Prior Knowledge and Transfer — Map of Content

Nobody arrives empty. Every user brings a complete, automatic model from whatever they used before, and that model runs first, by default, beneath conscious control. It is why some parts of your product are free for one cohort, why the near-misses are your most expensive errors, and why confusion clusters exactly where two things are almost the same.

This skill is a graph. Scan the lines below and follow only the [[wikilinks]] you need — don't read all eleven nodes up front.

Boundary. "It's too hard for this segment" lands here when the cause is the segment's prior system, and with duo-difficulty-calibration when the cause is their ability level. The distinguishing symptom is confident wrong answers rather than blank ones.

Start here — the shape of the problem

• [[references/difficulty-is-distance-from-what-they-already-know]] — difficulty is a relation between your product and each user's existing model, uneven across axes; stop shipping one rating.
• [[references/the-old-model-is-suppressed-not-erased]] — a redesign never deletes a habit, it inhibits it; budget the switch cost and expect regression under load.
• [[references/interference-clusters-among-near-neighbours]] — confusion concentrates between things that are almost the same, and false friends produce confident errors with no detection signal.

Diagnosing the errors you are already seeing

• [[references/mismatched-mappings-make-errors-structural]] — when your model splits what theirs merged, the errors are in the mapping; cluster reports by source system and say the mistakes are expected.
• [[references/write-the-path-for-their-source-system]] — pace onboarding by transfer profile rather than by level, and clone paths across structurally similar sources.