Chapter 43
Coherence Clusters and Their Nerve
43.1 Coherence Clusters
A coherence cluster is a maximal compact subpopulation: a compact population \(\mathcal{C} \subseteq \pop\) such that for any agent \(A \in \pop \setminus \mathcal{C}\), the population \(\mathcal{C} \cup \{A\}\) is not compact.
Maximality means the cluster cannot absorb any additional agent without losing its consensus property. The boundary of a coherence cluster is the locus where one more agent would introduce an irresolvable oscillation. This is the precise topological meaning of the intuitive notion of a coherence horizon.
The coherence clusters are the topological refinement of the communication-radius clusters from Section 41. The communication radius \(B_\epsilon(A)\) gives a metric notion of cluster; compactness gives a topological (and therefore coordinate-free) notion. The relationship between these two notions — and in particular whether they coincide under reasonable conditions on the GSLT — is an open question noted in Section 45.
43.2 The Nerve of the Covering
Let \(\{\mathcal{C}_i\}_{i \in I}\) be the collection of all coherence clusters of \(\pop\). This collection is a covering of \(\pop\): every agent belongs to at least one coherence cluster.
The nerve \(\nerve\) of the cluster covering is the simplicial complex whose:
vertices are the coherence clusters \(\mathcal{C}_i\);
\(k\)-simplices are collections \(\{\mathcal{C}_{i_0}, \ldots, \mathcal{C}_{i_k}\}\) of \(k+1\) clusters with non-empty mutual intersection: \(\mathcal{C}_{i_0} \cap \cdots \cap \mathcal{C}_{i_k} \neq \emptyset\).
The nerve \(\nerve\) encodes the global topology of the population at the cluster scale. Its vertices are the coherent communities; its edges are the communities with common ground; its triangles are the communities that all share a common ground; and so on. This combinatorial object is the coarse-grained map of the population’s ontological landscape.
Two clusters with empty intersection share no common ground. No consensus is possible between them, even in principle. They are ontologically disjoint: there is no HML formula that both communities can agree on. In the context of the hierarchical population, these are the communities that have drifted so far apart in their world models that communication between them is not merely difficult but meaningless.
Two clusters with non-empty intersection share a compact subspace on which they agree. This overlap is the fragment of ontology that both communities can coherently discuss. Cross-cluster communication is possible, but only through the shared vocabulary of the overlap region. The overlap is the common ground in the precise topological sense.
43.3 The Hierarchy
The degradation argument of Section 41 applies recursively. The coherence clusters are themselves agents (by assumption on the population scale). Among these cluster-agents, communication degrades at the inter-cluster scale. Applying the same argument, the cluster-agents fragment into clusters-of-clusters. The process iterates.
The ontological hierarchy of a massive population \(\pop\) is the sequence of nerves \[\nerve_0 \;\leftarrow\; \nerve_1 \;\leftarrow\; \nerve_2 \;\leftarrow\; \cdots\] where \(\nerve_0\) is the nerve of the base-level clusters, \(\nerve_1\) is the nerve of the cluster-level clusters, and so on. The arrows are coarse-graining maps (simplicial maps that collapse finer structure).
The ontological hierarchy is the computational analogue of the hierarchy of scales in physics: quarks \(\to\) hadrons \(\to\) nuclei \(\to\) atoms \(\to\) molecules \(\to\) cells \(\to\) organisms \(\to\) planets \(\to\) galaxies \(\to\) superclusters. Each level appears self-contained to an observer at that level. The coarse-graining maps are the analogues of effective field theory truncations: they discard the fine-grained structure below the observer’s resolution.