Notes on War: Grand Strategy

by johnswentworth11 min read18th Jun 202117 comments



First things first: I am not any sort of military expert. But most discussions or portrayals of war which I see among rationalists make me face-palm, especially discussions around the role of AI in future warfare or AI “takeover”. Seriously, guys, it’s not about autonomous drone swarms.

This post is just a brain-dump of some models/frames which I wish more people had. We’ll walk through a few different “use-cases” of war (defending an invasion, territory grab, takeover) and talk about grand-strategy principles relevant to each. If people find this interesting, I may write more brain-dumps like this on strategy, tactics, etc; the post quickly became far too long to fit it all in one.

Epistemic status: any particular claim is low-confidence, but I am more confident in the usefulness of the high-level frames.

Defending an Invasion

For concreteness, let’s think about Azerbaijan invading the Nagorno-Karabakh region of Armenia (though admittedly I did not follow that conflict closely so we’re not going to talk about real details too much). At a high level, how might Armenia go about defending against that invasion?

The obvious option is “kill invading troops until they leave, and kill them some more if they come back”. I claim this is wildly inefficient: it’s treating symptoms, not causes.

On the Azerbaijani side, the decision to invade was presumably a political decision. There were factions in favor, and factions against. If the anti-invasion factions gain control of government decision-making, then the invasion will likely end. Killing invading troops might help make that happen - making the war visibly costly may make it less popular. On the other hand, killing invading troops may just create martyrs and solidify the image of Armenians as Enemies. If that’s the case, then killing invading soldiers may only solidify the political faction pushing for war, and troops will just keep invading until one side runs out of people/resources. It could go either way.

So: how could Armenia fight in a way which destabilizes the pro-invasion faction’s control over Azerbaijan, rather than stabilizing it? Some possibilities:

  • Directly attempt to kill key people in the pro-invasion faction, e.g. assassinations. Note that this could backfire in the same way as killing soldiers.
  • Target the power base of people in the pro-invasion faction. For instance, in many third-world countries the military has a lot of control over decision-making, because they might realistically execute a coup if they don’t get their way. In that case, killing soldiers (and officers) directly reduces the military’s ability to execute a coup, and therefore directly reduces their influence over decision-making.
  • Similarly, if the invasion is driven by a regional or religious faction with significant wealth/income, destroy their capital assets.
  • If a particular faction/stakeholder has a “swing vote” or functional equivalent, try to focus the costs of the war on that stakeholder - i.e. whenever the Azerbaijani forces take some territory or kill some Armenian soldiers or destroy something, try to retaliate by destroying something valuable to the “swing vote” stakeholder.

Alternatively, the Armenians could try to directly change the political calculus of the pro-invasion faction:

  • Whatever benefits the Azerbaijanis expect from invasion, reduce them. Organize resistance groups and weapons caches before an invasion happens. Rig explosives to destroy capital assets (i.e. bridges, roads, rails, power grid) remotely in case an area is lost. Make sure the Azerbaijanis know about all this, to dissuade invading in the first place.
  • Increase the costs of invasion. Build second-strike capabilities - weapons that can be used to retaliate after an attack. Use proportional retaliation. Again, make sure the Azerbaijanis know about all this beforehand.
  • Maybe it’s a populist war - i.e. the Azerbaijanis just really hate Armenians and will back any politician promising to attack them. In the long run, increasing the popularity of Armenians among the Azerbaijani populace is probably a good strategy, so “take the high road”: avoid civilian casualties, treat prisoners well, send aid (food, medical supplies) to the enemy populace, etc. And more importantly, make sure to loudly broadcast all those actions.
  • In the short run, to deal with an Azerbaijani populace which really hates Armenians, try to shift the populace’ attention to domestic affairs. A recession could do this, so long as it’s obvious that defeating Armenia will not make the recession go away. Same with a famine, though that’s harder to induce while avoiding long-term blame. Maybe target the sort of economically-crucial companies that people love to hate, like banks.
  • Also implicit in all of the above: propaganda, obviously.
  • Taking a different tack: if the actual priority of Azerbaijani leadership is to generate popularity with a populace which hates Armenians, it may be possible to give them a symbolic victory without actually losing much. “Win-win”, in some sense.

