Anki with minimal failing of cards

Although I too often find Anki tedious in many respects, I persevere nonetheless because of the good results I attain from it, as I outlined in the other thread: I am a teenage Anki whore. But I am always thinking of alternate ways to use it that might not be quite as tedious.

There are a relative handful of cards, like maybe 2 dozen out of 10K+ in my German deck, that I rarely/never fail despite repeatedly not recalling the definitions. These are words from CEFR type coarse vocabulary lists which I regard as junk vocabulary, but just am too stubborn to delete them. I have noticed though that if I never fail such cards, but instead use the minimal "known/good" interval, I still manage to learn most of them just by the repeated exposure, even though it is less exposure than if I failed them.

Learning new cards and re-reviews of failed cards add significantly to overall number of reviews and daily review time. But the experience above made me think of possibly using Anki on some decks without ever failing a card, unless the minimum review time got to around a month and I still did not know it. I think if I still did not fail one then and let the minimum time get to increasingly longer intervals, then I would never learn them, and the effect would be the same as suspending leeches (which I don't do).

If in addition to the above, one simply transferred very easy cards to an archive/easy type of deck (I would never delete them in case I had to stop studying for a prolonged period and needed to start from scratch), then that would further reduce daily reviews and review time. When a deck gets large enough like my German one is, then even cards that only will come up in months still occur a lot each day simply due to the overall size of the deck.

To recap:
-treat most cards as normal with next interval decided on basis of how hard it is to recall
-but never fail cards when not knowing a definition, even in the learning stage, just use the shortest next interval
-transfer very easy cards to another deck for possible future use (like "German_EZ")

Has anyone ever done something like this with Anki, and if so what was your experience?

I use anki in all sorts of different ways. For Tagalog, I have a "conversational" deck. I put questions in Tagalog on one side (stuff like "Where do you work?", "How many siblings do you have?", "Describe your dog.") and a few useful sentences in Tagalog on the other side ("I work at a school." "I'm a teacher." "I like my job."). I have that deck set up with only 3 new cards per day (with a current backlog of 20). When I read the question, I say as much as I can out loud, then flip the card and read those sentences out loud, comparing them to what I said. As my spoken skills improve, I add new sentences or details to the back of the card.

I never "fail" any of these cards, nor do I mark them as "easy" (even if they are). I've been automatically marking them as hard, so they'll keep coming back at the lowest interval. I'm really pleased with the experience so far! It's made my Skype conversations much more fluid.

I don't expect this to be a "forever" deck like my vocabulary deck. Once I'm feeling comfortable enough with basic conversations, I'll delete it. But for now, it's very useful.

Of course, this is absolutely nothing like what you're saying - but I am using the shortest next intervals on anki, so it's kind of relevant.

Also, I have a quizlet account that I use for vocabulary sets that are really difficult for me. In Tagalog, for example, most adjectives start with "ma", making them very hard to memorize at first. I got tired of constantly failing them on anki. Since every single day I found myself with a dozen or more "ma" adjectives, anki obviously wasn't working.

I made a quizlet deck using only those words (currently up to 56 adjectives, all of them starting with "ma"), and play the scatter game for 5 minutes each day.

(my quizlet deck)

In anki, I mark those same words at the shortest next interval until I feel confident with them. The dual exposure really helps, and I don't have the frustration of constantly failing the same word. It's working! My lowest interval on many of the words is up to 10 days, and I'm passing most of them. If I fail a card and the interval is under 10 days, then I just mark it as "easy". But if I fail and the lowest interval is at 10 or more days, then I fail the card and put it back into the lowest interval rotation.

Stelle,

That's actually fairly similar, i.e. the main point is that you are not failing cards. Naturally cards will eventually reach a month and then even longer, so you won't be getting much effective exposure unless you do fail them then.

With those Tagalog adjectives, which to me are similar to German prefixed verbs, have you tried to emphasize the non-prefix part on your cards? Like:
maliit: maLIIT or maliit
malaki: maLAKI or malaki
matapang: maTAPANG or matapang

That would be like treating "ma" as an article that is learned with the main lexical item.

In the past when I was trying Anki, I always felt that if I remembered something after 2 weeks then I shouldn't be using time to study it anymore. I couldn't find an automated way to make a card like that to drop out though. Also, if the first time I answer a card I know it outright, then I should also be done with that one.

Could one use Anki for just "exposure"? No feedback from the user, the cards just come back according to a predetermined curve?

As a side note, this is something that always kind of bugged me: If Anki always asks you questions just as you are about to forget them, then how come I forget things? shouldn't the curve be back one day so that I get them all right?

