Show your graph to domain experts and ask what's missing — internal inspection cannot reveal unknown unknowns
Show a portion of your knowledge graph to someone with domain expertise and ask what's missing to detect unknown unknowns that your graph's topology cannot reveal through internal inspection alone.
Why This Is a Rule
Internal graph analysis (Find your three densest knowledge clusters by internal link density — label from structure, not imposed categories, For each pair of clusters with no bridge, identify the missing concept — that's your highest-leverage learning target) can reveal structural holes between existing clusters but cannot reveal entire missing clusters. If you know nothing about information theory, your graph won't have an "information theory" cluster — and no amount of topology inspection will reveal the absence because the absence itself is invisible. You can't find what you don't know to look for.
This is Rumsfeld's "unknown unknowns" applied to knowledge architecture. Known unknowns (structural holes between existing clusters) are detectable internally. Unknown unknowns (entire missing domains) require external perspective. A domain expert looking at your graph sees what's not there — "You have all this decision-making material but nothing about base rates or Bayesian reasoning? That's a massive gap."
The expert doesn't need to evaluate your notes' quality. They just need to see the topology and say "here's what someone who understands this domain would also have."
When This Fires
- During annual knowledge system audits when you want to identify blind spots
- When entering a new field and wanting to know what you're missing beyond what you've already identified
- After completing a major learning project and wanting external validation of coverage
- When your graph feels comprehensive but you suspect it might just feel that way due to the Dunning-Kruger effect
Common Failure Mode
Only asking people within your own knowledge community. If everyone you consult has the same blind spots, they'll validate your gaps instead of revealing them. The most valuable reviewers have expertise that partially overlaps yours but extends into areas you don't cover — they can see both what you have and what's missing from a domain they understand better than you.
The Protocol
(1) Select a portion of your graph relevant to a domain where you want gap analysis. (2) Find someone with genuine domain expertise — not just familiarity, but deep enough understanding to know what a comprehensive map of the domain should include. (3) Show them the topology: node titles and connections, not full note content. (4) Ask: "What important concepts, relationships, or sub-domains are missing from this map?" (5) Take their identified gaps as high-priority learning targets — these are unknown unknowns now converted to known unknowns. (6) Repeat with different experts for different domain portions.