One important thing to keep in mind: Armenia is no more a single unified agent than Azerbaijan; political considerations will inform their actual choices in much the same way. If the Armenian populace really hates Azerbaijanis, then the politically-popular choice will likely be to kill lots of soldiers and probably also lots of Azerbaijani civilians, destroy their stuff, etc, regardless of whether that is actually a smart strategy for winning the war.

Another important thing to keep in mind: war is not a zero-sum game. A war of attrition leaves both sides worse off, so there’s positive-sum gains in avoiding that outcome. That means that sharing information is sometimes a good move: if destroying our stuff will be costly to the enemy as well, then we want the enemy to know that. Also, from a politician’s perspective, a war can solidify the political position of leadership on both sides - a positive-sum outcome, in some sense, though probably not good for the citizenry or soldiers on either side.

Territory Grab

Invasions have a lot more variety, depending on the objective. First, we’ll consider the Azerbaijani side of the previous example. Their goal, presumably, is to take political control of the Nagorno-Karabakh region. (Of course in practice this may not be the “real” goal - e.g. if it’s a populist war, the real function may just be to generate political support among the population by symbolically Punching Bad People In The Face.)

Broadly speaking, if the Azerbaijanis plan to absorb the new territory, then they generally want to not break it. Don’t blow holes in a ship you plan to steal. That means:

  • Avoid civilian casualties
  • Minimize damage to capital assets (i.e. infrastructure)
  • Don’t piss off the local population: no pillaging, provide food and medical supplies, be very strict about discipline with Azerbaijani troops, etc.
  • Broadcast how nicely Azerbaijan treats the locals. As usual, propaganda.

There are exceptions to all of these, depending on the objective and resources. If the plan is to wipe out the local population and resettle, obviously that changes things. If Azerbaijan has a lot more capital investment capacity than Armenia, then it could leverage that advantage by destroying local capital assets, which will be easier for Azerbaijan to replace than for Armenia to replace. If the plan is an iron-fisted occupation, then pissing off the locals once or twice may be useful in order to make a few public examples. But the general underlying principle is: don’t break what you intend to steal.

Beyond that, building governing infrastructure - courts, tax collection capabilities, regulatory enforcement, etc - takes a lot of work. An important strategic question for Azerbaijan is whether to take over the existing infrastructure or replace it. If the latter, then attacking existing government institutions may weaken Armenia during the war - but at the cost of Azerbaijan needing to rebuild from scratch once the war ends. More discussion of that trade-off in the next section.

In terms of combat priorities, Azerbaijan’s problem largely mirrors Armenia’s problem. Armenia’s decision to fight or negotiate a settlement is a political decision, and Azerbaijan can influence that decision in basically the same ways as the previous section, so long as they avoid damage to whatever they intend to take.

Full Takeover

Now, we’ll think about the US invading Japan in WWII. (Or communists taking over China in the 1949 revolution; similar principles apply, and indeed I hear that Mao’s writing on the topic is quite similar to the discussion here.) The objective is to overthrow and replace the existing government.

This runs into the same tradeoff as the previous section, but to a much greater extent: whatever destruction the invader causes will be the invader’s problem to clean up, assuming they win. Destroying capital assets or undermining government institutions may make it much easier to win the war, but it will make the aftermath much more difficult.

First question: is it a strategic priority to preserve the existing enemy government? Wiping out leadership can make it a lot easier to invade, but it also leaves nobody who can surrender on behalf of the whole country. Probably lots of decentralized resistance. Also, if the plan is to force a surrender, then enemy leadership needs to be able to enforce the terms of surrender on their own people. Preserving the enemy government’s ability to do that is a strategic necessity, in that case.

