Robert Nozick’s arguments in Anarchy, State, and Utopia rely heavily on the notion of rights. But he distinguishes two ways that a moral theory can incorporate the idea of rights. One way is to view the nonviolation of rights as a goal to be sought: we should try to achieve a world which “minimiz[es] the total (weighted) amount of violations of rights” . This would have uncomfortable implications:
A mob rampaging through a part of town killing and burning will violate the rights of those living there. Therefore, someone might try to justify his punishing another he knows to be innocent of a crime that enraged a mob, on the grounds that punishing this innocent person would help to avoid even greater violations of rights… 
By contrast, Nozick tends to regard nonviolation of rights as a moral side constraint, meaning: we should not perform actions which violate rights, period, under any circumstances. That’s certainly appealing in this example. But it’s easy to dream up counterexamples where his view’s consequences seem absurd, such as this one from David Friedman:
A madman is about to open fire on a crowd; if he does so numerous innocent people will die. The only way to prevent him is to shoot him with a rifle that is within reach of several members of the crowd. The rifle is on the private property of its legitimate owner. He is a well known misanthrope who has publicly stated on numerous occasions that he is opposed to letting anyone use his rifle without his permission, even if it would save hundreds of lives. 
According to Nozick, “[s]ide constraints upon action reflect the underlying Kantian principle that individuals are ends and not merely means” ; he stresses the separateness of persons: “There are only individual people, different individual people, with their own individual lives. Using one of these people for the benefit of others, uses him and benefits the others. Nothing more.” 
Should the notion of the separateness of persons really lead us to believe in moral side-constraints? In the mob example, suppose I punish the scapegoat. It may be that my action hurts him and benefits the others in the town, “nothing more.” Now suppose instead that a meteorite strikes the scapegoat’s home, and, seeing the harm this does to him, the mob is placated. It’s equally true that this event hurts him and benefits the others in the town, “nothing more.”
Is it good that the meteorite falls? I could understand saying that it is good, on the grounds that (by hypothesis) this leads to fewer bad things happening overall. I could also understand saying that no evaluation of the event is possible, on the grounds that the harms and benefits to the townsfolk and the scapegoat affect separate people and therefore cannot be weighed against one another. But I don’t see any reason to think it would be positively preferable for the meteorite not to fall.
If we return to the case where I’m personally punishing the scapegoat, the consequences are the same, and all the people affected are the same (except me). The only difference is that the decision to placate the mob is decided by a human rather than left to chance. So the side-constraint against punishing the scapegoat is apparently not motivated by focusing on him, on his intrinsic value, on the impossibility of anything compensating him for his sacrifice, etc. Rather, it’s motivated by a focus on the person doing the punishing and on his act of carrying out the punishment. This seems backwards and narcissistic, as though the really important fact about the situation were not “an innocent person is being harmed” but rather “I’m sullying my hands by punishing an innocent person.”
Nozick claims that “no moral balancing act can take place among us; there is no moral outweighing of one of our lives by others so as to lead to a greater overall social good. There is no justified sacrifice of some of us for others.”  If we accept this, then any explanation of moral obligations will probably have the ‘narcissistic’ tone I complained about above. We would not be able to objectively evaluate “person A gets what he wants and person B’s rights are violated” as being a worse state of affairs than “person A does not get what he wants and person B’s rights are not violated,” so any reason for A to respect B’s rights at the cost of foregoing his own desires would have to be rooted in his own nature. 
But the notion that a “moral balancing act” is never possible is totally contrary to common sense. If, for example, you were forced to choose between your friend Dirk losing a toe and your friend Roxy losing an arm, surely the former is better overall, even though Roxy keeping her arm does not compensate Dirk for his loss.
So I don’t think Nozick’s observations present a very compelling case for the existence of inviolable moral side-constraints. Of course, our intuitions about the mob example (among other dilemmas for consequentialism) still have be to justified somehow or else explained away, but I’m done for now.
 p. 28
 pp. 28-29
 pp. 30-31
 p. 33
 Consider some of Nozick’s own speculations about what could justify moral side-constraints, which he does not expand upon in this book, but do not seem promising to me:
Suppose, for example, that one could show that if a person acted in certain ways his life would be meaningless. Would this be a hypothetical or a categorical imperative? Would one need to answer the further question: “But why shouldn’t my life be meaningless?” Or, suppose that acting in a certain way toward others was itself a way of granting that one’s own life (and those very actions) was meaningless. Mightnt’t this, resembling a pragmatic contradiction, lead at least to a [qualified in a way explained earlier in the chapter] conclusion of side constraints in behavior to all other human beings? [pp. 50-51]