A mathematician wakes up to find his house on fire. He frantically looks around before seeing the fire extinguisher on the far wall of the room. "Aha!" he says, "a solution exists!" and goes back to sleep.

-- Popular math students' joke

There has been much discussion of coulds, woulds, and shoulds recently. Agents imagine different counterfactual states of their own minds or actions, then select the most desirable. Something similar seems to happen during political discussions, but the multiplicity of agents involved muddles it a little.

I recently read a letter to the editor in my local paper. The city was launching a public education campaign against binge drinking, and this letter writer thought that all the billboards and lectures and what-not were a waste of time. She said that instead of a flashy and expensive public awareness campaign, the real solution was for binge drinkers to take responsibility for their own actions and learn that there were ways to have fun that didn't involve alcohol.

This struck me as a misguided line of thinking. Consider this analogy: pretend that the city government was, instead, increasing the number of police to prevent terrorist attacks. And that the writer was arguing that no, we shouldn't get the police involved: the real solution was for terrorists to stop being so violent and attacking people. This would be a weird and completely useless response.

Attempts to solve political problems are counterfactuals in the same way attempts to solve other problems are. In Newcomb's Problem, I modify the "my decision" node and watch what happens to the "money in the box" and "money I get" nodes. When I say "Increasing local police would prevent terrorist attacks," I am modifying the "local police" node, and positing that this would have a certain inhibitory effect on the connected "terrorist attacks" node.

The hypothetical second letter-writer's argument, then, is that if we counterfactually modified the "terrorists' attitude" node, then the "terrorist attacks" node would change. This is correct but useless.

But it's harder to see exactly why it's useless. Consider the original argument "We should raise the number of police and train them in counter-terrorism techniques." In this case, I would be counterfactually modifying the attitudes of (for example) the chief of police. But I'm not the chief of police, any more than I'm Osama bin Laden. If I'm going to let myself modify the chief of police's attitude just because it would be convenient, I might as well let myself turn Osama into a pretty decent guy. Yet "the chief of police should train more policemen" sounds like a potential solution, whereas "terrorists should be nicer" doesn't.

Here's one possible resolution to the problem: it's much more likely I could convince the chief of police to train more policemen than that I could convince terrorists to be nonviolent. Since the police chief shares my goal of stopping terrorists, all I need to do is tell them why training more police would accomplish this goal, and the problem is solved. So the reason why "train more police" counts as a solution is that, as soon as the statement and the evidence supporting the statement reaches the right person, the problem will be solved.

In this case, changing the "chief of police" node is really a proxy for changing the "my actions" node. The solution could be rephrased as "You or I could go to the chief of police and sugget they add more policemen, thus preventing further terrorist attacks." Counterfactually changing your own actions is entirely kosher. This also throws into greater relief the problems with "You or I could find and approach all terrorists and convince them to be nicer," or "You or I could go to every single binge drinker in the city and convince them to be more responsible."¹

Actually, when phrased like that, the binge drinking example doesn't sound so bad. Add a comprehensive plan for doing it, enough funding to reach them all, and some idea of how you're going to phrase the "be more responsible" point, and it sounds like, well, a grassroots public awareness campaign. Which is kind of ironic, seeing as the letter started out as an argument against a public awareness campaign, and maybe a sign that I'm taking the Principle of Charity too far here.

...in more realistic situations

It's more complex when there are only small probabilities of your own actions having any effect, but the principle stays the same. For example, I recently heard a doctor say that a single-payer system would best solve the US' health care woes, but since that was politically infeasible he was backing Obama's plan. This one doctor's support will have minimal effect on the chances of Obama's plan passing, but it will have even less of an effect on the chances of single-payer passing. If the expected utilities multiply out in such a way as to make supporting Obama more likely to gain more utility than supporting single-payer, the doctor is justified in his strategy of support for Obama's plan.

One more example from real history I learned recently. Suppose you are a Communist, and your fellow Reds are proposing ways to create a socialist paradise. One says that you must incite the workers to violent revolution. Another says you must petition the current government to support labor reform laws. A third says you must petition the current government to oppose labor reform laws.

Before you expel the third communist from the Party, let them make their argument. They say that the Party doesn't have enough resources to incite violent revolution, and the workers don't want to revolt anyway. Counterfactually modifying the "workers' actions" node to a revolutionary state is a waste of time, because there's no link between any modification of your own actions and that node reaching the state you want. Likewise, modifying "government policy" is useless, because the Communists don't have any clout in the government, so even if you found a wonderful value for that node that would make all workers happy forever, you couldn't change it.

Instead, she says, oppose labor reform laws. These are already unpopular, and even a small party like the Communists would probably have enough power to get them shot down. When there's no labor reform, workers will get angrier and angrier, until they gradually revolt and overthrow the system, getting you what you wanted in the first place.

There were communists in the early 1900s who actually tried this third approach. It didn't work, but I admire their thought processes. They ignored solutions that would never happen, and found an action they thought they could enact, that they thought they would raise the chance of revolution significantly. Compare this kind of cunning to the vapidity of the letter-writer who says "Binge drinkers should become more responsible."

...as an unrealized ideal

I like this way of viewing the problem, because it explains why a certain class of argument feels wrong: arguments that go "The solution to binge drinking is more personal responsibility" or "The solution to poverty is for the poor to work harder," or so on. But do people actually think this way?

The most glaring reason to believe they don't is that most people who "solve" societal problems have no interest in actually enacting their solution. The attitude is something like "Hey, if the federal government passed a single-payer health plan, then all our health troubles would be over!" and then don't bother to write a letter to their representative about it or even convince their next door neighbor.

For reasons that have been discussed ad nauseum on Overcoming Bias, politics is very much a signalling game. In particular, it seems to be a game in which you counterfactually propose different states of the "government policy" node and explain why these would have the best effects, and whoever can give the best explanation gets rewarded with higher status. Sometimes you're also allowed to edit the policies of large private organizations, or of influential individuals. In this case, the problem with the original letter writer wasn't just that she had no plan to enact her solution, but that she was breaking the rule which said that you're only allowed to play with relatively unified, powerful organizations, and not things like "the set of all binge drinkers".

In a way, this isn't so bad. When enough people play this game, their opinions get out to the voters, consumers, politicians, and business leaders, and eventually do change government and private policy.

The point is that if your goal is to actually personally affect things in a direct, immediate way, you can't just apply the rules of this game without thinking beforehand. Communists, when discussing politics for "fun", would never say "I think the government should oppose labor reforms," but that might be the winning move for them when they're actually trying to increase utility. Likewise, libertarians spend a lot of time discussing different ways the government could implement libertarian policy, but when they actually have to take action, the best choice might be seasteads or charter cities or something else that doesn't involve policy at all.

And if you are content to just play the game, at least keep it interesting. No fair counterfactually editing things like "terrorists' behavior" or "poor people's work ethic" or "how responsible binge drinkers are".

Footnotes

1: Or, here's an alternate interpretation of "Binge drinkers should be more responsible": it's not worth trying to prevent binge drinking, because binge drinkers could prevent it themselves if they were more responsible, so it's their own fault. This is not illogical, but applying the argument to a case like "Drunk drivers should be more responsible" would be. There, even if we have no sympathy for drunk drivers, we still need to prevent drunk driving because many of the victims are innocents. The other issue is that people process the two statements "The solution to binge drinking is for binge drinkers to be more responsible" and "We don't need to solve binge drinking; it's the drinkers' own fault and we need not care" differently; the first sounds wise and reasonable, the second callous. For both these reasons, I don't think this interpretation is entirely what the original letter writer, or other people who use this sort of argument, are thinking of.