Second question: is it a strategic priority to maintain the competence of the existing enemy government? Mao notes that incompetent bureaucrats or officers in the enemy hierarchy are an asset for the rebels, and guerilla fighters should avoid targeting them. Conversely, competent bureaucrats or officers should be targeted. (Besides the obvious benefit, this has a bonus effect: if the enemy leadership knows about such a policy, then they will trust their middle management less, which will further slow organizational information-passing and decision-making.) However, this is a short-term-oriented view; those same competent bureaucrats could become assets for the new administration after the war. Training and installing new people, en masse, is difficult and expensive.

In the case of a guerilla rebellion, the same question applies not only to competence but to hostility/abuse. Bureaucrats or officers who abuse their power will generally drive support for the rebellion, whereas bureaucrats or officers who are liked by the populace will generally drive support for the regime. On the other hand, once the war is over, regardless of who won, getting rid of abusive bureaucrats/officers is part of the reason for fighting the war in the first place.

Third question: which enemy institutions are to be kept? This is where the political considerations from earlier come in: within each institution, there will be a decision to cooperate with the takeover or not, there will be factions and key decision-makers, etc. If the top-level government structure is to be maintained, then the same considerations from earlier sections apply. But if only lower-level structures are to be maintained - e.g. courts, military, school system, road maintenance, government hospitals, etc - then high-level decision makers at each of those individual institutions must be persuaded, coerced or replaced. Furthermore, the ability of those high-level decision makers to enforce their decisions on the rest of the institution must be maintained.

A Continuum Between Politics and Warfare

When we open the black box of “enemy government” and start manipulating internal gears of decision-making - e.g. political factions, particular institutions - it becomes clear that war lives on a spectrum. There’s a lot of ways to control the decision-making process of a government.

At one end, there’s lobbying and advertising/propaganda, which are often completely legal and legitimate methods of influencing government decisions. Further down the spectrum are bribes, and then assassinations. Still further along is outright guerilla warfare - which is often really just a mix of propaganda, assassination and targetted destruction of capital assets. Finally, there’s outright invasion and occupation.

I don’t think there’s a clear dividing line between illegal manipulation of government decisions (bribes, assassination) vs outright guerilla warfare. To a large extent, it should be possible in principle to achieve the same sort of goals - i.e. government takeover - with relatively little, highly targeted illegal activity. This, however, would require very precise information and models. For instance, if one knew exactly where a particular senator’s funding came from, one could in-principle physically destroy the capital assets which provide that funding (assuming insurance did not cover the loss - ideally one would want access to the details of the insurance contract in order to fake a form of destruction not covered). Or, if one knew exactly which aid wrote the text of a thousand-page bill which nobody would ever read, a bribe or threat could create a subtle loophole or a structural change which made the actual effect quite different from the symbolic meaning of the bill. Or, one could directly target the bureaucrats in charge of implementing a particular law (this is already one of the main functions of lobbying, as I understand it).

This sort of strategy relies mainly on very precise information and models; it’s exactly the sort of area where I’d expect AI tools to convey a massive advantage.


The enemy is not a single unified agent or a black box. The internal gears of decision-making in an enemy organization can be manipulated.

War is not zero sum. Attrition is (usually) costly to both parties, therefore avoiding attrition is positive-sum. It is sometimes useful to share information with the enemy so that they know which actions will result in mutual attrition (e.g. credible threats).

Destroying the enemy’s organizational capacity - e.g. killing leadership, removing competent people, making leadership unable to enforce their decisions on others - can impair their ability to wage war. However, this also impairs their ability to enforce terms of surrender on their own people. Also, if the goal is to take over the enemy government, then any organizational capacity destroyed will have to be rebuilt after the war, which is expensive and difficult; don't destroy things you intend to take.


16 comments, sorted by Highlighting new comments since Today at 6:03 AM
New Comment

You claim that "just fight the war" is a wasteful and inefficient way to defend against invasion compared to clever strategies like taking out the enemy's leadership or deploying a propaganda campaign to change the invading nation's opinion about your nation's citizens. But doesn't most invasion-defense mostly consist of just fighting the war? If assassinations are so easy and are obviously the right thing to do, shouldn't they happen more often? When was the last time assassinations were used to end a war anywhere? Without examples, the ideas in this post seem unmoored from any real assessment about what's hard vs easy.

