The war that began in Gaza three months ago is increasingly becoming a regional system of interconnected confrontations. Hamas deputy political leader Saleh al-Arouri was killed in Beirut on Tuesday, in an explosion widely attributed to Israel, although Israel has not formally claimed responsibility. The location matters enormously: Arouri was killed in the southern Beirut suburbs where Hezbollah is deeply influential, meaning the strike did not merely remove a senior Hamas figure but demonstrated that the conflict could reach directly into territory associated with one of Israel’s most powerful regional adversaries. Hezbollah has promised a response, while Israel and Hezbollah are already exchanging fire regularly across the Lebanese border.
At the same time, Yemen’s Houthis continue attacking commercial shipping in the Red Sea, American forces have intercepted missiles and drones, Iranian-backed militias have attacked US forces in Iraq and Syria, and the United States is attempting to deter these groups without becoming another direct combatant in a widening regional war. These fronts are connected politically through the Gaza conflict and through Iran’s relationships with armed organisations across the region, but they are not controlled through one simple command structure. That makes escalation more difficult to manage because several actors possess enough independent agency to create consequences for everyone else.
The conventional way to imagine escalation is through intention: one side wants a larger war, takes increasingly aggressive actions and eventually gets one. The Middle East currently demonstrates a more disturbing mechanism. Actors can collectively produce escalation while individually trying to avoid all-out war. Each takes limited actions intended to restore deterrence, demonstrate resolve or support an ally. The opponent then responds for exactly the same reason. Nobody necessarily chooses the final level of confrontation at the beginning.
This is the logic of an escalation ladder. Imagine every military action occupies a rung. Hamas attacks Israel. Israel responds in Gaza. Hezbollah increases attacks along the northern border. Israel strikes Hezbollah positions. The Houthis attack shipping. American and allied naval forces intercept them. Militias attack American bases. Washington retaliates against militia facilities. Each individual movement may appear limited relative to the previous one, yet the system as a whole continues climbing.
The problem is that stepping down can look strategically dangerous. If Hezbollah attacks and Israel does not respond, Israeli leaders may fear Hezbollah interprets restraint as weakness. If Israel kills a senior Hamas leader in Beirut and Hezbollah does nothing, Hezbollah may fear losing credibility among supporters and allies. If the Houthis repeatedly threaten shipping and Washington does not impose costs, American officials may fear attacks become normalised. If Washington retaliates, the Houthis or their partners have incentives to demonstrate they were not deterred.
Deterrence therefore creates escalation incentives precisely because every actor wants to prevent future escalation.
That sounds contradictory until we recognise what deterrence requires. To discourage an opponent from taking an action tomorrow, a government sometimes needs to impose costs for the action taken today. The punishment is intended to prevent repetition. But the opponent observes punishment and may conclude it needs to retaliate to prevent the punisher from believing future attacks are cheap.
Both sides therefore describe retaliation as defensive.
This is one reason the language of self-defence cannot by itself tell us whether a conflict is becoming safer. Almost every participant can interpret its own actions as responses to somebody else’s previous action.
The sequence matters.
So does the point at which somebody decides not to continue it.
Restraint becomes a strategic resource, but it is difficult to use because restraint needs to be distinguishable from inability. A powerful actor choosing not to retaliate can potentially communicate confidence and create space for de-escalation. An actor perceived as weak may worry the same decision encourages further pressure.
Effective de-escalation therefore requires credible restraint: the opponent needs to understand that you could escalate but are deliberately choosing not to under specified conditions.
Diplomacy exists partly to communicate those conditions.
Without communication, military behaviour becomes the principal language through which intentions are inferred.
That language is dangerously ambiguous.
A military mobilisation intended to deter attack can look like preparation for attack. A limited strike intended to close a confrontation can look like the opening of another phase. An air-defence deployment intended to protect forces can convince an adversary that offensive operations are imminent.
Capabilities are visible.
Intentions are not.