Nope, because the problem with the algorithm is that it doesn't have any notion of the inherent difficulty of an item. A word may be easy because it's got a close cognate in your own language (interesting, interesante) or it might be difficult because it's made up of complex consonant clusters that don't exist in your native language (Russian, j'accuse!). If they adjusted the algorithm to stop you forgetting the difficult stuff, you'd get reeeeaaally bored with the easy stuff repeating too much.

As far as I know there is no automated way to move a card. You have to open the browser, select the cards, and then transfer same to a different deck. If one uses smaller sub-decks as I do, which means the cards you have just reviewed are easy to find, then that should not be too time consuming as you can use the control key to select multiple cards to move at one go. Perhaps Cainntear can wow us with some of his new found python skillz and write an addon to add a button to move a card to a predetermined deck .

If you always opted to use the same next interval (hardness button), both in the initial learning and subsequent reviews, then that should make that be the result. As an alternative, select a deck (again sub-decks or at least tags need to be used unless you want to keep track of where you left off in the list) in the browser, and then click the "preview" button. Then you can step through the selected cards, though of course you would have to keep track and manually apply the curve. You can also just use the cram feature described in the Anki manual to effect the same result on an entire deck.

What Cainntear said. Essentially though, by user chosen levels of hardness that apply to individual cards, you are fiddling with the algorithm.

I started paying attention to that when I saw the same roots coming back in verbs and adjectives. I still had a bit of a mental block, though! It's getting easier.

...and there's the other main problem with Anki. SRS was designed for discrete facts, but words are linked, and the same roots recur frequently. SRS will never know that.

I gave up on using Anki for individual words and phrases without context, because I found it incredibly painful. Today, almost all my Anki cards come from actual books. I use two main formats:

1. Understand a word in context. Boldface one or two words in a sentence, and put the definitions on the back of the card. Pass the card if I understand the word.
2. Fill-in-the-blank. A.k.a MCDs or cloze cards. Toss a bunch of interesting text on the card, and hide a word, or even just a syllable. Make multiple cards from the same sentence if there are multiple interesting things going on. Pass the card if I can fill in the blank.

These cards tend to be very easy—I hit the Easy button close to 50% of the time, and I fail very few cards. In fact, if I fail a card a couple of times, I just throw it out. But despite the general ease of these cards, they've improved my vocabulary considerably. For me, the key insight is that Anki is not some kind of permanent cybernetic memory where I remember everything for every. Instead, I treat it like a "memory amplifier": it makes my memory several times stronger than normal, but if I still forget something, who cares?

The premise of this thread was to use minimal failing, and the only reason I suggested failing after the interval of a month for a still unlearned word was because it would then start to get so far in the future that such longer times between repetitions would have no learning value. So it that still objectionable?

As I said in the other thread I don't delete, and know that it is obsessive. But in any frequency band some words are just much harder to learn than others, at least to me. While it may not make much difference if one deletes past a certain frequency band, surely it must at the lower bands, and especially the 0-3000 range which is of paramount importance for a learner. How will one learn hard words in that range if one deletes them? With an amount of extensive reading that takes orders of magnitude more time than Anki?

Re context, I personally always put example sentences on cards if I can find them. I find the ones on wordreference.com and dict.cc to be mostly worthwhile and represent the language as spoken, as opposed to very contrived sentences in many grammars.

For myself, there is no way I could have learned or partially learned over 10K words in German in one year without Anki (or Iversen type word lists), at the expense of an hour a day reviewing and probably a similar amount of time making cards for the future, so 2 hours a day. Only if I could dedicate a full-time effort each and every day could I have accomplished that without Anki, and I am doubtful even then, due to shortcomings of extensive reading (the topic of another thread here).

My purpose in making this thread was to see if some learners used Anki with minimal failing as a way to partially alleviate criticisms of the effects of failing cards.

@emk: re your comment on Anki being a cybernetic memory, it is just that for me. It helps me keep my knowledge active and fresh as possible with a minimum amount of time. And just as important, it represents a way back, with all the creation of cards work already done, in case I were forced to set aside language learning for an extended period. If I did not put so much effort in creating useful cards (in depth definitions and example sentences/phrases), I would think differently though.

I am always open to more efficient methods of acquiring the necessary vocabulary (15-20K) for a language in what is to me a reasonable time frame (1-2 years), with less time and effort and stress. I just have not found one. So for me, I would rather focus on trying to come up with ways to use Anki with less time, effort and stress.

Obviously Anki is not for everyone, and other persons' judgments of what is reasonable and efficient differ from my own.

Pere, what are your opinions on the Goldlist system? I gave up Anki when I tried it out. I think of it as SRS with a different approach.

Let me answer this question in a separate thread.