3 Lessons about Anki from MS1
Published:
Contents
- 1. Make better Anki cards
- 2. Actually practice spaced repetition.
- 3. Quality-of-life improvements in Anki.
1. Make better Anki cards
For the first days of MS1, I relied primarily on cards made by upper-year students. I didn’t like that they often used large or multi-item cloze deletions, so I starting making my own using smaller cloze deletions. While my cards isolated concepts better, they were often verbose regurgitations of lists, which promoted spatial or contextual recognition and hindered true conceptual learning. I didn’t resolve to change my ways until the beginning of the 2nd term, after some exposure to the AnKing USMLE decks and applying Supermemo’s 20 Rules for Formulating Knowledge. Here are some of the main things I look for in good cards now:
- Atomization. Each card should test an individual concept. To do this, cloze deletions should be small and independent.
- Contextualized, within reason. There should be enough context to identify the specific concept(s) in each card. For example, I like to write sentences like, “In [disease], [symptom] is caused by [process].”
- Add context in extra fields. This is something I started doing early on, even in my 1st semester. Since I use the AnKing card templates, I have several collapsible fields where I typically add lengthy explanations, figures, tables, etc. This helps when I return to the card after a while or am struggling with a concept.
- Use specific approaches (e.g. templates) for learning lists. I often make 2-3 cards per list. For example:
- A card with cloze deletions for each item, or groups of related items.
- A card where all items are in the same cloze deletion.
- A one-by-one card.
- An overlapping cloze card, often parameterized as
[1,1,true]
or[0,0,true]
.
- Add separate cards and associated images for mnemonics. If I have a mnemonic, I like to have cards for both the list (above) and the mnemonic itself. Sometimes, I’ll also add an image, usually generated with AI.
- Organized tags. This is something I learned from the upper-year decks, which were tagged by week and lecture. I found this to be very helpful when creating filtered decks.
2. Actually practice spaced repetition.
Within my first days of MS1, I switched my Anki scheduler from the traditional SM-2 algorithm to the FSRS algorithm, hoping to achieve a long-term retention of 80-90%. However, I was inconsistent and did not complete my scheduled reviews. By December, I had hundreds of new cards and thousands of due/learning cards. It was an absymal attempt at spaced repetition.
Learning from this, I made it a top priority in my 2nd semester to flush out all my reviews each day. Initially, I had some difficulty because I was primarily studying from a deck I had made from my own reading, rather than lecture material. However, following the midterm, I made sure to diligently focus on lecture material and study as the algorithm commanded. As such, I would review material from all previous weeks each day, rather than focusing solely on new content for one week. By the end of April, I was averaging over 1000 reviews a day, with a maximum of about 3200.
The biggest limitation of relying solely on Anki is that retention is determined by the algorithm, at least in theory. Furthermore, Anki is oblivious to our examination schedule, and I was/am not interested in applying any workarounds for this issue. As a result, my true retention (and expected exam performance) would, on average, be whatever target the algorithm was set to achieve.
In my opinion, this is in every way superior to the alternative(s). While I may score lower than expected on exams because a fraction of cards would be scheduled beyond the examination date, as long as I kept up with my reviews, I could maintain a consistent knowledge base without too much extraneous cognitive effort. I can confidently relax and allow the algorithm to dictate my desired intelligence.
3. Quality-of-life improvements in Anki.
Perhaps the biggest quality-of-life (QOL) upgrade was when I purchased an 8bitdo Zero 2 controller. This not only made it easier to review Anki while sitting in front of the computer, as I was already doing, but it also allowed me to perform my reviews during activity and exercise. This led to a major increase in my exercise frequency and enjoyment of Anki, as one can grind flashcards for only so long without developing some form of psychosis or dissociation.
I also found a number of Anki addons that provided additional QOL benefits. Here are some of my favourite:
- Audiovisual Feedback. I used the ‘kitten rewards’ mode for a while, but it’s honestly a bit distracting, so the default scheme is what I use right now. The other configurations are even more distracting, so I’ve only ever used them once. Regardless, it’s a great addition to the Anki experience, as it provides some much needed stimulation to stay focused.
- Opening the same window multiple time.
- TabOut of Cloze
- Remove Cloze Button and Hotkey
- Highlight Search Results in the Browser
Last but not least, I was (and am) very picky about my card templates, especially their colour schemes. Following the original AnKing templates’ style, I specify separate colours for bolded, italic, and underlined text. I aim for maximal yet comfortable contrast in dark mode, and also apply separate colour schemes to the different collapsible fields. Although the effect may just be placebo and a waste of time, I feel it helps with both review speed and stamina.