The problems with both schemes for justifying punishment led Rawls to try to combine them.[1] He made an important distinction between justifying specific acts of punishment and justifying the institution or practice of punishing. He treats the  justification of specific acts of punishing on essentially retributivist grounds and justify the practice on broadly utilitarian grounds.  The retributivist "justification" of the practice of punishment turns out, as we saw, to be almost a tautology.  We should only punish wrongs sounds like a definition that presupposes the practice.  When we knowingly inflict harm on someone other than in answer to a previous wrong, we would call it something else (pointless torture, aversion therapy, parental education, human sacrifice, or sadism). 

Rawls then gives a clever utilitarian argument for the institution of punishment by exploiting this distinction between act and rule utilitarianism. Act utilitarianism asks what the utility of this specific act is and rule utilitarianism asks what is the utility of everyone’s obeying this rule. Thus he abandons any attempt at a direct utilitarian justification of each act of punishing. Instead, he asks if the "practice" or institution of punishment might be justified on "practice utilitarianism" grounds. 

Since the meaning of 'punishment' is bound up with the practice, he clarifies his point by contrasting it with a practice of hurting people on teleological (consequential or utilitarian) grounds which he calls "telishment." (Don't bother to look—you won't find this word in the dictionary.  Rawls invented it to make his point.)  Now we can think of the choice between two rival practices or institutions: punishment and telishment.  Telishment would include inflicting harm on someone prior to their doing anything wrong to deter them or others from wrongdoing.  It would also include inflicting harm on someone innocent of that crime (merely pretending to have found the real culprit) in order to deter the rest of the population.

These are the things retributivists accuse utilitarians of favoring. Rawls then argues that a utilitarian would not choose the institution of telishment over that of punishment. The consequences of telishment as an institution would be far worse than those of punishment.  People would become excessively insecure and distrustful of government.  The web of lies and deceptions about guilt would be costly to the individuals and institutions involved in the lying as well as those lied about.  Maintaining the institutional lie would require draconian restrictions on open communication from the government to the people and thus affect the cohesion of society as a whole. Thus, any argument of comparative benefits and harms clearly favors using the institution of punishment.

Once the practice is in place, of course, we make judgements on individual cases using that practice and its rules. A utilitarian should not do things that undermine just or good social practices.  Once the institution is justified, that is, the utilitarian's judgment in individual cases is just like that of the retributivist.

Note, however, that the argument solves only the first problem for the utilitarian.  It shows why a rule utilitarian would not punish the innocent. However, it doesn't deal with the other worries about absence of any empirical evidence that punishment really achieves a better society or deters criminality.  (It never really deals with Confucius' argument that punishment will lead to more selfish people.) 

Perhaps we should have neither institution. Remember the Confucian point‑‑our social choices "exercise" natural human morality  or natural human selfishness.  A system of inflicting harm on people rather than engaging in moral education may be easier for rulers to employ, but in the long run it's not the best way to have safe streets.

Rawls implicitly accepted one of the retributivists key criticisms.  It is wrong to punish someone who has committed no crime. On this, in effect, the utilitarians and retributivists can now agree. Notice, however, that it is a principle of when one may not punish given that we have the practice.  It does not show that we should have the practice. The problem for retributivists is that they need to prove that one should punish whenever one commits a crime. That is, they need to justify the practice. 

Kant, as we noted, is credited with a famous (almost absurd) illustration.  If the end of the world were just one day away and only you and I were left alive and you were guilty of murder, then I would have the duty to kill you before tomorrow rather than let you die naturally.  This sets a harder task for the utilitarian justification of the institution.  Why should we have either institution? 

However, we can use a different kind of appeal to retributivism to justify the rule of law.  We would not then decide if punishment is justified or now.  We just observe that the rule of law sets the conditions outside of which punishment is clearly wrong. Following the rule of law, it is not so definitely wrong. But it is also not definitely right, either.

Let us now go back to some of the more sophisticated retributive lines of argument we looked at before. The linking element is a Kantian theme.  Punishment is the right of the criminal--in fact should be treated as rationally "willed" by the criminal.  In punishing him, we are abiding by the rational rule or maxim that he uses in committing his crime.  Thus we take an eye for an eye, a life for a life. 

Retribution is, thus, not merely vengeance. It's difference is that it has a highly rationalized structure. For one thing, it requires that the punishment be rationally proportional to the crime, more severe punishments be for more serious crimes. One problem we saw was finding rational guides for the whole range of crimes.  It is easy enough to say a life for a life, but what about three torture murders?  Torture the criminal almost to death three times then kill him? Rape the rapist?  Steal money back from the thief?  We can make sense of some crimes being more serious than others and thus deserving longer jail terms, but it doesn't tell us what term length "fits" any particular crime non-comparatively.  A utilitarian defense of the institution might give more plausible answers to these balancing the harms of the punishment with the degree of deterrence and harm of the crime. (Still, as we noticed last time, it may still have some counter-intuitive institutional results—like executing ordinary bus-drivers who speed on the highway.)

