Computer generated transcript

Warning!
The following transcript was generated automatically from the content and has not been checked or corrected manually.

Meta- learning LEARNING HOW TO LEARN – ANKI MARK.JACUNSKI@NHS.SCOT0-9 min 9-23 min 23-40 minBasics of useAnki is automated optimal flashcard studying Free software (Apple app is paid, unfortunately, but can use web browser version) https://apps.ankiweb.net/ Personal nugget: You make flashcards (or get them from others, download them) I mostly made/make my own cards, https://ankiweb.net/shared/decks/ but certain tedious topics to make flashcards for I used others’ decks, https://www.reddit.com/r/medicalschoolanki/ e.g. anatomy Cards reintroduced to you based on when you need it to rememberAnki gives you a dashboard of your performance Personal nugget: The whole point of this is not to be burdened by over-analysing your study style. I only have checked my stats: -early on, when figuring Anki out -whenever my study style changed, such as after med school -at most 6 monthly to make sure my leeches aren’t taking too much of my timeStudying and card outcomes Study something too quickly: time is wasted AND learning is worse Study something too slowly or rarely: learning is worse Cram: learning is poorer and short-term Several possible outcomes for a given card: A simple fact that you remember well will rapidly be relegated to your memory (ie the app will only reintroduce it after a few years, just to make sure you know it) A collection of facts that you want to remember or a detailed list may only gradually become less commonly seen, based on how high your standards are for rememberingCards are learned then graduated There are two main states of a card: Learning – new cards or cards marked as forgotten, undergo a slightly different repetition rate to allow for learning Graduated – once cards pass the card repetitions set for Learning cards, they join the remainder of your deckIf you haven’t set up Anki yet, hang tightRating a card By default, there are four ways you can rate a card: Again, Hard, Good, Easy The meaning of these ratings is determined by the algorithm and how you use them, but generally: Again means inadequately remembered/understood Hard means adequately remembered but possible misunderstanding/detailed omitted Good means adequately remembered Easy means already deemed remembered/in knowledgeLearning cards behave differently For Learning cards, the ratings apply only to the learning process -Again will reset cards to be re-learned -Hard will slow down the learning process Personal nugget: As being unsure is part of the learning -Good will continue the set learning process process, I tend not to use Again unless I continue to feel unconfident when the card is -Easy will skip the learning process and make the going to be graduated (ie. next interval is 1 initial interval slightly longer day) I use Easy at the beginning of the process if it is clear I remember the contents of the cardAgain and Hard marking graduated cards For Graduated cards, the ratings will apply indefinitely, only becoming irrelevant if the card is manually reset Personal nugget: The default settings are suboptimal, which will be discussed in the I use Again on graduated cards Algorithm section whenever I forget a simple fact or a fact I really want to know Generally: Again restarts the learning process, shortening the interval and For the usual simple concepts, I increasing interval slower (e.g. 28 days -> 0 days or 7 days, depending avoid Hard, but I do use it for certain on the settings… to be discussed) lists and groups of facts (discussed in Advanced Use section) Hard still increases the interval by default, but not by as much, and decreases further intervals by 15% (e.g. 28 days + 1 day modified by 1.3 = 38 days, assuming initial interval was modified 2.5x, the next “Good” ranking will be 38 + 1 day modified by 2.4 = 94 days)Good and Easy marking graduated cards Good increases interval by at least one day, and the standard multiplier which starts at 2.5x (e.g. for a 28-day card marked Good twice in a row, the next interval is (28+1)*2.5 = 73 then Personal nugget: The key is to keep it simple, 185 objective, and reliable. Accordingly, I ◦ As you see, once you remember something for a month, it is expected never use Easy. that you remember it for 2 months, then for 6 months. The amount of additional study that avoiding the easy button causes is Easy increases interval by an initial bonus and increases the minimal, perhaps a few cards more multiplier by 15% (e.g. for a 28-day card marked Easy, the next per year. interval is (28+1)*2.5*1.3 = 94 days, with the next multiplier being 2.65x)Vacation, holidays, and days off Everyone needs days off; holidays and vacation are worth valuing as community, family, friend, or personal time. Anki will take into account any days off. However, if you plan on having consistent days off (e.g. every Saturday), try to find an add-on that will spread the load rather than dump everything onto the next day. Alternatively, you can study ahead. Click on a completed deck -> custom study -> review ahead and choose how many days ahead to study. A