I have been using Anki, the spaced repetition tool, off and on for two years now. For context, here are my usage graphs:

As you can see, I’m much more inactive than active. Aside from summers and a new year’s resolution in 2024, I haven’t been consistent in my memory practice. I started using Anki because many people I follow on the Internet used it. I had several reference sheets lying around from finals, and decided to add information from those as cards to start.

I want to reevaludate how I use Anki as I start back up. My goals have changed, and I want to use my previous experiences to inform how to effectively use spaced repetition.

Successes

I have had several moments where investing the time in Anki have visibly paid off, usually in classes where retaining material was important. There have also been less visible successes, but here’s a list of things I know mainly because of Anki.

• I memorized density and mass functions of probability distributions, and expected values, variances, and connections between them. This helped when recalling distribtuions in advanced statistics classes.
• I memorized two poems, Jabberwocky and Ozymandias
• The first summer, I understood terminology from a theory of computation course after adding it in the first week.
• Learned preprocessor directives in C, some Makefile syntax, and random trivia about <stdio.h>.
• Thanks to a few random prompts from one of the reference sheets, I knew about the six types of files (regular files (-), directories, symbolic links, named pipes, character devices, and block special devices) in Unix before encountering them in a systems programming course last spring.

Failures

However, not all cards are created equal. Some notable failures that come to mind from the first two years include:

• I created cards to memorize facts about LEDs and LCDs when breadboarding, but all of those cards are leeches now.
• The vocabulary questions from the theory of computation class no longer fit outside of the context of the class, and the cards don’t stand up alone.
• I memorized the poems above, but didn’t spend time sitting with them the way Michael Nielsed did when memorizing various text passages, which has led to brittle memory, no understanding, and difficulty memorizing trickier passages.
• I memorized other probability distributions and facts without understanding them, breaking the cardinal rule of spaced repetition: understand before you learn.

Analyzing a bad prompt

As I start to weed out my deck, I want to be mindful and intentional about how I use Anki. Each card I suspend or flag is an opportunity to reflect on why I created that card, what I don’t like about it now, and how to avoid creating cards like it in the future.

Often, my cards have poor prompts. I have observerd several mistakes I made when first writing prompts: the questions assume but don’t provide context, the answers are too detailed, or too vague (what is a correct answer?), there’s only one card for a topic that should have several. Andy Matuschak has an exstensive guide on how to write better memory prompts. He lists five properties of effective retrieval prompts: focused, precise, consistent, tractable, and effortful.

Here’s an example of a bad card I want to assess:

Evaluating based on the qualities above, what’s wrong with this card?

• This answer is not focused. The response is over twenty words, more than double the length of the question, and I am expecting myself to recall it verbatim. Additionally, it relies on details outside the scope of what I want to recall.
• The prompt is not precise. Instead of providing context for the passage, the card implies context about how I broke up the sections when first writing cards. It includes nothing specific about the passage itself (doing deeds, the great devotions, etc.). Rather, it asks for the “third” part of the passage, an arbitrary distinction I made before memorizing, and relies on me remembering enough of the passage to know where the “best” keyword is.
• As a result of the lack of precision, the prompt is not tractable. When I came back to this card after ten months, I couldn’t remember if the “third part” referred to the end of the poem or this section. My brain lit up in reference to the poem, but not the specific lines I wanted to recall here.
• The card is not very tractable, as adding it felt like relearning this specific passage.

If the card is so bad, why does it exist at all? Prior to The Man in the Arena, I had previous success memorizing Ozymandias by breaking the poem up according to its narrative structure: setting the desert scene, the sculptor’s description, and the rise and fall of the king. I tried to apply a similar tactic to this passage, and the semicolons make for great cutoffs.

However, I failed to take into account how different this passage is. There is a narrative structure, but it is less aparent, and I didn’t take the time to understand it before creating structure-based cards. I also failed to appreciate how complex the passage was, and didn’t make enough cards in general for it.

Another general thing I struggle with is evolving my deck. Once I’ve created a card, it can feel like it’s etched in stone. Rewriting or discarding a card feels like cheating: instead of recognizing a tool that’s not working, I’m cheating on a personal test, changing the answers to fit my style. I fail to realize what my purpose is when using Anki. I am trying to build knowledge, not to evaluate it. If I am consistently getting a card incorrect, it is far better to modify or discard it than to memorize brittle knowledge, or get discouraged from getting many questions wrong and become inconsistent.

Lessons to apply going forward

Now that I’ve looked at this card, how will I avoid creating similar cards going forward?

1. Spend some time understanding this passage. I’ll sit down with a pencil, noting word choices that are interesting, surface details, orderings of clauses, and learning about the context of the speech.
2. Create (more, better) cards. Now that I have a framework to evaluate previous prompts around, I’ll create atomic, specific notes (and more of them).
3. Delete previous cards. I won’t be afraid to change what isn’t working, suspend and reevaluate cards, and modify the ones I create now.
4. Get more creative in what kinds of cards I use. Instead of solely recall, I can add images, cards to prompt me to think, opinion-based cards that I can reflect on and modify. I’ve stuck to the prompt-then-regurgitate mold for too long now.

One of my goals this summer is to memorize The Rime of The Ancient Mariner. It is considerably longer than anything I’ve tried before, and will benefit from my experience with smaller passages.

I also want to add cards to help me learn more quickly at work. This will take more time and thought, but evaluating what I’ve done so far is a first step in that direction.