Consequentialism for beginners

The “other side” in the game is not Saddam (a pawn), but the Democratic opposition (and in extension, “Old Europe”). The game is called Reelection, and since GWB doesn’t foot the bill or carry the human toll of the war, he won.

“..an optimal set of rules would be so detailed and precise as to yield the optimal decision in every case.”

suggests the problem that Godel spotted.

Although I agree that Posner’s analysis is not very helpful, I disagree w/ your criticism here. The fact that the relevant political actors will not be following bayesian theory does not prevent us from using it to consider the wisdom of starting a particular war. Further, the fact that political actors are, in fact, irrational makes your second point less relevant.

I think that one of the functions (perhaps even _the_ function) of rule-based morality is that it gives us guidance when we can’t do the cost-benefit analysis, which is most of the time.

I confess to being a beginner, so the headline indicates you are writing for me.

If the starting point is that we evaluate an action based on its consequences, whether on an act or rule basis, that excludes taking action knowing the consequences are deleterious, just because you choose to act in accord with principle.

For instance, the peace activists hypothetically flocking to Iraq to act as human shields for police stations are acting out of principle; even though they personally will suffer costs in the action up to and including death, their principles offer a rationale of potential good if others adopted their ethic, or are merely deterred from murder by their presence.

Similarly each of us with a neighbor suffering domestic violence has the choice; knowing WE can act but will get no thanks – just the risk of murder, the possibility of assault and the certainty of abuse by either or both parties. This consequentialism is the root of the ‘not my business’ response to domestic violence.

I hope you’ll elaborate on the second point about war being a negative sum game. Intuitively, this seems wrong. I would think it easy to construct an example, from history or otherwise, of a war with positive gains in sum. Off the cuff, imagine a war were the attacked lay down their arms without a shot fired. The victors march in, institute democracy, grow the economy and are adored by the conquered. Wouldn’t this ‘war’ be positive sum?

Also, your arguments on this point seem to just have the effect of increasing the costs as discussed in Posner’s post. You’ve discovered some ‘hidden costs’ in the Posnerian calculation. That’s fine. Realize those costs and then redo the calculation.

Walt Pohl,

You said, what I think, better than I could. Thanks.

Our deontological rules should be based on utilitarian outcomes over long periods of time, and should be adhered to, unless, at a minimum, the utilitarian case is very strong.

But will that still work in our rapidly changing world?

Now, you say it better.

Will, that’s silly. Of course we can construct an imaginary example of a ‘win-win’ war; it’s just that the preconditions very rarely obtain in reality. The neocons are rightly considered fools for kidding themselves that Iraq was an exception.

If the past teaches us anything, it is that war tends towards ‘lose-lose’ with occasional ‘win-lose’. The likelihood of ‘lose-lose’ has to enter our prior calculations. And as others point out calculations geared to ‘win-lose’ ignore the welfare losses of the losers.

Since war is a negative sum game, rational decision makers do not fight wars

Brilliant. Even Desert Storm or Afghanistan (which wer as “positive” as any wars could have been) were undoubtedly lose-lose.

Precision strikes and in & out guerilla actions (SEAL / Special Forces ) may be zero-sum or better, but they do not like constititue “war”.

Bigmacattack: Thanks!

Perhaps I’ve missed something important.

But Posner certainly needs to differentiate between “rational”, on the one had, and “good,” “moral,” or “just” on the other.

This distinction is clear within classic utilitarian thinking, but has seemingly dropped by the wayside in modern faux-thinkers like Posner.

Act A (or rule A) might be “rational” for Person X in the sense that it maximizes whatever they are maximizing. But this does not mean that A is “rational” or “good” from the point of view of society: X plus those who are not X. In this case, those who are not X include those who are attacked preemptively.

Of course, we can define those we attack as non-persons who don’t count. Or was this assumption made?

“Since war is a negative sum game, rational decision makers do not fight wars”

I don’t know what kind of game “War” is, but if you’re thinking of a repeated prisoner’s dilemma with war/war the Nash equilibrium of the one-shot subgame this has nothing to do with negative sum, and war/war is a Nash equilibium strategy for rational actors: Under the expectation that the other side wages war, war is the rationally optimal response for both sides.

