Three Densities

Nighthaven⛺︎ April 5, 2026
Source
  1. Overview A platform grows. User count rises, follow edges multiply, the graph thickens. The metrics say the space is thriving. The participants say the space is dead. This divergence is not a feeling. It is a structural phenomenon. Growth indicators and community vitality can move in opposite directions — not because someone measured wrong, but because they measured different things. This essay models the divergence. Three layers of density — graph, feed, perceived — carry three distinct quantities. The conversion between them is non-conservative. What determines whether a space is alive or dead is not the layer easiest to measure. It is the layer hardest to measure. The argument crystallizes observations first laid out in "The Still Water." That essay described. This one models.
  2. Definitions Density is not user count. Density is the probability that any two active participants share context: overlapping projects, mutual acquaintances, compatible ambitions. When this probability crosses a threshold, a subculture forms. Below it, you have a population. Above it, you have a scene.
  3. Three Densities D₁ (Graph Density) — Structural density based on follow relationships and mutual connections. Approximated by node count, edge count, clustering coefficient. Externally measurable. D₂ (Feed Density) — The probability that a participant's feed surfaces posts from context-sharing contacts. Transformed from D₁ by feed algorithms, custom feeds, mute settings. D₃ (Perceived Density) — A participant's subjective perception of density. Composed of three conditions:
  • Identity: a felt sense of what you are in this space.
  • Role: recognition that your presence contributes to the texture of the space.
  • Sustained anticipation: the sense that something might be happening even when you are not looking.
  1. Propositions Proposition 1. The life or death of a space is determined by D₃. Not D₁. Not D₂. High D₁ with sub-threshold D₃ produces a space that participants experience as dead. Low D₁ with supra-threshold D₃ produces a functioning scene. Proposition 2. The conversion D₁ → D₂ → D₃ is non-conservative. Density is not preserved across layers. A rise in D₁ does not guarantee a rise in D₃. Proposition 3. The divergence between D₁ and D₃ is the research object of Network Perception. Conventional network analysis measures community health by D₁. This model places the divergence itself — its magnitude and direction — at the center of analysis.
  2. State Classification False Growth is the distinctive prediction. Platform operators read D₁ — user count, DAU, follow edges — and see growth. Participants read D₃ and feel stagnation. Two opposite diagnoses of the same space coexist.
  3. Divergence Mechanisms Three mechanisms attenuate density on its path from D₁ to D₃. Context Dilution — Users increase but new arrivals share no context with existing participants. The denominator grows. The probability that any two people share context drops. D₁ rises while diluting itself from within. Feed Occlusion — Two nodes are connected on the graph but the feed algorithm does not surface one to the other. Engagement optimization accelerates this: low-context popular posts crowd out high-context posts from known contacts. Response Void — A participant sees posts from context-sharing contacts in the feed but receives no context-aware response to their own posts. The feed works. The loop does not close. The space is comfortable and quiet. The three mechanisms operate independently. When multiple mechanisms act simultaneously, divergence escalates.
  4. Corollaries Corollary 1. Product quality has a non-monotonic effect on D₃. Better moderation and removal of algorithmic rage keep D₁ and D₂ healthy. But removing friction can undermine the "sustained anticipation" component of D₃. The relationship between product quality and D₃ is not monotonically increasing. Corollary 2. The threshold for scene formation exists on the perception side. The threshold is a function of participant perception, not graph structure. The same D₁ can yield different D₃ values. Corollary 3. Density and scale are independent variables. A small community can have high D₃. A massive platform can have D₃ of zero.
  5. Formalization Sperber and Wilson (1986/1995) defined relevance as proportional to cognitive effect and inversely proportional to processing effort. It clarifies directionality: what increases relevance, what decreases it. The same move applies here. L ∝ D₃ / D₁ Liveness is proportional to perceived density and inversely proportional to graph density. This says one thing: liveness is not how large the graph is but how much of that graph converts into perceived density. When D₁ rises and D₃ does not follow, liveness drops. A growing platform feels dead because the denominator grew. D₂ does not appear in the formula. It operates as a transmission coefficient between D₁ and D₃ — the rate at which graph structure converts into feed exposure. When D₂ is low relative to D₁, the conversion is lossy. The formula captures the endpoints. D₂ explains why the conversion fails. This is not a computational model. It is a directional claim in the form of a ratio. The value of the formalization is not calculation but compression: one expression that answers why user growth and community vitality can move in opposite directions.
  6. Open Questions This model defines D₃ conceptually but does not operationalize it. Identity, role, and sustained anticipation are qualitative descriptions. At least three paths toward quantification are conceivable: self-report surveys measuring subjective community experience; response-network analysis constructing a graph of who responds to whom with context-aware content; or computational approaches inferring D₃ from time-series patterns in posting behavior. None of these have been attempted. The model has drawn a map. The surveying has not begun.

Discussion in the ATmosphere

Loading comments...