The Israel-Hezbollah confrontation illustrates the danger particularly clearly because both sides possess memories of the 2006 Lebanon War, which caused enormous destruction without eliminating Hezbollah. Hezbollah is now substantially better armed than it was then, possessing a large arsenal of rockets and missiles capable of striking throughout Israel. Israel possesses overwhelming conventional air power and could inflict extraordinary damage on Lebanon, while Hezbollah could impose costs on Israeli cities far beyond anything Hamas can generate from Gaza.
That creates mutual deterrence.
It also creates mutual vulnerability.
The same weapons that make each side reluctant to begin a full war make every limited exchange potentially more dangerous because both know how catastrophic escalation could become.
This is similar to the logic of nuclear deterrence at a much lower level of destructive capacity: the possibility of enormous retaliation can stabilise the highest level of conflict while making competition below that threshold persistent.
Actors probe for the amount of pressure they can impose without triggering the larger confrontation.
Political scientists sometimes describe this as competition inside the “grey zone” below full-scale war. Hezbollah can fire limited numbers of rockets. Israel can conduct targeted strikes. Both communicate that escalation remains controlled.
The difficulty is that the boundary separating controlled violence from uncontrolled war is not marked physically.
Nobody knows exactly which strike crosses it until somebody decides that it has.
Arouri’s killing tests that boundary because location changes the meaning of an operation. Killing a Hamas leader in Gaza during an active war would be interpreted differently from killing a senior Hamas figure in Beirut. The military target may be similar while the sovereign territory and political audience differ.
Geography therefore changes escalation risk.
Hezbollah now needs to decide whether the strike requires a qualitatively stronger response because allowing senior allied figures to be killed inside its Beirut stronghold could weaken its deterrent reputation.
Israel simultaneously needs to estimate how much Hezbollah can retaliate without requiring another Israeli escalation.
This is bargaining conducted through violence.
The enormous problem is that bargaining through violence kills people while information is being exchanged.
Diplomatic bargaining allows actors to communicate preferences through words.
Military bargaining communicates through costs.
Misunderstanding becomes much more expensive.
Iran’s role makes the system more complicated because Tehran has developed relationships with armed groups across the region partly as a form of forward deterrence. Hezbollah, the Houthis and militias in Iraq and Syria give Iran ways to impose costs on adversaries without relying entirely on direct Iranian military action.
From Tehran’s perspective, this creates strategic depth. An Israeli or American attack on Iran can potentially generate responses from several geographical directions.
From Israel’s and America’s perspective, the same network appears as an architecture of regional coercion.
Again, one actor’s deterrence becomes another actor’s threat.
This is the security dilemma operating through proxies.
The word “proxy” can itself become misleading if it implies complete Iranian control. Iran provides funding, weapons, training and political support to several groups, but these organisations possess their own leadership, constituencies and interests. Hezbollah is deeply connected to Lebanese politics. The Houthis emerged from Yemen’s own political conflicts. Iraqi militias operate inside Iraq’s fragmented security and political environment.
Iran can influence them without necessarily determining every tactical decision.
That creates a principal-agent problem. A sponsor benefits from an armed partner imposing pressure on adversaries but does not necessarily control how much pressure the partner applies. The local organisation may accept greater risk than the sponsor prefers because its own political incentives differ.
The sponsor can therefore become entangled by actors it helped empower.
This is another route through which wars expand without central intention.
Suppose a militia kills a large number of American soldiers in an attack Tehran did not specifically order. Washington may retaliate against Iranian interests because it considers Iran responsible for enabling the militia. Iran then faces pressure to respond to an American attack it did not originally seek.
Agency becomes distributed.
Responsibility becomes contested.
Escalation becomes easier.
The Houthis demonstrate this particularly clearly. Their Red Sea campaign serves several objectives simultaneously. It allows them to present themselves domestically and regionally as defenders of Palestinians, strengthens their position within Iran’s broader regional network and gives them international influence far beyond Yemen’s economic weight.
The attacks therefore may remain politically valuable even if Iran itself prefers avoiding direct confrontation with Washington.
