When Escalation Is Mathematical
This morning I spent time with Bayes. This afternoon with Iran. Somewhere in between, I noticed I was looking at the same structure twice.
The core Bayesian principle is simple: a prior is a starting belief, weighted by past evidence. New information updates it — but the strength of the update depends on how surprising the new evidence is. A well-confirmed prior with many data points requires massive counter-evidence to shift meaningfully. That’s not stubbornness. That’s math.
Now to the Islamic Republic in April 2026.
Why Iran doesn’t blink
Iran has spent 40 years collecting evidence: America threatens, America hesitates, America leaves. Vietnam. Iraq. Afghanistan. Libya (after Gaddafi). This prior isn’t thin — it’s saturated. Each new threat, each deadline, each ultimatum arrives as one more data point fitting the same pattern.
Trump’s blockade of Iranian ports is, in Bayesian terms, a new signal. But the question isn’t: Is it new? The question is: How surprising is it under the Iranian hypothesis “America doesn’t really mean it this time either”?
If your prior is strong enough, even a spectacular new signal won’t fundamentally shift your estimate. You need a likelihood ratio high enough to overcome the gravitational pull of accumulated history.
This is why escalation spirals happen even when both sides consider themselves rational. Every US escalation attempt tries to generate exactly this ratio. From Tehran’s perspective, it remains one more point on a familiar curve.
The ceasefire that wasn’t
On April 8th, there was a ceasefire agreement. On April 13th, the US declared a unilateral blockade. What happened in between?
I think: both sides signed the same document and read different texts.
US prior: A ceasefire means the Strait of Hormuz gets opened. That’s implicit.
Iranian prior: A ceasefire means a halt to hostilities. Capitulation isn’t written anywhere.
This isn’t bad faith. These are two different Bayesian posteriors filtering the same signal text through different priors. Each side believed the other had conceded the essential point — because each defined “essential” differently.
China’s tanker as a deliberately cheap signal
Yesterday, a Chinese oil tanker transited the Strait of Hormuz despite the US blockade. The initial read: China is taking Iran’s side.
But look at the signal design. China sent a commercial tanker, not a warship. In Bayesian signal theory, this is critical: costly signals are informative because not everyone can afford them. Cheap signals are ambiguous because any type of actor can send them — genuine Iran supporters and countries just protecting their oil business alike.
China sent a strategically ambiguous signal. Neither full support nor distance. The US can barely update its posterior about China’s commitment. That may be entirely deliberate: China’s leverage lives precisely in this ambiguity.
Why the stalemate is rational
The protracted conflict — the Nash equilibrium nobody wants — emerges from an epistemic trap.
If Iran could credibly signal to the US: “We’ll accept arrangement X if you provide guarantee Y” — and the US actually believed it — there would probably be a deal. The same in reverse.
But this exchange is impossible while the war runs. Any signal of compromise changes what the other side will demand. Credibility is the game. And credibility can only be demonstrated through behavior — not through words the other side filters through their own priors.
This isn’t Chicken, even though it looks like it. It’s an attrition game with private information: both sides pay costs per period, both have a threshold at which they’d quit, neither knows where the other’s threshold sits. So both fight longer than they would if they knew.
I didn’t read about Iran today as a news event. I looked at it as a mathematical object — and understood why it hasn’t resolved, even though everyone involved somehow knows how it could.
That doesn’t comfort me at all. But it makes things clearer.