new deck will be created, and these cards will be returned to the original deck for the appropriate interval once completed. Personal nugget: If I reduce my New cards 2 weeks in advance, I know that I will need only about 30 minutes daily for my main deck, which is easily doable while resting, getting ready, etc. Therefore, I don’t really take any days off, but would encourage people to find whatever works for them.AlgorithmBasic settings step-by-stepShort-term workload Personal nuggets: New cards will require frequent input High to start (20 is a good number but 50 over the following 1-2 months, therefore will increase your short-term may be ok if there is lots that you already workload. I want mine to be under 1 h remember), then can find a good balance (5-10) per day so keep the number low, 0-4 new cards Unless extremely detailed and rapid learning, should not be reaching maximum. This should be set to your absolute ceiling of attention/time I don’t expect to study 450 cards/day any more, but this would probably take commitment/caring. me about 4 h, which would still be physically possible most days for me. Therefore, I have left it at this number for the time being.Short-term memory Personal nuggets: I found that some more complex cards Recommend increasing learning steps (e.g. detailed cards for Behcets from the default of 2 (1 min and 10 min) to disease) were difficult to remember with just 3 steps, so I added another at least 3, exponential increase (approx. one (at approximately 17 h). 10x longer by step, 1 min, 10 min, 120 min) With this additional learning step, I tend to be more forgiving with the 10 Graduating interval is the first interval = 1 min and 2 hours repeats, relying on day, I recommend the default. the next day (17 h) to confirm whether Easy interval is the first interval if you I have really learned the card press Easy during learning. 4 is the default but 1-6 is reasonable. I would keep the insertion order randomErrors and forgetting Personal nuggets: When you press again, these are the steps You want to know which cards you taken, with cards acting as “Learning” can’t remember well (leeches) cards because these will take a high amount of your time, brainpower, and Analogous to the graduating interval, I recommend the default, 1. frustration. Determines how many times you get a card My recommended ways of dealing with them: wrong before the programme identifies it -reset them to start from scratch or as problematic. much more conservatively (i.e. if interval is 2 years, may be better to Tag or suspend. If you are keen on change to 6 months rather than 1 day) changing ineffective cards, then switch to -separate out complex facts & suspend, otherwise keep as tag only. concepts -use mnemonicsMaintaining your deck The most important aspect of maintaining a deck is fixing leeches (cards you get wrong repeatedly). These are time wasters. If you can’t learn it or fix it, you may as well remove it and accept that you will never know it. But you will inevitably make mistakes in your cards as well. Mark them (control+k in Windows) or Flag them. Schedule time to fix these and read through your problematic cards.Burying sibling cards I recommend burying (delaying until the next day) any sibling cards (other parts of a card, e.g. front-to-back then back-to-front) Seeing the answer in the same day will obviously not check your memory in the same way Unfortunately this can only be done for cards from the same note (i.e. from the same entry), and not for other cards that you set as related. Where the magic happens Maximal interval (how rarely the card can come) set By what proportion a card’s interval changes when “Good” pressed once graduated, to 100 years, some recommend 10 or even 3 years also known as “Ease”. Keep this at 2.5x (+/- 0.25 is probably fine) In addition to “Good” pressing, by what proportion the interval will change when “Easy” pressed. Remember that the Ease will be permanently increased by 0.15 (i.e. when you first press Easy, from 2.5x to 2.65x). Keep this at 1-1.3. By what proportion a card’s interval will change regardless of what button is pressed. This is the key number to change if you want to adjust the algorithm (more in Advanced Use). Recommended 0.8-1.0. By what proportion a card’s interval changes when “Hard” is pressed, including the Interval Modifier but not Ease. Therefore, if Interval Modifier is 0.9 and Hard interval 1.1, the interval will not change. However, Ease permanently decreases by 0.15. Generally recommend against using the hard button (more in Advanced Use). Assuming you aim to remember 90% of your cards, and are By what proportion a card’s interval changes when “Again” is pressed, excluding the currently remembering 88% of Mature cards, Interval Modifier Ease, and after a card is re-learned. Ease is decreased permanently by 0.2. Default is 0 would have to be 0.82, but this would vastly increase your which is a total reset, but I recommend reducing to the previous interval that was workload. successful (i.e. assuming an Ease of 2.5x, reducing interval by 3x should be effective). Check the Anki website for add-ons Add-ons https://ankiweb.net/shared/addons/2.1 