emsenn

Proxy Networks as Complex Adaptive Systems

8 min read

Proxy Networks as Complex Adaptive Systems

1. The wrong question

The asymmetric escalation analysis identified “coordination ambiguity” as an intelligence problem: the degree to which proxy actions are directed by Iran versus independently initiated is itself a question the intelligence system must answer. The standard framing asks: does Iran control its proxies? The question assumes a spectrum from full control to full independence, with the intelligence task being to locate each proxy’s position on that spectrum.

Complex adaptive systems (CAS) theory suggests this question is malformed. In a complex adaptive system, macro-level behavior emerges from micro-level agent interactions. The proxy network does not need to be “directed” to produce coordinated behavior — coordination emerges from the agents’ responses to shared conditions, shared incentives, and shared signals. The question “does Iran control Hezbollah?” may be as misleading as asking “does the queen bee control the hive?” — the answer is simultaneously yes (she provides essential coordination signals) and no (the hive’s behavior emerges from the interactions of thousands of agents responding to local conditions).

2. The proxy network as CAS

Iran’s network of allied non-state actors — Hezbollah in Lebanon, Ansar Allah (Houthis) in Yemen, Kataib Hezbollah and other militias in Iraq, Palestinian Islamic Jihad and Hamas in Gaza — exhibits the defining properties of complex adaptive systems:

Multiple interacting agents with local decision-making

Each proxy is an independent political entity with its own leadership, constituency, political survival requirements, and strategic calculus. Hezbollah is a Lebanese political party and military organization; its decisions reflect Lebanese domestic politics, its institutional interests, and its assessment of its own military situation after the 2024 war. The Houthis’ calculus incorporates their position in the Yemeni civil war, their control of territory, and their own political identity. Iraqi militias operate within Iraqi factional politics, the U.S. military presence, and competition with the Iraqi state.

These agents make decisions locally. A Houthi commander deciding whether to launch anti-ship missiles in the Red Sea considers local conditions — target availability, munition stocks, his own unit’s readiness, the political messaging value within Yemen — alongside any signals from Tehran. The decision is locally generated even when it is broadly aligned with Iranian interests.

Emergent coordination without central direction

The proxy network can produce what appears to be coordinated behavior without requiring specific orders from Tehran. The mechanism is shared signals rather than command authority:

  • Shared trigger. The 2026 strikes constitute a shared shock — every proxy in the network receives the same stimulus (attack on the patron state, assassination of the Supreme Leader) and responds according to its own local logic. The responses may look coordinated because the stimulus is shared, not because a coordinating authority directed specific actions.
  • Shared ideology. The “axis of resistance” narrative provides a common framework within which each proxy interprets events and selects responses. This narrative is a coordination mechanism — it produces aligned behavior without requiring specific instructions.
  • Pre-delegated authorities. Iranian military doctrine reportedly includes pre-delegated authorities for certain contingencies — proxy commanders have prior authorization to initiate specified operations if specified triggers occur. These pre-delegations allow coordination to survive the destruction of the central node (Khamenei’s death, communication disruption) because the coordination was embedded in the system before the disruption.

Path dependence

Each proxy’s post-strike behavior is shaped by its specific history — the 2024 Hezbollah war’s outcome, the Houthi campaign against Red Sea shipping, the Iraqi militia attacks on U.S. bases. These histories constrain future behavior: Hezbollah’s degradation in 2024 limits its escalation capacity; the Houthis’ successful disruption of Red Sea shipping provides a proven model they will repeat; Iraqi militias’ operating environment depends on the Iraqi government’s position. The network’s past shapes its future in ways that a snapshot assessment of current capabilities cannot capture.

Nonlinear effects

The 2024-2025 period demonstrated CAS-characteristic cascading effects. Hamas’s degradation after October 7 contributed to Hezbollah’s weakening, which contributed to the fall of Assad’s government in Syria. The cascade was disproportionate to any single node’s importance — the network’s interconnection meant that degradation at one point propagated nonlinearly through the system. The 2026 post-strike environment may produce analogous cascading effects, but CAS theory warns they are not predictable in advance — the direction of cascading (further fragmentation? compensatory escalation by surviving nodes? emergence of new network configurations?) depends on local conditions at each node.