As for cultural/propaganda solutions, these all seem far too slow. Once the enemy's tanks are rolling, the war will be decided in a matter of days or weeks -- no time to go about changing the cultural attitudes of an entire population! (And how might we expect to shift their attention to domestic issues overnight, when we have to compete with the headline that their country has just declared war??) I could see some of these defensive tactics working as a way to try and prevent invasion from ever occurring in the first place (like Taiwan's situation), or as a way to make the best of a small international incident (like how the occasional India/Pakistan flare-ups are played by both governments to score domestic political points), or that they would become relevant amid a long, drawn-out stalemate. But if you're the victim of a fast-moving surprise invasion, no clever cultural shenanigans are going to stop the hard power streaming across your borders.

Once the enemy's tanks are rolling, the war will be decided in a matter of days or weeks -- no time to go about changing the cultural attitudes of an entire population!

Contemporary war happens in two phases.  The first phase involves tanks and planes and lasts days or weeks.  The second phase involves putting boots on the ground and asserting the victor's will over the victim.  As you may imagine, the second phase involves a lot of human rights abuses.

America is great at the first phase, but is generally unwilling to admit that the second phase actually exists.  Therefore, American wars tend to involve quick kinetic victories that degenerate into insurgencies and civil wars almost immediately.  Think Iraq or Afghanistan.  

Successful second-phase war looks like what China is doing in Tibet, Xinjiang, and Hong Kong.  Constant direct subjugation of the losing population, lots of propaganda, and creating an environment where any hint of resistance is directly confronted and destroyed.  Eventually the population just gives up and stops resisting.  The process seems to be essentially complete in Tibet, well in progress in Xinjiang, and just beginning in Hong Kong.   

War is hell, but there are different circles in hell.

That's certainly the case for some wars; I'm certainly not claiming that no war is won quickly by an overwhelming force.

On the other hand, look at the US wars in Vietnam, Iraq or Afghanistan. The outcomes of these wars were determined much more by political forces (in both of the relevant countries) than by overwhelming force.

If assassinations are so easy and are obviously the right thing to do, shouldn't they happen more often?

Regarding assassinations specifically: they are not obviously the right move, in many circumstances. In many cases, one leader killed will quickly be replaced by another of similar competence. Or a group will be replaced by another group. Nonetheless, they are a useful tool, and if you think they don't happen often then I wonder why you believe that to be the case. Would you have heard about it?

Without examples, the ideas in this post seem unmoored from any real assessment about what's hard vs easy.

I basically agree with this. The lack of examples is not because they're hard to find, but because I didn't want to spend a week on the post (and make it 3-4x as long).

I’m just finishing Nixonland, which focuses on the military/political intersection of the Vietnam war from the American perspective.

In Vietnam, the decision of whether or not to get involved in the first place, and whether or not to pull out, was indeed highly political. We got entangled (after initially supporting Hochimin) because France claimed that they’d fall under Soviet influence if we didn’t help them keep their colony. Then it was the domino effect theory. And then it was (under Nixon) a hard to parse concern that losing Vietnam would harm America’s stature as a superpower.

Perversely, it was these political concerns that drove some of the intensity of the punishing American bombing. Nixon wanted to show off how violent he could be. There’s a parallel here to the decision to bomb Nagasaki and Hiroshima partly to show off our willingness to use a nuclear bomb to the Soviets.

As another example, a key victory in driving the French out was when Hochimin slaughtered an entire dug-in French army corps in the valley of Dien Bien Phu. He lost 3x the men, but it showed France just how much death they’d have to suffer to continue colonizing Vietnam. They did assassinate bureaucrats too. But a lot of those bureaucrats were Vietnamese mandarins. The French already thought they were inferior people, so why would they care? There’s nothing like wiping out a huge number of enemy troops to make them think twice about pressing on.

