19 hours ago
- Spaced repetition works best for systematically organized key-value mappings with short keys and values, such as the NATO phonetic alphabet or periodic table.
- For biology and language learning, spaced repetition is effective because the underlying conceptual models (e.g., 3D shapes in biology, a language center in the brain) come naturally, leaving mostly facts to memorize.
- Highly conceptual knowledge like math is difficult to encode into flashcards because it requires building a mental model first, and creating short, unambiguous questions from abstract facts is challenging.
- Relational facts (e.g., 'caffeine is metabolized by CYP1A2') are easier to encode as binary predicates, while stand-alone assertions (e.g., 'all unitary matrices are invertible') are problematic because simple yes/no questions are biased and broader questions are ambiguous.
- Using AI to write flashcards is often misguided because it cannot access an individual's internal mental model, what they already know, or which facts need more reinforcement.
- The author suggests that case studies showing how to turn textbook passages into flashcards could help derive general rules for encoding complex conceptual knowledge.