3. What CAS analysis changes

The intelligence question shifts

From: What is Iran ordering its proxies to do? To: What behavior will emerge from the proxy network given the current conditions at each node?

This shift has collection implications. The standard approach — intercepting communications between Tehran and proxy leadership (SIGINT), recruiting sources in the coordination chain (HUMINT), monitoring Iranian weapons shipments (IMINT) — targets the hierarchical model. CAS analysis requires additional collection:

  • Local-level monitoring. What are the conditions at each node? What is Hezbollah’s domestic political situation? What is the Houthis’ military posture? What are Iraqi militia factional dynamics? These local assessments are at least as important as the Tehran-to-proxy communication channel.
  • Network topology monitoring. How are the nodes connected? Are new connections forming? Are existing connections breaking? The network’s topology — not just its nodes — determines its emergent behavior.
  • Pattern recognition. CAS behavior is characterized by patterns that are recognizable in retrospect but difficult to predict. The intelligence system should look for emerging patterns in proxy activity — changes in tempo, target selection, geographic distribution — that signal a macro-level behavioral shift before it fully manifests.

Decapitation effects are uncertain

The assassination of Khamenei does not necessarily paralyze the proxy network. In a CAS framework, removing the central node produces one of several possible outcomes:

  • Fragmentation. Each proxy reverts to fully autonomous operation, producing uncoordinated but individually threatening behavior. This is harder for the intelligence system to track (no central communications to intercept) but may be less strategically coherent.
  • Emergent coordination. The proxies self-organize around a new coordination mechanism — an IRGC successor, a regional figure, or simply the shared ideology and pre-delegated authorities that survive the central node’s destruction. The behavior may be less precise but more resilient than hierarchical direction.
  • Compensatory escalation. Individual proxies escalate independently to demonstrate continued relevance and commitment, producing a period of increased activity that is locally rational but not centrally planned. This creates an analytical illusion: the activity looks like a coordinated retaliatory campaign but is actually multiple independent actors responding to the same stimulus.

CAS theory does not predict which outcome will occur — it predicts that the outcome is not predictable from the decapitation event alone and depends on the local conditions at each node, the network’s topology, and the path-dependent histories of each agent.

Indicator sets must change

Indications and warning for a hierarchical adversary monitors command-level indicators: leadership communications, mobilization orders, logistics movements. I&W for a complex adaptive system must monitor emergent indicators:

  • Changes in network interaction rate. Are the proxies communicating more or less with each other? With Tehran? With new external actors?
  • Local-level tempo shifts. Is any individual proxy changing its operational tempo in ways that its local conditions alone do not explain?
  • Novel behavior. Is any proxy employing tactics, targeting, or messaging it has not used before? Novelty in a CAS signals adaptation — the system is reconfiguring in response to the changed environment.
  • Network topology changes. Are new connections forming between previously unconnected actors? Is the network restructuring around new hubs?

4. Assessment

The intelligence system’s default model of Iran’s proxy network — a hierarchy with Tehran at the top issuing orders that flow downward to executing proxies — is a legibility operation. It makes the network comprehensible in categories the intelligence system can process: command relationships, communication channels, weapons flows. CAS analysis does not reject this model entirely — Iranian influence over its proxies is real and material (funding, weapons supply, training, doctrinal guidance). But it warns that the hierarchical model underestimates the network’s emergent properties: its capacity to produce coordinated behavior without central direction, to adapt to the loss of nodes, and to generate responses that no single decision-maker planned.

In the post-strike environment — where the central node has been destroyed, communications may be disrupted, and each proxy faces its own local conditions — the CAS model is likely more accurate than the hierarchical model. The intelligence system that monitors only the Tehran-to-proxy channel will miss the behaviors that emerge from the network’s internal dynamics. The system that monitors local conditions at each node, network topology changes, and emergent patterns will produce better assessments of what the proxy network will do — even if it cannot predict specific actions with the precision that hierarchical models promise but do not deliver.

Graph View

Backlinks