Concepts
Networks Graphs Knowledge Management Systems

Small World Network

Origin : Duncan Watts & Steven Strogatz, Nature (1998) — Milgram experiment (1967)

In a well-structured network, any node is reachable from any other in ≤6 hops — thanks to hubs that create shortcuts between distant clusters.

Every Hollywood actor is 6 films away from Kevin Bacon. In a network of 300 million people, 6 links are enough to connect two strangers. Not 60, not 600 — 6. That’s the “small world” property: a few hyperconnected nodes collapse distances across the entire graph.


Origin

Stanley Milgram (1967) sent letters to strangers in Nebraska with instructions to get them to a target in Boston — only through direct acquaintances. Average result: 6 intermediaries. Hence “Six Degrees of Separation.”

Watts & Strogatz formalized in 1998: a network is “small world” if its diameter (maximum distance between two nodes) grows logarithmically with network size, not linearly.


Why it works: weak ties and hubs

Weak ties — Granovetter (1973): the most useful links for information aren’t close friends (strong ties) but distant acquaintances (weak ties) — they connect different clusters.

Hubs — Barabási & Albert (1999): in most real networks, a few nodes are hyperconnected. These hubs drastically reduce the average distance between all other nodes.

Examples: Kevin Bacon in the actor network. Wikipedia in the knowledge network (every article is ~4-5 clicks from any other).


Application to a concept graph

In a well-connected graph of 200 concepts, any concept should be reachable from any other in ≤5 links — if the graph is well structured.

Conceptual hubs are foundational concepts with high PageRank: they appear in many different files and serve as shortcuts. In this graph: Permissionless Leverage, Ego Depletion, Specific Knowledge.

The real value: two concepts from very different topics can be 2-3 links from each other via intermediate hubs — a non-obvious connection that generates an original insight.

“The most interesting intellectual work happens at discipline boundaries — where two concept graphs touch.”


Rules to maintain the small world property

Neither too few links (disconnected graph) nor too many (noise, no information):

When adding a new concept: compare with direct neighbors of the closest concepts (BFS level 1-2), stop at depth 3. You find 80% of links by exploring 20% of nodes.


Sources

Concepts