You’re probably thinking of a game with endowments. If both sides start off with endowments E1, E2 and playing peace/peace increases the endowment each round while war/war (and peace/war for the dovish side aka sucker) reduces it, starting a war is not a rational strategy for the side with lower endowment (assuming a payoff between sucker and penalty in eternity after the war is lost), but might be for the party with higher endowment (although it might be dominated by extortion). Of course all of this only holds if the cost of war is known a priori. If the comparative cost of war, and in extension time until victory/defeat, has to be established by Bayesian updating war might very well be a rational strategy for the lowly side, depending on the priors. It requires overconfidence though, i.e. war must be more costly or lengthy (or the spoils of victory less profitable) than anticipated at least for one side, if not both. This might be what you were trying to get at.

OGMB, since wars by definition involve the diversion of productive resources into producing expensive pieces of capital equipment which are delivered to people who don’t want them and then explode, they can’t be positive or zero-sum events. Any war is Pareto dominated by a contract under which both parties agree to the outcome that the war would have produced without fighting.

I’d add a third bullet to John’s points in favour of rule-utilitarianism above:

* There is no basis for assuming that the probabilities one would need in order to be a fully rational Bayesian decision maker actually exist, let alone that there is a sensible basis for constructing an assumed prior and likelihood function.

As I said in an earlier thread, I’m sceptical about giving consequentialism the role that John wants it to play in thinking about war and peace.

We ought to distinguish, though, between two roles for consequentialism: (1) we use consequentialism as a decision-making tool, either directly in the evaluation of acts or via rules; and (2) we use consequentialism as the explanation for the system of prohibitions, permissions and so-on that we employ.

I don’t want to rule out the thought that (2) might provide the best explanation for our norms (though I doubt it). What I do think important, though, is that whether or not we think that the underlying explanation/justification is provided by consequentialism or by something like a casuistic reflection on the justification for individual self-defence (etc.), we have good reason to settle on a system of norms to govern war and peace, norms that are common knowledge among the participants, can predictably guide their actions (and reactions) and so on …

Two reasons for being skeptical about giving consequentialism the role that John wants it to play in the actual decision-making process, (1) we’re unlikely to achieve a consensus on what the relevant consequences ought to be and (2), for Chou-en-lai like reasons, we can’t calculate them.

To illustrate: WW2, and specifically the decision of the UK and France to declare a state of war in September 1939. With respect to the calculability issue, we don’t know the consequences even now! (And given the way consequences ramify, we never will). On a narrow and crude utilitarian calculation (a Benthamite one if you like) my money would be on the decision to go to war being _unjustified_ . The victims of Nazi terror are so vastly outnumbered by the other dead and maimed, that there is, at least, a real possibility that a narrow u calculus comes out against war.

I think there’s something obviously wrong with this conclusion, though. What’s wrong is that the narrow u calculus is _too_ narrow: we ought to count other things, such as great injustice, in a reasonable calculation of consequences. But I doubt that we’ll achieve consensus on what are the relevant values to build into a wider consequentialist scheme and, depending on how expansionist/inclusivist we are, our justification is going to start looking a lot less … consequentialist.

I do think, though, that a broad consequentialist approach can usefully _supplement_ and inform deliberation within the framework of a set of public norms. Take, for example, the case of the Iraq War. Like the liberal hawks, I believe that the norms, rules and permissions governing just war ought to include at least a permission to intervene to remove a regime that grossly and systematically violates human rights. But unlike them, I think that — especially given the unpredicability of war — we need to be pretty sure that the outcome of our action will be better in broad consequentialist terms (i.e with values like justice built into the calculus) than inaction, or action at some other time. So whilst I think that it would sometimes be justifiable to wage war against a Saddam-like regime, I think this actual war is too uncertain in its (broad) consequences to be justified.

[Short addendum: As I noted in a comment on Kieran’s Posner thread, any conseqentialist calculation ought to be with respect to the costs and benefits from an impartial pov, and _not_ the c&bs to the nation initiating the decision.]

