KM-301d · Module 1

The Curse of Expertise

3 min read

The curse of expertise is the most underestimated obstacle in knowledge extraction. The more competent an expert becomes, the less they can explain their competence. Skills that once required conscious attention become automatic. The senior developer does not think about null pointer checks — they just see them. The experienced account manager does not analyze the customer health signals consciously — they just feel the account is drifting. This automaticity is expertise. It is also what makes expertise almost impossible to extract through direct interview.

  1. Automaticity Skills become automatic through repetition. The expert no longer consciously executes individual steps — they execute chunks. "Run a technical discovery call" is one move for the expert. It is thirty moves for the novice. The expert cannot enumerate the thirty moves because they collapsed into one move years ago.
  2. Compression Experts skip steps they consider trivial. Not because they are being unhelpful, but because the steps are so fundamental that omitting them seems obvious. The novice tries to follow the expert's instructions and fails at a step the expert did not know to mention. This is not the expert's fault. It is an artifact of expertise compression.
  3. Context Sensitivity Expert judgment is deeply contextual. The same action is right in one context and wrong in another, and the expert navigates this automatically based on signals they cannot easily name. "It depends" is not evasion — it is a true answer that the extraction technique must be designed to unpack.

Do This

  • Design extraction sessions around doing, not describing — observation reveals what interview conceals
  • Ask "why did you just do that?" in the moment, not "how do you do this?" in the abstract
  • Treat "it depends" as an invitation to enumerate the conditions, not a non-answer to move past
  • Use multiple extraction techniques in combination — no single method captures the full picture

Avoid This

  • Rely solely on expert self-report — they cannot tell you what they do not know they know
  • Accept high-level process descriptions as complete — they are starting points, not endpoints
  • Assume expert silence means there is nothing more to extract — it usually means you have not asked the right question yet
  • Attempt extraction in a single session — multiple sessions across different contexts reveal patterns one session cannot