While my mnemosyne is well and truly out for the count, and I can't go through my scheduled cards, I thought I'd share a few of my thoughts and ideas on SRS in general and see what you guys think.
1. I seem to remember reading somewhere (on the supermemo site?) that the algorithm was designed to present fcs when you had a 90% retention rate. If so does that mean after every session of scheduled cards you should have forgotten about 10%? And you should forget each card one out of every ten times. I imagine however different users will want/expect different amounts of recall and change the way they grade the cards accordingly. I know I have changed the way I grade the cards over time. Could this result in an endless battle between algorithm and user, one trying to shift the spacing to acheive a 90% retention rate and the other grading the cards harder and harder to achieve a higher retention rate?! 2. Indeed, how do people deal with the necessity of different levels/ types of recall required for different flashcards. For example unless you planning on winning quiz shows, immediate recall is not required for most "factual" trivia type flashcards, ie the exact date of a King's reign for example. You might want this knowledge so that when you come across about some other historical event you can put in temporal context or whatever. However some other set of flashcards might deal with something more relevant to you in your immediate life and you need faster recall. Is this simply a matter of revising flashcards more often? To a certain extent I imagine this would be true: the more you revise a flashcard the better and the faster you can remember it. If this is the case, would it make sense to be able to tag different flashcards/categories with the level of retention required, ie. instant recall, slow recall etc. This would change the frequency that the flashcards would be scheduled at, whilst allowing the user to still grade all their flashcards by the same criteria, well remembered, not well remembered etc. 3. I have to say I was a little skeptical about learning anything other than factual or vocab style little knowledge units using flashcards. Reading the supermemo site persuaded me that a much larger range of knowledge could be divided up and effectively remembered using SRS. However after 6 months of using mnemosyne I'm back to being skeptical about the actual real life effective breadth of application of SRS. Although I have become much better at paring down wordy flashcards to the bare minimum it seems to me that it is all too easy to remember flashcards answers as words or information but not as meaning. That is to say in response to a certain cue, ie a simple cloze deletion phrase, I can remember the word that is associated with it, but that does not mean that I am processing the meaning of the phrase, or the word remembered and the information recalled is basically pointless. Recently I've noticed with some old flashcards, that I can read the cloze deletion phrase, think I've never seen it before but when it comes to the missing word, I'll unerringly blurt out the exact answer without any hesitation and complete confidence. This tends to be after very long gaps, and I have obviously remembered the flashcard very effectively, but it is a meaningless memory. Also I find that I'm getting very good at recalling stuff in the quiet of my room in front of my computer, when I'm in "mnemosyne" mode. However I often can't remember the same stuff when I'm out and about in the real world and feel like testing myself. I think the basic problem is that it's easy to forget that SRS is a tool, you devolve the responsibility of useful memory to the software. I try and combat this by making sure I'm not focussed on just remembering the word, but I'm remembering the concept, the meaning and everything behind it, everytime I come across a certain flashcard. There is no doubt that this actually slows the memory creation and makes a less precise memory, often I will remember the wrong words in a cloze deletion but with the same meaning, but it seems to be a more meaningful memory. Another related impression, is the feeling that some of the memories I've created using flashcards are like a factoid that I've written on a piece of paper that I can have a look at whenever I want. However the second I want to really think about the factoid, process it, combine it with other knowledge etc (the very reason I learn most of the stuff I do), it's gone, absent. I have to "drop" all other mental processes, go back to my mental "piece of paper" and read the factoid again. Only when I've repeated this many times can I "use" this "memorised" fact in the real life. This is especially true I think for facts remembered using devices such as peg words, mnemonics etc. The end result is that the "knowledge" memorised using the flashcards is very similar to just having a copy of wikipedia on your ipod. Anytime you want to think about something, you still have to go to your wikipedia, "search" for the term, read about it and process it all over again. 4. Have there been other ways of thinking about how to get the user to "grade" flashcards. There seems to be a lot of leeway for users to choose between the different grades. After working on a lot of non- standard flashcards, (ie a bunch of sentences I half translate/half remember back into a foreign language) I thought of this method. It involves a number of grading steps which makes it more complicated, but at the same time more specific criteria and so less leeway. The first grading occurs before you see the answer. It is a confidence question. How confident are you that you are right? Very, Somewhat, Not at all. This helps identify mistaken memory, where you've remembered the wrong answer. It's also interesting to see that sometimes you have no idea, and you are sure you are wrong, but actually you've remembered the card perfectly (don't know how the algorithm should deal with that situation though). Next one you've seen the card you grade again, not on how difficult it was to remember, or anything about the interval, but whether you were correct, slightly wrong, or completely wrong. I imagine by removing any user feedback on quantifying the interval would remove some of the user manipulation of the interval. Finally the timer would see how long it took you to give your answer and ascertain any final information from that. I guess this kind of grading would need a very different algorithm and some kind of theory related to mistaken memory, and how to correct it etc... To sum up, over the last 6 months it seems to me that reducing effective, useful, real life memory to flashcards and software is much more tricky than it initially seems, however many times you might read the Supermemo 20 rules of formatting knowledge! It is very much an art, and far more time needs to be spent coming up with the right flashcards than actually then learning and reviewing them. These days I seem to be treating the flashcards as memory prompts, (to think about the concepts involved) rather than the aim of the memory itself (did that make sense?). Anyway, sorry for the long post, what do people think? -- You received this message because you are subscribed to the Google Groups "mnemosyne-proj-users" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/mnemosyne-proj-users?hl=en.