It’s the radical nature of war that presents difficulty here, because this ‘consequentialist’ approach seems perfectly reasonable when appied to a more banal situation, like a tax policy or something.

It’s just that human behavior in extreme situations is pretty much unpredictable. If you ask me: say, a guy is threatening you with a knife, what do you do – run away? attack him? break down in tears? Well, the answer is: I have no idea what I would do. C/b analysis doesn’t work here.

So, if we accept that the actors in this particular situation probably won’t act rationally, this throws the whole premise out of the window; nothing much to analyze here.

Dsquared, I agree with much of what you say, especially the ceteris paribus presumption, but that doesn’t settle the problem that bothers all non-pacifists, namely *which* wars are to be supported. War isn’t the only thing that costs huge numbers of lives (see Rwanda and Cambodia); this is why I think it’s important that John’s point about human bias is going to apply to military inaction as well as to action. And if your key points are the numbers dead and the waste of resources involved in delivering ordnance to people who don’t want it, then your strictures should apply as much to resistance to either internal or external aggression as to the aggression itself. Do you think we have the same kind of reason to avoid defensive wars, or revolution against tyranny, as you’re arguing we have to avoid war simpliciter?

JQ: ‘Since war is a negative sum game, rational decision makers do not fight wars’.

That being the case, are all states which spend money on armed forces led by irrational decision makers?

If not, why not?

If their rationality stems from the fact that they are spending money on armed forces for deterrent purposes with no intention of actually using them in war, doesn’t that rational intention destroy the deterrent effect?

Btw, I thorougly recommend Alan Bennet’s play ‘The History Boys’ for those who can make it to the National Theatre in London. The saddest character is named ‘Posner’, although I have no idea whether this is deliberate.

“Brilliant. Even Desert Storm or Afghanistan (which wer as “positive” as any wars could have been) were undoubtedly lose-lose.”

Afghanistan is lose-lose only if you cling tightly to the idea that the invasion had nothing to do with 9-11, and if you believe that letting the organization which commited that war crime run the Taliban would not have led to further atrocities against Americans.

Are you assuming that 9-11 was a rational act to be analyzed under this structure? Was the 9-11 attack a win for Al Qaeda?

I also have to agree with eve that you can’t ignore the costs of inaction in your calculus–a cost almost always ignored by those who argue along pacifist lines.

Before we go completely off-track here: neither Afghanistan nor Gulf War I were preventative wars.

Dsquared: since wars by definition involve the diversion of productive resources into producing expensive pieces of capital equipment which are delivered to people who don’t want them and then explode, they can’t be positive or zero-sum events. Any war is Pareto dominated by a contract under which both parties agree to the outcome that the war would have produced without fighting.

Pareto dominatedness doesn’t equal irrationality, and negative sum in accounting terms doesn’t mean it’s a negative sum game. You’re also assuming enforcible contracts and a “good” peacetime allocation of resources, neither of which are givens in international relations. But I get the point about war being mutual destruction of resources, therefore a “negative” event. But this doesn’t imply irrationality. Either side, in order to prevent war, would be willing to pay R-P (the delta between peacetime and wartime payoffs) to the other side each round, but on the other hand also faces the prospect getting up to R-P from the other side, so there’s a lot of money on the table to fight over. A decision for war can be rational if discounted expected postwar payoff minus discounted war payoff exceeds discounted payoffs without a war. Since by assumption P < R this requires that both sides’ expected probabilities of winning sum up to more than one–iow, overconfidence. But overconfidence is not necessarily irrational because economic rationality only requires best responses to the expected actions of the other side.

A comment on war being a negative sum game: this only holds for the political actors involved, which is important for game theory reasons but should not extend to being a general criticism of war as some here have implied. The Afghanistan war is an excellent example of a war that has a positive sum potential when we include the wellfare of the people of Afghanistan instead of just the interests of the Taliban ruling them. In the case where a regime is not representing the interests of those being ruled, there is always the possibility of a positive sum when that rulership is changed, be it through internal revolution or war against an external foe. If such a war were never desirable, then revolution against an unjust government would also have to be always undesirable.

