SRS for sequential sets

rjgoif's Avatar


06 Jan, 2014 10:49 PM

I've attached a brief document proposing a way to tackle the "list" problem that is (hopefully) fully compatible with Anki/SRS scheduling. If anyone has time, give it a read and discuss below. Maybe I might even convince a programmer to give it a try.

Sorry if this is a little rambly, but I tend to get that way when I am dictating instead of typing. Also be forewarned: it isn't proofread at all because I have somewhere to be shortly.

  1. 1 Posted by mnhende2 on 06 Jan, 2014 11:31 PM

    mnhende2's Avatar

    @rjgoif: This is very nice. I'm not a programmer or a decision maker
    on this, but I commend the effort and the quality of the thinking. I'd
    love to see if any research has been done on this, it would seem that
    somewhere out there some type of research would be able to lend support
    to this idea (or the opposite). But lacking such research at hand, I
    would say that what you write seems logical and reasonable.

    I in now way speak for Damien, but I know that right now Damien he has
    his hands full with much more pressing issues of working out bugs,
    wanting to try the new QT webkit in Anki, etc. So, my gut tells met
    that this idea won't come to fruition quickly, but the idea does seem
    to have some merit, and if used, could potentially help make Anki and
    even more distinct product offering a potentially valuable tool that I
    haven't yet seen in any SRS software. I would be nice to see something
    like this potentially eventually implemented.

  2. Support Staff 2 Posted by Damien Elmes on 08 Jan, 2014 12:56 AM

    Damien Elmes's Avatar

    I'm afraid I'm not really convinced this is a problem that needs solving. "Recite all the greek alphabet" or "Give the fourth greek letter of the alphabet" only appear in tests; in reality you only need to be able to read and write individual letters. If you take other examples like poems and lines from a script then I have a hard time imagining the system would scale, as having to start from the start every time you make a mistake would lead to a lot of excess repetition on larger items, a problem that the existing solution of making cards with a little forward context does not suffer from.

  3. 3 Posted by rjgoif on 08 Jan, 2014 04:51 PM

    rjgoif's Avatar

    The examples I gave are illustrative and generally accessible, not meant to be "real-world". In my own experience and when talking to my peers in medicine, this is a functionality that would be useful.

    As far as "Give the fourth greek letter of the alphabet" only appearing in tests: that's my point. Memorizing a sequential list by matching items to their individual positions is not practical for real-world applications. The "minimum information" guideline suggests that this is the best way to go, as it recommends splitting a list into discrete units and memorizing them in relative isolation. I do not think it is the most effective or the most useful in rapid recall.

    "Recite all the greek alphabet" may not be a real-world example, but it one abstractly.

    A more realistic scenario: a patient's heart stops in the hospital. The code team responds rapidly and follows a sequential, evidence-based algorithm to attempt to prevent cardiac death (see

    Less dramatically, a patient presents with nephrotic syndrome. Knowing a list of most likely → least likely causes helps the physician to judiciously order tests and prevent needless medical costs and polypharmacy treatments. It takes more mental effort, and thus time, to recall the #1 cause, then the #2 cause, then the #3 cause... (and the diagnostic tests + treatments for each) compared to just remembering causes 1–7 in decscending order. My suggested method helps to memorize that list by building it stepwise while reinforcing the relationships between the items.

  4. 4 Posted by rjgoif on 10 Jan, 2014 05:29 PM

    rjgoif's Avatar

    I thought about this off and on for the last few days and came up with an easier, more broadly applicable solution: on the card template page, how about a "tier" field that takes a numerical argument specifying the unburying order? Example: I make 8 cards from one note. The first 4 cards are tier 1, so they are unburied from the get go. Once and only once the tier 1 cards are learned, the tier 2 cards are unburied. So long as all cards with a lower tier number are learned, a given tier is unburied. If a card from tier 3 lapses, all tier 4 cards are buried until all tier 3 cards are relearned.

    No fancy algorithms or interval modification. Just a simple if/then statement with a numeric input. Thoughts?

  5. Support Staff 5 Posted by Damien Elmes on 12 Jan, 2014 03:34 AM

    Damien Elmes's Avatar

    If it's important that you be able to remember all the material at once, I'm afraid I don't see what your method is offering over a single card with all the material on it, as any mistake on such card will require you to fail it and start again, just as your suggested approach would. A single card also means you don't end up having to study a bunch of cards unnecessarily if you remember most of the content already.

  6. Soren Bjornstad closed this discussion on 07 Feb, 2014 04:08 PM.

Comments are currently closed for this discussion. You can start a new one.

Keyboard shortcuts


? Show this help
ESC Blurs the current field

Comment Form

r Focus the comment reply box
^ + ↩ Submit the comment

You can use Command ⌘ instead of Control ^ on Mac