This is why destroying individual Houthi missiles may not change the incentive producing attacks.
Deterrence requires altering the expected political and military payoff.
That can involve defensive interception, economic pressure, diplomacy or offensive action.
Every option carries costs.
Striking Houthi positions could reduce their military capability while increasing domestic support for a movement able to portray itself as confronting the United States.
This is the martyrdom and mobilisation problem common to asymmetric warfare. Military punishment can weaken an organisation materially while strengthening its political narrative.
The same problem appears in Gaza. Israel can destroy Hamas infrastructure and kill commanders while simultaneously creating anger among Palestinians that another militant organisation could eventually mobilise.
Military effectiveness and political effectiveness therefore need to be evaluated separately.
This is why regional escalation cannot ultimately be solved only through superior firepower.
The United States possesses vastly greater military capability than the Houthis or Iraqi militias.
Israel possesses vastly greater conventional capability than Hamas.
Neither fact automatically produces political control over the system.
Weak actors survive partly because they do not need to defeat stronger actors conventionally.
They need to impose enough costs that the stronger actor changes behaviour.
This is asymmetric warfare in its most basic form.
The weaker actor chooses the domain where the power difference matters least.
Hamas uses tunnels and dense urban terrain.
Hezbollah uses rockets and missiles.
The Houthis use drones and maritime chokepoints.
Militias use dispersed attacks against bases.
The stronger actor then has to decide how much force to apply against an opponent embedded inside a wider civilian and political environment.
Superior military power can therefore produce diminishing strategic returns.
The regional economy is already absorbing some consequences. Red Sea shipping diversions increase transport costs, tourism across parts of the region faces uncertainty and Lebanon’s already devastated economy would be extraordinarily vulnerable to another major war.
This creates a strong economic incentive for restraint.
Economic incentives do not guarantee restraint.
Governments sometimes value security, ideology or political survival more than economic welfare.
Lebanon is a particularly striking example because the country has experienced one of the world’s worst modern financial collapses. A major Israel-Hezbollah war could destroy infrastructure and deepen poverty dramatically.
That gives Lebanese society enormous reasons to avoid war.
Hezbollah’s strategic calculations include factors beyond Lebanese GDP.
This demonstrates why states and armed organisations cannot always be modelled as actors maximising national economic welfare.
Political organisations maximise their own objectives.
Those objectives can diverge from the aggregate welfare of the populations among whom they operate.
This is another principal-agent problem, except the principal is society itself.
Democratic institutions attempt to reduce that gap by making governments electorally accountable.
Armed non-state organisations possess different accountability structures.
That makes predicting their risk tolerance harder.
Israel faces its own domestic pressures after October 7. The attack represented a catastrophic failure of Israeli security institutions, and residents of northern Israel have been displaced because of continuing Hezbollah fire. The government therefore faces strong pressure to demonstrate that citizens can return safely.
A temporary reduction in Hezbollah attacks may not be sufficient if residents believe the underlying threat remains.
This creates pressure for a more durable military solution.
But eliminating Hezbollah’s enormous arsenal through war could require a campaign producing extraordinary destruction without guaranteeing the organisation disappears.
Again, the strategic objective needs to be defined.
What does victory mean?
No rockets tomorrow?
No Hezbollah forces near the border?
Disarmament?
Political transformation inside Lebanon?
Each objective requires a radically different level of force and diplomacy.
Undefined objectives create escalation because military action continues without a clear condition for stopping.
The Gaza war already suffers from this problem through the phrase “destroy Hamas”.
Regional escalation could reproduce it on a much larger scale.
The United States therefore has a strong interest in forcing political objectives to become more specific. Washington supports Israel’s security while simultaneously attempting to prevent another front in Lebanon and protect American forces across the region.
Those objectives can conflict.
Supporting Israeli deterrence may require allowing or supporting actions that increase short-term escalation risk.
Preventing regional war may require restraining an ally after it suffers attacks.
Alliance politics therefore involves influence in both directions.
