shows a very different picture: people who don’t read docs are usually the more educated and smarter people. Surely, we want them to be here.
I wouldn’t have cared if this was a small issue but it’s not as we’re soon moving away from SM-2. So, we’d need to make sure things are right even if you don’t have the time to read Anki’s 200 page documentation (if you print it out on letter-sized papers).
By the way, dae is considering having a 2 button mode but we’d need to have a shared reviewer first. Though, from his recent comments he might be willing to accept a 2 button mode soon. Let’s see.
According to the abstract, they are more educated. That does not mean they are also smarter.
I do not have hs access anymore and this claim does seem off to me. How did they explain that more educated people don’t tend to read the docs? What programs were tested specifically? E.g. if someone is more educated in a certain field (e.g. accounting), this person is a) more educated and b) probably has enough experience with accounting software, so he doesn’t have to read the manual.
So if someone is educated about learning, spaced repetition and the likes (like e.g. psychologists), they probably don’t need to read ankis manual regarding these topics. As another example: we used JASP in hs to do statistics. But if we have to use something else like SPSS instead, we can deduce most things since they are very similar.
Besides: Not reading the manual doesn’t mean not getting information from somewhere. Maybe more educated people tend to ask other humans. Or they tend to search on the internet (like forums) instead of having to read lengthy manuals.
And this claim from the paper doesn’t apply to a lot of professions either, where high education is common. E.g. in IT you basically have to read documentations all the time. In clinical psychology you have to read manuals regarding tests (e.g. to test for depression), not only for carrying out the test but also to evaluate and standardize the test results.
I don’t think we should be going in length about this here but some clarification:
This is about product documentation so not relating to manuals about tests in clinical psychology.
They tested products for domestic/personal use and not things used in professional settings.
It’s actually somewhat popular among technical writers so don’t expect obvious flaws. Also won the Ig Nobel.
The understanding among technical writers is most people like to use the products and learn by using instead of reading anything. I’ve no idea why educated people would read less though. I think I read about it but I can’t remember.
There is a way to detect misuse: the intervals get too large and people repeatedly fail to remember their cards: Keep hitting ‘Hard’ and the card keeps getting put off for the future. Hit ‘Again’ and the card reappears relatively soon.
Also there are the literal meanings of the words “hard” and “again” to clue people in (I can’t speak to the translations).
Yes, if the users detect it. There does not yet seem to be a way for auto detection by the program.
However I think these are the concerns:
If I remember correctly there was no info in the manual until recently that “Hard instead of Again is a misuse”. In SM2 using Hard instead of Again is rarely a problem, so it is possible to use Hard instead of Again at the user’s own risk. So even if users read the manual long-term Anki users may not be aware of this change. If they have already read and understood the old manual, there is very little reason to re-read the manual.
I think there are a several cases when it is reasonable to intentionally press Hard instead:
a. The user could not recall a card but that card is not important, so they postpone with Hard instead of suspend.
b. The user does not have time to relearn the cards because the study load is too high, so they press Hard to force the reduction of the review cards.
c. The Hard due date is well before the user’s exam, so the user postpones learning with Hard.
So in my opinion it is almost impossible for users to find out on their own and even if they did know they may intentionally press Hard.
This info is posted on Anki’s subreddit but about 95% of Anki users do not use Anki’s subreddit. I think that to almost certainly notify all of the Anki users, ideally the warning should be displayed when FSRS is enabled.
Since users that misuse hard tend to not link their issues regarding retention with the misuse of the hard button, the only way would be to automatically detect misuse by using something like an algorithm.
But how would one do that? If a user grades themselfes with the buttons, the FSRS algorithm has to assume that the user did it accurately enough. The manual states some percentages of how often one should press the buttons:
But
No idea what those percentages are based on (real data or estimates?)
It depends on what you’re learning (e.g. if I had to study historical events with their names and dates, I’d get them wrong most of the times. It just refuses to stick. So my again and hard button usage count would skyrocket)
The above means that one cannot simply monitor the button usage, because it will be inacurrate and error prone.
In short: we’d need some way to automatically detect misuse, which cannot be reliably done (just yet?).
I have already stated elsewhere that I think a 2 button mode is probably the best course forward, and I know you’ve already seen that, so your hyperbole is not helpful. It’s been a particularly busy few months, and I’m still playing catchup. Please be a bit more patient, and direct your enthusiasm into more productive avenues such as the great user education and guides you do.
