Every learning model is really a race between reinforcement events strengthening memory and natural decay eroding it.
retention(t+1) = retention(t)*(1-decay) + reinforcement
- Concept nodes — discrete concepts or skills the learner is acquiring.
- Network connectivity — how related the concepts are — related concepts reinforce each other.
- Reinforcement rate — how much a practice/study event strengthens a concept's retention.
- Forgetting threshold — retention level below which a concept is treated as forgotten and must be relearned.
Spaced-repetition apps like Anki are built directly on this reinforcement-vs-decay race, timing reviews just before predicted forgetting.