Agent-based modeling

What this method is for

Agent-based modeling (ABM) creates populations of agents with simple rules and observes what emerges from their interaction. It is a method for mechanism discovery — understanding how collective patterns arise from local dynamics.

ABM is particularly suited to questions about:

• Emergence — how macro patterns arise from micro interactions
• Heterogeneity — how diversity affects collective outcomes
• Thresholds — where regime changes occur
• Feedback — how local and global dynamics couple

What it can tell us

ABM can identify sufficient conditions for phenomena. If a model produces cascade failure when load-bearing agents lack relief pathways, this demonstrates that the mechanism is capable of producing the effect — not that it is the only cause in real systems.

ABM is a conceptual instrument — a sandbox for testing ideas, not a predictive simulator of specific real-world systems.

What it cannot tell us

ABM cannot:

• Prove that a mechanism operates in any particular real system
• Predict specific outcomes in complex real-world contexts
• Replace empirical observation or fielded practice

The relationship between model and world is always interpretive. Models generate insight, not truth. The map is not the territory.

Why it is appropriate for this inquiry

The coherence vs entrainment question is fundamentally about mechanism: what structural features of coordination regimes produce different failure modes under stress?

ABM allows controlled experiments impossible in real collectives — systematically varying parameters, running thousands of replicates, tracking agent-level dynamics at high resolution.

We’ve run 600+ experimental runs using NetLogo BehaviorSpace testing the Coherence Theorem, identifying that relief pathways (not diversity per se) are the causal mechanism preventing cascade failure.

Tools

• NetLogo — Visual, accessible ABM environment
• BehaviorSpace — Systematic experiment running
• Python/Jupyter — Statistical analysis of exports