Let’s imagine that your city hires Robocop, who (fortunately) has a publicly readable source code in this hypothetical, and is programmed to shoot anyone who commits a violent crime from this moment on.
An evil mastermind would immediately look up the source code, realize this, and put their criminal plans on hold until they’ve found a way to circumvent Robocop. They would be subjunctively deterred.
A regular hoodlum wouldn’t realize that anything was different until they started seeing other criminals riddled with bullets. They would be causally deterred from committing crimes when they saw the punishment actually being meted out.
The point Silas is making is that, if all your city’s criminals are of the first type, then Robocop will successfully deter crime even without actually shooting anybody.
That is, if there’s a bug in Robocop’s source code that in practice will cause it to never actually load its weapon, but the bug is subtle enough that even an evil mastermind won’t notice it based on an evaluation of the source code, then the evil mastermind is deterred “subjunctively” while the regular hoodlum is altogether unaffected (though the evil mastermind presumably quickly updates on the existence of non-bullet-riddled criminals).
If all the city’s criminals are evil masterminds, OTOH, they are all deterred.
I may be splitting hairs, but causal versus subjunctive deterrence aren’t mutually exclusive as described here. In fact causal deterrence is always subjunctive and subjunctive deterrence is always causal. The difference is on what you’re focusing: the historical learning experiences versus the lessons learned from those experiences. When you talk about causal deterrence, you’re talking about the learning experiences themselves—but they are learning experiences only because of the lessons learned. Meanwhile, when you talk about subjunctive deterrence, you are talking about the lessons learned—but the lessons always require a learning experience. Let’s apply it to these examples:
An evil mastermind would immediately look up the source code, realize this, and put their criminal plans on hold until they’ve found a way to circumvent Robocop. They would be subjunctively deterred.
The evil mastermind knows how to interpret source code—but he has to have at some point learned how to do this. If he’s like the typical skilled programmer, he has a long history of having run code and seeing what happens. Even if you think, “in principle he could learn language simply by reading the language specification”, he needs to have learned how to read and understand language, which again involves lots and lots of experience.
A regular hoodlum wouldn’t realize that anything was different until they started seeing other criminals riddled with bullets. They would be causally deterred from committing crimes when they saw the punishment actually being meted out.
In order for the regular hoodlum to learn from the experiences of others as he does here, he must draw an inference about the future, such as, “if I were to commit a crime then Robocop would shoot me”—which belief about the future behavior of Robocop conditional on his own behavior, if it deters him from commiting crime, is subjunctive deterrence.
I still think that the two types of deterrence ought to be distinguished, as they differ on the key point of whether a punishment has to actually happen in order for deterrence to take place.
Recall the hypothetical raised here on LW wherein the US, immediately after WW2, declares that any other country found to be pursuing nuclear weapons will be immediately nuked. Ignoring the many moral and practical difficulties of that policy, it at least is an example where leaders would strongly prefer subjunctive deterrence to suffice.
I agree that it’s an important point that the punishment doesn’t actually have to ever happen in order for the deterrent to work. Agree 100% on that. Nor do I think that I’m really saying that there isn’t a distinction—what I’m arguing is that the distinction is which side of the coin you’re looking at.
However, to split another hair, the WW2 example that you mention (I didn’t see the original mention so I may be wrong about this) depends strongly for its deterrent effect on the fact that the US actually did drop atomic bombs on Japan, which created an example for the world to look at and learn from. I don’t doubt that there would be a deterrent effect merely by demonstrating the bomb on fake houses and people, but actually dropping the bomb on populated areas probably greatly increased the seriousness with which everyone takes the atom bomb.
Let’s imagine that your city hires Robocop, who (fortunately) has a publicly readable source code in this hypothetical, and is programmed to shoot anyone who commits a violent crime from this moment on.
An evil mastermind would immediately look up the source code, realize this, and put their criminal plans on hold until they’ve found a way to circumvent Robocop. They would be subjunctively deterred.
A regular hoodlum wouldn’t realize that anything was different until they started seeing other criminals riddled with bullets. They would be causally deterred from committing crimes when they saw the punishment actually being meted out.
The point Silas is making is that, if all your city’s criminals are of the first type, then Robocop will successfully deter crime even without actually shooting anybody.
And possibly without being able to shoot anybody.
That is, if there’s a bug in Robocop’s source code that in practice will cause it to never actually load its weapon, but the bug is subtle enough that even an evil mastermind won’t notice it based on an evaluation of the source code, then the evil mastermind is deterred “subjunctively” while the regular hoodlum is altogether unaffected (though the evil mastermind presumably quickly updates on the existence of non-bullet-riddled criminals).
If all the city’s criminals are evil masterminds, OTOH, they are all deterred.
I may be splitting hairs, but causal versus subjunctive deterrence aren’t mutually exclusive as described here. In fact causal deterrence is always subjunctive and subjunctive deterrence is always causal. The difference is on what you’re focusing: the historical learning experiences versus the lessons learned from those experiences. When you talk about causal deterrence, you’re talking about the learning experiences themselves—but they are learning experiences only because of the lessons learned. Meanwhile, when you talk about subjunctive deterrence, you are talking about the lessons learned—but the lessons always require a learning experience. Let’s apply it to these examples:
The evil mastermind knows how to interpret source code—but he has to have at some point learned how to do this. If he’s like the typical skilled programmer, he has a long history of having run code and seeing what happens. Even if you think, “in principle he could learn language simply by reading the language specification”, he needs to have learned how to read and understand language, which again involves lots and lots of experience.
In order for the regular hoodlum to learn from the experiences of others as he does here, he must draw an inference about the future, such as, “if I were to commit a crime then Robocop would shoot me”—which belief about the future behavior of Robocop conditional on his own behavior, if it deters him from commiting crime, is subjunctive deterrence.
I still think that the two types of deterrence ought to be distinguished, as they differ on the key point of whether a punishment has to actually happen in order for deterrence to take place.
Recall the hypothetical raised here on LW wherein the US, immediately after WW2, declares that any other country found to be pursuing nuclear weapons will be immediately nuked. Ignoring the many moral and practical difficulties of that policy, it at least is an example where leaders would strongly prefer subjunctive deterrence to suffice.
I agree that it’s an important point that the punishment doesn’t actually have to ever happen in order for the deterrent to work. Agree 100% on that. Nor do I think that I’m really saying that there isn’t a distinction—what I’m arguing is that the distinction is which side of the coin you’re looking at.
However, to split another hair, the WW2 example that you mention (I didn’t see the original mention so I may be wrong about this) depends strongly for its deterrent effect on the fact that the US actually did drop atomic bombs on Japan, which created an example for the world to look at and learn from. I don’t doubt that there would be a deterrent effect merely by demonstrating the bomb on fake houses and people, but actually dropping the bomb on populated areas probably greatly increased the seriousness with which everyone takes the atom bomb.