Here are the best practices I’ve collected from others and from my own experience over a decade creating and reviewing 10,000+ prompts over a wide variety of topics (languages, yoga, computer science, geography, biology, d&d, fiction, neuroscience, philosophy, physics, nutrition, math, etc) and a variety of media (books, articles, talks, podcasts, personal experience, etc).
Principles
- Train your memory because you can’t think about what you can’t remember.
- Encode useful knowledge that matters to you, in the most effective form for your brain to retrieve it when needed.
- Limit to 5-10 new cards per learning session: at first you won’t be able to separate signal from noise and almost everything is noise. You can always go back later to the same material to increase breadth or depth but with the benefit of a better understanding.
- Expect creating effective prompts to be tiring: it requires understanding, critical thinking, creativity, predicting, editing. The quality of your prompts is a reflection of the quality of your thinking.
- Don’t use other people’s prompts: they don’t have your goals , priorities, level of understanding or wiring / mental hardware.
Prompt creation checklist
- Main questions (prompt goal & type):
- What exactly are you trying to remember?
- Why remember it? How is it helpful or meaningful to you? Don’t speculate. The default is “no”. If in doubt say “no”.
- In what situations do you want your memory to retrieve it?
- Which prompt(s) (words, images, sounds) will pass through your mind when you’re trying to retrieve this knowledge?
- Which prompts will “fully load” this idea into your mind?
- What type of knowledge is this? Factual, conceptual, procedural, “open” list, “closed” list, an idea that you want to keep in mind?
- What type of prompt(s) are you trying to create? Retrieval (memory)? Creative? Salience?
- General tips:
- Write short, simple prompts to avoid pattern matching on sentence shapes.
- Use pictures: our brains are wired for images, not words.
- Make them personal: emotion and associations accelerate storage & retrieval.
- Prefer “Why” prompts: easier to remember because they give cause-effect chains and they also help you better understand ideas. Encode etymology in your vocabulary cards.
- Avoid lists & sets.
- Break down procedures into lists or keywords (verbs, adjectives, nouns).
- Avoid binary prompts, use open-ended instead (How/what/why/etc).
- Avoid “Give an example” prompts unless you’re going for a creative prompt (“Give an example you’ve not given before”) and update the card with it.
- For important ideas create multiple prompts using different lenses.
- For salience prompts (“I want to remember this idea when I’m meditating” or “I want to remember this idea throughout the day”) use a different spaced repetition schedule (e.g. use a different deck in Anki with different lapses), add as much context as possible and use your imagination to see yourself in that situation.
- If there are keywords in the question or answer, highlight them in Bold to accelerate reading.
- Before clicking ‘Save’:
- Is the prompt focused on one thing? One complex prompt with N parts is terribly ineffective and frustrating, N atomic prompts are way more effective and a pleasure to review.
- Is the prompt precise enough? Rework until there is only one valid answer.
- Is it easy enough? Use cues, aim for 10-30s of total review time over first year.
- Is it hard enough? No pain no gain.
Lenses to create multiple prompts for a deeper understanding of concepts:
- Lateral Thinking lenses
- Andy’s lenses:
- attributes & tendencies
- similarities & differences
- parts & wholes
- causes & effects
- significance & implications
Prompt review checklist
- If don’t care or the prompt isn’t useful anymore, add “Why this matters” to it or delete it.
- You won’t always remember to follow the above tips so be on the lookout for violations and correct (e.g. refactor if prompt is causing inconsistent answers, break down if it’s not atomic, etc.)
- Destroy prompts freely: your priorities, interests, understanding and experience with repetition change over time. Human knowledge also evolves.
Anki tips
Customize your note types. Example fields I use in my current “Basic” note, in addition to “Front” and “Back”:
- “Connections”: Free-form input for personal connections images, etc.
- “Source”: Where did I get the information within this card? URL, book, person, etc. Useful for when I think the content might be wrong or outdated.
- “Note TODOs”: For cards that need to be edited, this lists what needs to be edited. Normally I use Anki’s flag system (I have renamed every flag so instead of “Red”, “Green”, “Purple”, now I have “Split”, “Ambiguous”, “Unretrievable”, etc) to flag cards that need some editing, but in some cases I need to give my future self details about what exactly needs to be changed.
Besides adding new fields, you can also modify the card’s HTML, CSS and also embed logic with inline Javascript. I use this to show a card’s tags, show the new fields mentioned above, use visual cues (color, font size, alignment, etc) to make parsing cards more efficient, and also to add new behaviors like altering questions randomly (e.g. when a card has a question like “What are [capital|fixed] assets?”, it will randomly choose “capital” or “fixed”.)
Plugins I’ve found useful:
- Image Occlusion Enhanced
- Review Heatmap
- Purge collection from ( mess up card displays in mobile)
Further reading
Writing effective prompts
- Knowledge structuring and representation in learning based on active recall (Wozniak, 1994)
- Effective learning: Twenty rules of formulating knowledge (Wozniak, 1999)
- Augmenting Long-term Memory (Nielsen, 2018)
- Rules for Designing Precise Anki Cards (Bjornstad, 2020)
- How to write good prompts: using spaced repetition to create understanding (Matuschak, 2020)
Spaced repetition
- How to remember anything forever-ish (nicky case, 2018)
Lateral thinking
- Lateral Thinking by Edward De Bono