On the “can wars be win-win?” question: First of all we have recognize the fundamental shift that happened between WW1 and WW2: the treaty of Versailles to the Marshall Plan. Versailles was driven by the millenia-old rule of war, To the Victor Go the Spoils. The Marshall Plan on the other hand, rather than transferring assets from loser to winner used Allied money to invest in the defeated Axis countries. The payoff for the victors therefore moved from immediate (by seizing assets) to mediate (by creating allies and estblishing trade partnerships).

Iraq, despite the claims of Halliburton profiteering, was designed as a win-win war. The people who think win-win is impossible implictly assume that the post-war economic order is the same as the pre-war order, only disrupted by loss of human lives and industrial assets. This isn’t true, and the losing side changing its political system is the historical rule rather than the exception. Falkland was a “win-win” war in that it was fairly restricted and it directly lead to the toppling of the Argentine junta, therefore giving the Argentinians at least the possibility to create a better social order than they had before. Afghanistan could have been a win-win if the GWB administration hadn’t prematurely redirected its focus to Iraq.

The problem with thinking of Iraq as a win-win war is that it assumes that 1. dictatorship is the worst kind of social order, and 2. democracies just snap into place once you remove the dictator. At least domestically most dictators are preferable (and come into power as remedies) to absence of social order (aka the Libertarian Utopia), and the key rationale for keeping them in power is to keep the country from slipping into chaos. Even if there is a prospect, there are no guarantees that the post-war social order is better than the pre-war order, so fighting a (preventive) war on a win-win premise should make damn sure that the post-war order (quickly) improves on the status quo. The total blindness of the Rumsfeld-Wolfowitz-Feith fascia to this simple doctrine is the main failing of the current administration in re: Iraq.

ogmb – “The problem with thinking of Iraq as a win-win war…”
The problem with thinking of the “war” in Iraq as a war waged for the officially stated purpose – to unseat a dictator – is that in even the most well-intentioned and rational mind it creates an argument against something that doesn’t exist. It diverts oppositional energy into a cul-de-sac, from which it can’t emerge except in defeat.
Inasmuch as your clear paragraphs emphasize the illogic and danger of what’s been done, and refute the given reasons for it, they help greatly. But it would help even more if you attacked the heart of the matter with that same clarity.
The soldiers on the ground, some of them, may believe we’re there for democracy, the Christmas shoppers at the mall may believe it, but no one in Washington believes that, and very few people in the rest of the world do either.
Too often an otherwise sound thinker pounces on these bogus explanations, like a dog capturing a sock, ignoring the hand that holds it for the sake of the game.

Since war is a negative sum game, rational decision makers do not fight wars

However, a rational decision maker of post-draftable age may encourage others to fight a war after investing heavily in military hardware manufacturing, with positive-sum result for himself.

Both logically and chronologically NE (mutual best response) as a solution concept is an upshot of the economic definition of rationality, not vice versa. Although I agree I should’ve used the more general term consequence rather than response since part of game theory is about shaping the opponent’s responses, not just responding to his/her actions.

I’d like to see an argument for that, because I’m pretty sure it isn’t a logical consequence. Several of the reasons for thinking that neither NE, nor any other equilibrium concept, can be derived directly from principles of individual rationality are set out in Bob Stalnaker’s recent work, e.g.

* Robert Stalnaker, “Knowledge, Belief and Counterfactual Reasoning in Games”, Economics and Philosophy. 12(1): 133-163

* Robert Stalnaker, “Extensive and strategic forms: Games and models for games”, Research in Economics (1999) 53, 293–319

* Robert Stalnaker, “Belief Revision in Games: Forward and Backward Induction”, Mathematical Social Sciences, vol. 36

The one-line version is that to get from Bayesian rationality to NE (or anything like it) you need assumptions much stronger than common knowledge of mutual rationality, and those assumptions are, on the whole, wildly implausible even as idealisations. Since in war we have quite a bit _less_ than common knowledge of mutual rationality, the relevance of NE is far from obvious.