Straight Reward can help you automatically regain Ease when a card is marked Good several times in a row. Allows for other Ease settings. Personally, the only two I use are: -Hierarchical tags (helps organise tags like decks) -Cloze Overlapper (no longer valid for up-to-date versions, but can recreate most of the effect in latest version)Advanced useLearning card options Recommended options: Steps Advantage Disadvantage -Short learning period -Same-day transition to -Early memory lapses and card 1 min, 10 min graduated card difficulty will determine long- (default) -Achieves card’s appropriate term outcome ease quickest -Less versatile -Vocabulary and definitions -Assumes exponential learning in short term -Most cards do not need this 1 min, 10 min, 120 min, 1000 min (my -Allows for early “resets”, on much repetition in first day setting) day 2 (after 1000 min = 17 -Less versatile than multi-day h) without affecting long- learning periods term Ease 15-20 min, 1440 -Longer learning process -Easy new cards are introduced min, and optional likely better recapitulates much later than usual 4320 min, 8640 actual learning -Takes longer to establish min -Less rapid repetition appropriate card EaseAlgorithm manipulation: starting ease Recommended options: Starting ease Advantage Disadvantage Other Slower learning Usually paired More versatile if Early on, “Again” with a higher 1.3 used with Easy will have little interval modifier (though less long-term (e.g. 1.9) which objective) disadvantage can then be adjusted Appropriate initial Complex or interval growth if difficult topics wilMost popular and 2.5 using interval rapidly have to be most underlying modifier as a relearned until evidence modifier of study appropriate Ease time and success determinedAlgorithm manipulation: easy bonus Recommended options: Easy bonus Comment -No short-term alteration of algorithm from its simple 1.0 mode (Again vs Good) -Excessive short-term repetition for genuinely known cards 1.3 -Default, appropriate for rare use but generally avoidedAlgorithm manipulation: hard interval Recommended options: Hard interval Comment -Acts to maintain previous interval and slows further interval growth 0.4-0.6 -Will rapidly shorten intervals without the benefit of relearning -Progresses a card at same proportion as before but 1.0 slows further interval growth -If interval modifier is <1, will progressively shorten intervals without the benefit of relearning -Appropriate to maintain interval with interval modifiers 0.8-1.0 1.1-1.3 -Most widely used setting -Hard should seldom be used in any caseAlgorithm manipulation: relearning interval Recommended options: New/relearned interval Comment -Resets interval to 0 after relearning, as if it were a new card 0 (default) minor lapses in memory (e.g. bad day or minor detail or missed) -Will return interval to approximately the previous appropriate interval (e.g. assuming 2.5x increases, the 0.2-0.4 setting should be 0.4x) -The lower the value, the more stringent the learning expectations are -Can act more like a “Hard” button but with a relearning process 0.4-0.6 -Leads to excessive short-term repetition for the degree of inappropriately long long-term intervals Algorithm manipulation: interval modifier according to score Change this whenever your study habits or time dedication changes Consider reviewing on a yearly or two-yearly basis Check stats on mature cards Interval modifier adjustment = log(desired %retention)/log(current %retention) The closer to 100%, the more difficult to achieve 0.85-0.95 generally considered acceptable Example 1: at 0.85 but wanting to achieve 0.9 Example 2: at 0.9 but wanting to achieve 0.87 Log(0.9)/Log(0.85) = 0.65 Log(0.87)/Log(0.9) = 1.32 This would mean about 35% more studying! This would mean about 32% less studying! Algorithm manipulation: interval modifier according to score Interval modifier adjustment = Example 3: at 0.88 but wanting to achieve 0.9 log(desired %retention)/log(current %retention) Log(0.9)/Log(0.88) = 0.824 This would mean about 18% more studying, but what if your interval modifier was already around 0.83? Option 1) Multiple together = 0.68 Option 2) Tweak interval modifier down by a few centiles Example 4: at interval modifier 0.8, but spending too much time studying Readjust equation based on lower expectations OR tweak interval up by a few centiles Example 5: at interval modifier 1.4, but finding that marking too many Again to feel comfortable Readjust equation based on higher expectations OR tweak interval down by a few centiles Algorithm manipulation: interval modifier Recommended options: Interval modifier adjustment = Interval modifier Comment log(desired %retention)/log(current %retention) -Very stringent -Possibly appropriate for high-stakes low-volume 0.65-0.8 -For thousands of cards likely to be extremely high burden -Measured approach allowing modest changes 0.9-1.1 -Sufficient as long as content of cards is reviewed for appropriateness (in leeches, for example) -Low stringency, increased challenge -Possibly appropriate for low-stakes high-volume 1.2-1.4 memory -For thousands of cards likely to be low burden but low valueFeedback