Violence is a form of signaling. And signaling your willingness to use, escalate, and sustain violence by doing so is part of the political theater. So I’m not really convinced that history shows alternatives to violence to achieve political ends being the norm.

For every dead enemy soldier, you create 2 bereaved parents, 1.5 bereaved siblings, many bereaved friends and a country full of terrified draftees.

On the other hand, look at the US wars in Vietnam, Iraq or Afghanistan. The outcomes of these wars were determined much more by political forces (in both of the relevant countries) than by overwhelming force.

Insurgencies aren't a good comparison for conventional wars like the Nagorno-Karabakh war.

If assassinations are so easy and are obviously the right thing to do, shouldn't they happen more often? When was the last time assassinations were used to end a war anywhere? 

The issue with assassinations is that while they can be effective for the country, a leader ordering an assassination of a leader of the other side faces the prospect of having an assassination ordered against them.

A gentlemens agreement between leaders of both sides not to assassinate the leaders of the other side is personally benefitial for the leaders involved. 

The other reason, as noted by Clausewitz, is that the enemy leader is the only person who can order its army to surrender.  If you kill them, victory gets much harder to achieve.

I think you may overestimate how much control over an enemy's internal politics you can reasonably expect.  The enemy is going to be as hardened as possible against your influence and will assuredly establish strong social norms against yielding to your influence, for values of "strong" that look like "succumbing to enemy pressure is treason, punishable by death".  Nations pull together in war.

I expect this to usually not be much of an issue, though for two very different reasons, depending on the country:

  • Many third-world countries just don't have a strong nationalistic streak. People act in their own interest or their family's interest, even if the country is at war. (Same sort of attitude that gives rise to widespread corruption and nepotism.)
  • Social pressure itself is dominated by symbolism. Succumbing to enemy pressure does not necessarily look like succumbing to enemy pressure. Furthermore, political factions will actively spin anything to make their side look good.

In the first case you cite, you've misidentified your enemy.  You're not fighting the nation, you're fighting some subset of it.  The usual response is to identify a significant subset that opposes your enemy subset and supply them weapons.  Be careful - a lot of Afghan anti-American insurgents started out as a US funded anti-Soviet insurgents.  The enemy of your enemy often stops being your friend when your first enemy has fled.

For the second case - the enemy is probably not stupid or politically naive (they're leading a country, after all).  Anyone within their borders who impedes their prosecution of a war will very likely be imprisoned and replaced (and probably executed).  You may see the occasional Oskar Schindler type who's willing to take the risk and cunning enough to carry it off, but that's pretty rare and it's far too slim a reed to support a realistic military strategy.

Whoops, this was meant as a response to the post, not ChristianKI's comment.

The overall thrust here seems like an application of Clausewitz's maxim that "war is the extension of politics by other means". However, the specific politics suggested seem very unrealistic.

  • You suggest ways to impact Azerbaijan's internal politics by targeting harm to specific groups. I see no reason to believe that Armenia had any substantial ability to deal much harm to Azerbaijan at all, so this isn't relevant. In general, it would be much harder for Armenia to advance to deal significant damage to Azerbaijan's homeland than it would be to defend.
  • Assassinations are practically universally a bad way to change a country's politics; they usually result in a direct backfire.
  • Your advice seems to lack object-level knowledge of the conflict itself; in particular, Azerbaijan is not a liberal democracy; its current leader has been in power for decades, and won ~86% of the vote in the most recent election.
  • The second Nagorno-Karabakh conflict was indeed a populist war: the two main causes were racial conflict between Azerbaijan and Armenia, and the result of the First Nagorno-Karabakh war, in which Armenia overran and occupied Nagorno-Karabakh. "We will end racism against us" is not in fact a realistic short-term plan; it would take decades to have relevance. Resolving the conflict over Nagorno-Karabakh would presumably involve returning Nagorno-Karabakh to Azerbaijan; this is not a good plan for retaining Nagorno-Karabakh! More generally, you can usually prevent a war by just giving in, and maybe Armenia should have given that they lost the war. It's also not politically realistic, given Armenia's domestic politics.
  • Your plans for improving Armenia's popularity in Azerbaijan (a) wouldn't stop the war, (b) likely wouldn't help end the war, and (c) are irrelevant to actually existing war, since they assume that Armenia is overrunning Azerbaijan, occupying their citizens, taking significant numbers of prisoners, etc.
  • It is not clear to me how Armenia would go about creating a recession or a famine in Azerbaijan?
  • Pretty much all of these plans are underspecified outcomes, not realistic plans. For example, for "we should do propaganda" you haven't specified what Armenia should have actually done for it to matter. *In practice*, actually Azerbaijan had an *overwhelming* advantage in propaganda, making use of new channels like TikTok and Youtube to quickly disseminate videos of military successes.

