Main idea

In a wiki of notes built from books, each book gets a reading map that points to every note drawn from it. In a 1,013-note marketing theme fed by some twenty books, the ten most connected notes are all reading maps: the one for April Dunford's Obviously Awesome has 180 links, the one for Seth Godin's This is Marketing has 156.

That weight is enough to steer the automatic split into groups. A community detection algorithm groups what is densely linked, and what is densely linked is the notes from the same book around their reading map. Positioning, a single concept, ends up split across five groups, mostly by source book: Dunford on one side, Ries and Trout on the other. Paths suffer the same effect: the one connecting the chasm to willingness to pay goes through three reading maps, which is accurate but teaches nothing.

A knowledge base filed by concept in its notes thus becomes, in its graph, a base filed by source again.

Why it matters

This shows that a navigation structure is not neutral for analysis. A node created to orient the reader becomes, for a graph computation, the centre of gravity of everything it serves, and the map you get draws the library rather than the ideas.

It also gives a concrete adjustment: before reading the groups of a graph of notes, remove or lighten the navigation nodes, or at least know that the groups found carry their imprint.

Nuances and limits

Grouping by book is not wrong: an author builds a coherent system, and their ideas really are more linked to one another than to another author's. The mistake is to take that grouping for a map of concepts.

And reading maps also do the graph a service: they link notes that would otherwise stay isolated.

Open questions

• What weight should reading maps carry in the computation for groups to follow concepts?