The root of the "dilemma" just seems to be that we find it unlikely that a trustworthy model would really say such a thing. Human brains have trouble granting false premises, even hypothetical ones.
So yes, I would cut all punishments to zero if a sufficiently trustworthy model told me it would be for the best. And if a sufficiently trustworthy model told me I could fly by flapping my arms and quacking, I'd be outside pretending to be a duck.
The root of the "dilemma" just seems to be that we find it unlikely that a trustworthy model would really say such a thing. Human brains have trouble granting false premises, even hypothetical ones.
It's not a problem with human brains; that's how it should be. If completely eliminating all punishments for crime lowered the crime rate, there would have to be something very deeply weird going on, and it would be foolish to try to make policy judgments without knowing what that weird thing was.
Thinking about it for sixty seconds, the only ways I came up with to make "zero punishment yields smaller crime rate than small punishment" true, would be if the punishments were actual mislabeled rewards; or receiving punishment turns people into criminals somehow (such as by putting them in the company of other criminals). Both of these would be resolved by taking "severity of punishment" off a numeric scale and looking at it qualitatively.
You're asking, then, whether I would trade off the satisfaction of having other tribe-members help me punish the one who hurt me, against the probability of being hurt? In the specific circumstances you have set up, I absolutely would. The reason is that, if there's no punishment for murder, there is presumably likewise no punishment for retaliation in kind. So I don't need my tribe to enact any punishment, I can do it myself, gaining even more satisfaction. This really seems like the best of both worlds: You get to take personal vengeance on anyone who wrongs you, with no dilution through 'official' coalitions; while the probability of actually being wronged goes down.
Yes of course, you are free to do it yourself, but it is assumed on the large scale that that even including retaliations (which are crimes), crime rates would go down. And in a society with no punishments would it be rational to do that? (Given that the friends or relatives of that guy could come after you for coming after him for coming after you and so on..?
It's hard to reason about your hypothetical, because it seems to directly contradict actual experience. But it seems fairly straightforward to reason about what I, personally, would do: I want the chance of crimes against me minimised, so I accept the no-laws state as the best way of getting that. But if I nonetheless am among the unlucky ones, then I want revenge, so I get revenge. By hypothesis, the chance that the other guy's friends will successfully punish me has got to be small, because we've established that crime is very low in this hypothetical universe.
You know this doesn't have to be relegated to "thought experiment". It's pretty much the story of a large part of human history:
There were no state-enforced laws, per se, but you didn't kill people, because then their family/clan/tribe would kill you. Of course then your family would do the same, and well..you can see how this might be a never-ending downward spiral. A more sophisticated way would be to pay a weregild . (I killed your brother. Sorry. Here's some goats.)
If you are interested in seeing how this actually worked out, early Icelandic society is a pretty good microcosm. If you want to make learning it fun, read a saga. (Here's Njal's Saga )
second learning about early Icelandic society, it has some fascinating legal and economic structures.
People want both (as separate values) safety from crime and retribution for crime. In this thought experiment, zero punishment is stipulated to maximize safety from crime, but it simultaneously minimizes retribution. If marginal [value of] retribution at zero punishment is lower, then zero punishment is optimal, otherwise a non-zero amount of punishment is better overall. Alternatively, if the only available forms of punishment are such that the aspect of (undesirable) hurting of criminals outweighs the aspect of (desirable) retribution for the crime, then zero punishment is automatically optimal in this thought experiment.
crime rates anti-correlated with harshness of punishments imposed on criminals.
This sentence appears to have a typo; if crime anticorrelates with punishment then more punishment equals less crime.
So it helps people in general by decreasing crime, and helps criminals by not hurting them? Sounds like win-win to me!
Punishment hurts people. It's terminally bad. I'm only okay with it because it keeps other people from being hurt, so its effect is net good.
I am suspicious of thought experiments involving especially implausible situations, but I would say that if circumstances were as you described, the demobilization would be morally required.
I don't know whether the value I get from revenge is greater than the value I get from decreased danger, but I am almost sure that would depend on exact numbers and there would be a threshold T of probability for each crime. So, for example, let P = present law and H = hypothetical law which doesn't punish murder and p(M) = murder rate; I would prefer H to P iff p(M | H) is lower than T(murder). I am not certain what T(murder) actually is, but it may also depend on p(M | P). To say at least something, I would prefer P to H if p(M | P) - p(M | H) < 10^(-8) , and I would prefer H to P if p(M | H) < 10^(-8).
If you want to survey about values attributed to retribution, you could probably do it more directly without assuming an inplausible scenario. Just: "Would you choose to do action X, if the consequences were (1) reduction of crime rate by amount Y, and simultaneously (2) no punishments for crimes?" There is no need to assume that (2) is the cause of (1); doing so primes people to concentrate on the implausibility instead of the core of your question.
A reformulated remotely plausible version of the thought experiment: I am a rich experimental psychologist / experimental ethicist / sociopath. You are dictator of a city state with 1 million inhabitants. I offer you to install a super-powerful automatic security system if and only if you agree to abolish all punishments. The automatic system covers your entire city by surveillance cameras and detects all crimes. Moreover, there are robots which can prevent 99% of crimes (private retributions included) from happening, and experiments show that the increase of attempted crimes in the absence of punishment is only by factor of 10. This means that the overall number of committed crimes decreases 10 times. Would you accept my offer?
Least convenient possible world, of course: you have absolute power, so no real chance to be overthrown because of making unpopular decisions; the security system is safe and destroys all records after they are of no use for the crime-preventing robots, so no disutility associated with loss of privacy; etc.
One additional assumption: there is no significant target shift.
Armed robbery for taking week's profit from ten businessmen may be less of a strike to all of them in sum than armed robbery taking months' salary from a minimal-wage-working single mother. It would count for a greater crime rate, though...
If all this were true and if the model held around zero, then doing the thing that minimizes suffering would be good.
That doesn't mean we should not check if other forms of punishment would drop crime rate further.
Of course, the model would have to include a lot of stability checks. With 1¢ fine for every crime, people would use a lot of interesting weapons for self-defense in broad sense (including defending good neighbour's property from theft by force just in case). People wanting retribution would do the retribution themselves... Some other interesting effects would occur...
If that worked out to lower aggregate crime rate for each victim group - it would be nice...
Here is a thought experiment for you. There will be some bold assumptions here, and they may be regarded unrealistic. I am aware of that and the purpose of this query is not to propose some truths about society in general, but to isolate certain characteristics of preferences regarding the societal institutions of law enforcement and punishment.
Assume that there existed a highly trustworthy model that showed beyond reasonable doubt that crime rates anti-correlated with harshness of punishments imposed on criminals. So basically, if policies changed towards shorter sentences, lower fines and lighter penalties, the number of criminal acts decreased (in every category).
Further assume that this was empirically tested and each time penalties went down, fewer and fewer crimes was committed. But the dependence was not linear so if we would get rid of punishments all together - there would still be murders, rapes, robberies etc. But, the crime rates would be minimized in that case. To summarize: We knew that crime rates would be at minimum if there was no consequences at all.
With no penalties, somebody could simply kill or rape your mother, sister or child and move in next door and live a nice and happy life in front of your very eyes, without society doing anything about it! Bare in mind now that this is the situation where the probability of your mother, sister or child being abused, robbed or killed is minimized!
Would it be reasonable to go trough with this demobilization that would spare lots of innocent people all the pain of getting robbed and abused, given that those criminals still out there can do anything they want and go free?