So how should Armenia have retained Nagorno-Karabakh?  Given that Azerbaijan is about 3 times its size, and that it has substantial oil reserves that can be used to fund military spending, Armenia would have little chance on its own. Even worse, Azerbaijan is supported by their co-ethnics in Turkey, which is vast and wealthy in comparison to both states; Armenia would not realistically have been able to disrupt this relationship.

Armenia would need a powerful patron to counter this. Three options:

  • Iran supported Armenia in the First Nagorno-Karabakh War, but developed closer ties with Azerbaijan more recently. I don't know how realistic blocking this would be, I'm not very familiar with regional politics.
  • Russia has also historically backed Armenia, but didn't intervene until late in the conflict (Russian peacekeepers are now in the region). This was likely for two reasons: first, Russia is already somewhat overstretched, with concerns in ongoing conflicts in Ukraine, Syria, and Libya. Second, Armenia's 2018 Revolution brought to power a more democratic government that leaned away from Russia and towards the West. The answer here would be for Armenia to become a more dependent Russian client state, although this may have been impossible for domestic political reasons. Armenia will likely now pursue this strategy.
  • Armenia has an extensive diaspora in the West, which it used to mobilize political support; using the diaspora and the Armenian realignment towards the West to secure military aid and security guarantees could have been useful, although in practice I don't think Armenia could have secured anything of much significance.

I generally agree with most of this, in the context of that particular conflict. It was a very smart war to invest in from the perspective of Azerbeijani leadership; Armenia really didn't have a realistic approach to defend.

Th one part I object to is "Pretty much all of these plans are underspecified outcomes, not realistic plans.". The title of the post is "Grand Strategy"; the whole point is to talk about general approaches, not specifics. Realistic plans would be the domain of strategy.

So how should Armenia have retained Nagorno-Karabakh?

Use the Iraqi playbook.  In the kinetic phase of the war, Armenia is probably hopeless.  So make only a token show of resistance.  

Before Azerbaijan takes over NK, scatter weapons caches to your co-ethnics.  Train NK locals as insurgents.  Make sure your border is permeable to insurgents; give them a place to rest, recover, and prepare.

Don't let Azerbaijan consolidate its control.  Use ambushes, snipers, and IEDs to discourage Azerbaijani troops from leaving their compounds.  When the invaders make an enemy (and they will, lots of them), give that enemy a weapon.  When the invaders make a friend, give that friend and his entire family a hideous death.  Let people know that collaborators get closed-casket funerals (and then bomb the funerals).

Provoke the invaders into heavy-handed response, then put videos of the massacres on YouTube (CNN, if you can).  Make their allies pay in lives and embarrassment.  Portray your freedom fighters as heroes standing tall against brutal oppression.

It's a horrible project.  War usually is, and insurgency is worse than most kinds of war.  But it could be done. Eventually Azerbaijan would probably leave, simply because nobody sane wants to stay in the hellhole you've created.  Victory!

This reminds me of Bruce Bueno de Mesquita's work. His TED talk on how sanctions against Iran increase the chance that Iran builds an atomic bomb by shifting internal powerdynamics is great.