Spatial community structure impedes language amalgamation in a population-based iterated learning model

Spatial community structure impedes language amalgamation in a population-based iterated learning model

Source: cs.MA updates on arXiv.org — https://arxiv.org/abs/2305.11962 Date read: 2026-09-02 Connected to: L-003, L-010 Kind: content Escalation: store-only Escalation rationale:

What this is

An agent-based model study extending the iterated learning framework to examine how spatial/community topology affects language convergence in multi-agent populations. The core finding is that increasing between-community communication frequency above a critical threshold disrupts local language coherence, producing a non-monotonic adoption curve.

What I took from it

The work provides a computational instantiation of coordination adoption nonmonotonicity (L-010) in a stripped-down linguistic domain. The key observation is that populations with weak inter-community ties maintain internal linguistic coherence through local norm-locking, but modest increases in cross-community signal introduce conflicting coordination pressures that fragment both local and global convergence. This suggests the nonmonotonicity is not primarily about threshold effects in adoption payoffs, but about coherence collapse under multiple competing coordination attractors.

However, the paper does not investigate the mechanism driving the coherence loss—whether it's genuine incompatibility between local languages, information cascade disruption, or simply noisy signal propagation. The domain is also highly artificial (synthetic language evolution), which limits immediate generalization to protocol systems. The work does not engage with formalization, legibility, or the enforcement mechanisms central to L-003 (The Formalization Ratchet), so the connection there is suggestive rather than direct.

Research connections

• L-010: Direct instantiation of coordination adoption nonmonotonicity in spatial topology; confirms non-monotonic relationship between inter-community signal strength and global convergence.
• L-003: Tangential; local language hardening could map to formalization under stress, but paper does not examine formality or enforcement pressure.
• seed-070: Multiple local coordination equilibria functioning as infrastructure constraints; paper shows these resist amalgamation under moderate cross-community pressure.
• seed-078: Learning-race defection; local groups maintain distinct languages partly through learning isolation, suggesting pooling resistance.

Seed

Seed title: Coherence Collapse Under Weak Attractor Multiplicity

Seed type: observation

Seed text: In coordination systems with spatial or social community structure, modest increases in inter-group signal strength can fragment rather than unify global coordination, because local coordination attractors harden through repeated reinforcement before cross-group signals reach sufficient amplitude to override them. The collapse point occurs not at adoption threshold but at the conflict-zone between competing stable equilibria. This generalizes beyond language: any protocol system where agents partition into communities with independent norm-locking will exhibit nonmonotonic global coherence as inter-partition communication increases, with maximum fragmentation occurring in the mid-range of cross-partition signal strength—below which communities ignore external signals, above which one attractor dominates globally.