A useful ally is not one that agrees automatically.
Sometimes alliance value comes from having enough trust to tell a partner that another escalation does not serve its own long-term interest.
Whether Washington possesses that influence over Israel is increasingly important.
The same applies to Iran’s relationship with its partners.
Regional stability may depend partly on whether sponsors can restrain actors they previously encouraged to develop independent military capabilities.
This is the control paradox of proxy strategy.
Proxies are valuable because they possess local autonomy and can operate without direct state involvement.
That autonomy is exactly what makes them difficult to control during crisis.
The broader international community therefore confronts a conflict without one negotiating table capable of resolving every front.
A Gaza ceasefire would reduce some regional pressure but would not automatically resolve the Israel-Hezbollah dispute, Houthi power in Yemen, Iranian-American competition or militia activity in Iraq.
At the same time, Gaza provides the political centre around which several of these confrontations are currently mobilising.
Reducing violence there could therefore lower the temperature across the network even without solving the underlying disputes.
This is another characteristic of interconnected systems: one node can become disproportionately important without being the sole cause of instability.
The Middle East increasingly resembles a network in which several actors are connected through alliances, rivalries and armed groups.
Network systems can appear stable until enough interactions begin reinforcing one another.
Then local shocks propagate.
The assassination of one Hamas leader in Beirut therefore cannot be analysed only through the value of removing that individual.
Its strategic cost includes the probability that Hezbollah responds.
The value of Hezbollah’s response includes the probability Israel responds again.
Each expected future action should theoretically enter the calculation of the current one.
This is what makes escalation control so difficult.
No actor knows those probabilities perfectly.
Intelligence estimates intentions.
Leaders infer red lines.
Historical behaviour provides clues.
Then somebody makes a decision.
Errors accumulate.
Wars can therefore emerge from miscalculated expected values, not irrationality.
Every actor can make a decision that appears reasonable given its beliefs and collectively produce an outcome almost everyone considers disastrous.
That is perhaps the most frightening lesson from this week.
We often reassure ourselves that a regional war is unlikely because Israel does not want one, Hezbollah does not want one, Iran does not want one and the United States does not want one.
That information is genuinely reassuring.
It is not sufficient.
The First World War remains the historical warning against assuming catastrophic wars require every participant to prefer catastrophe. The analogy should not be applied mechanically because today’s Middle East is radically different from Europe in 1914.
The general mechanism remains useful.
Mobilisation.
Alliance commitments.
Credibility.
Retaliation.
Misperception.
Domestic pressure.
Each can narrow the space available for leaders to step down.
The challenge is therefore not merely convincing actors that regional war would be costly.
They already know.
The challenge is constructing off-ramps that allow them to avoid it without believing they have surrendered deterrence.
That requires communication.
Limited objectives.
Third-party mediation.
Clear red lines.
And sometimes deliberately refusing to retaliate at the maximum level available.
Restraint is not pacifism.
It is part of strategy.
The strongest actor is not necessarily the one that responds most aggressively to every provocation.
It can be the one confident enough to distinguish between threats requiring escalation and provocations better absorbed.
That judgement becomes harder after trauma, particularly for Israel after October 7.
It becomes harder for Hezbollah when senior allied figures are killed in territory it considers its stronghold.
It becomes harder for Washington when American personnel are attacked.
Every actor has a story explaining why this particular retaliation is necessary.
That is exactly how escalation ladders are climbed.
Nobody needs to announce that they want the top rung.
They only need to keep deciding that the next rung is safer than stopping where they are.
The Middle East enters 2024 with that danger becoming increasingly visible.
The war is still centred in Gaza.
It is no longer contained there.
The strategic objective for every government claiming not to want regional war should therefore be straightforward, even if implementation is not: stop treating every opportunity to demonstrate resolve as an obligation to use it.
Deterrence needs credibility.
So does restraint.
If the region loses the second while pursuing the first, nobody has to choose a wider war deliberately.
They can arrive at one one justified retaliation at a time.