Herbert Morris employs Rawls strategy but with a Kantian twist in place of the utilitarianism.  He also focuses on comparative justification of the institution of punishment--and is particularly concerned about the rule of law. He argues that the institution is justified because it is the right of persons considered as rational moral agents. He contrasts rule of law with what he thinks a utilitarian argument would justify, not telishment but a wide-ranging system of therapy (using psychology, drugs, hypnosis and so forth).  It might actually control crime better than a system of law and punishment. (Curiously, unlike Confucius, he does not consider the benefits of strong families and moral character education.) He argues that we have a right, nonetheless, to insist on the combined institutions of law and punishment because they treat us as rational decision makers, rather than as objects of scientific control. Law and punishment shows respect for our rationality.

Morris’s depiction of the utilitarian alternative is indeed frightening, but it is often not clear that he is finding something wrong with the use of therapy as much as of using it without the limitations of the rule of law. His examples draw their impact mainly from the absence of the procedural safeguards found in a mature scheme of the rule of law. It does not depend on anything clearly superior about punishment per se, except in the sense that doing anything to someone as a result of their commission of a crime makes that thing sort of like punishment.

Conceivably, we could employ legal and procedural safeguards in a therapy or "re-education" system as much as in a punishment. We could also design an education system of the sort that draws on Confucius’ ideas and build in the Kantian respect for moral agents that Morriss requires. Education could (implicitly would) respect our nature as reasoning beings capable of mastering and applying rules and guiding concepts.  The "re-education" could be based on the natural human tendency to conform, and could employ argument and persuasion rather than drugs and mind control.  Morris has given himself a victory by choosing a comparison with only the most horrific portrayal of what would be a utilitarian system (drugs, hypnosis and psychoanalytic therapy) rather than a more acceptable and rational system of education and moral suasion.

Still, we can appreciate one of the points he makes as applicable against even an education system with legal safeguards. The model of education is inherently a parent-child or teacher-student model.  It is appropriate for a certain stage in a person's life. However, the state is not a viable candidate for the "father" (it is not obviously wise) and adult human beings should not be continuously subject to the kind of moral domination that is characteristic of childhood training. The system would be authoritarian in a way that would offend his Kantian sensibilities.

It would be an affront to one's dignity as a mature rational agent to undertake to "train" her in our morality.  On the reverse side, I suspect advocates of punishment (like Morriss) underestimate the psychologically manipulative aspects of punishment.  It is hardly a better method of respecting and encouraging rational decision making.

Morris also appeals to another powerful line of argument for punishment, the one based on "balancing benefits and burdens."  Supposedly, a criminal gets an unfair advantage from her violation of the laws.  The rest of us forgo those selfish advantages in the interest of the whole community.  The unfair benefit obtained should be balanced by a corresponding burden (punishment) placed on the miscreant.  Our sense of justice is that a balance should be restored and punishment seems like a means to achieve that fair (as opposed to utilitarian) end.

The problem is that punishment usually fails to restore the moral status quo. The punishment typically puts the criminal back further than merely where he was when he committed the criminal act. Her reputation, standing in the community, employment possibilities, self-image etc. are damaged to the point that few become productive citizens again.  On the other side, so harming the criminal does nothing to restore what was taken from the victim.  A real policy of "balancing benefits and burdens" would place the state in the business of compensating victims of crime for whatever they lost. Punishing the perpetrator may satisfy a thirst for vengeance, but it does not restore the victim to his previous status.

The upshot is that philosophers still have not managed to justify punishment. As we saw earlier, punishment is what Confucius’ best argument was really attacking, not law itself. However, we may now start to see a line of argument that can justify the rule of law without justifying punishment.

The Kantian celebrates the fact that the agent brings the punishment on himself willingly or logically.  This is an important insight. The fact that the rules are known in advance, are clear, are interpreted consistently means that one can make a rational prediction that some action will bring the probability of punishment.  If we choose, in such a social context, with such a practice, to violate such clear guidelines, we have, in a sense, chosen at least to risk punishment. This justifies the use of the rule of law whenever punishment is threatened, even if it does not justify the practice of threatening to punish. In a sense, it makes it so the criminal has only himself to blame (does not have a right to complain about his own punishment).  That doesn't make it right to punish, but it makes it less wrong.

Then we can step back and place punishment in the context of human nature.  Humans do normally have a strong desire for vengeance and powerful humans, those with force at their disposal (rulers), tend to use that force to get their way, they are inclined to harm those who defy them.  Let us assume that these are natural human tendencies, whether or not they are justified. They are, we accept practically impossible to eradicate. Any adequate social philosophy, consequently, must find a way to deal with them and mitigate their bad effects.

Now a utilitarian justification of punishment does not contemplate an ideal situation.  It seeks for the best outcome we can get given these twin human impulses.  A rationalized system of punishment structured by the rule of law channels and controls both the human impulse to vengeance and the ruler's impulse to coercion.  We do not justify punishment, but we recognize the pervasive and destructive urge to use it.  The rule of law and retributive systems specify when it may be permissible (though never a duty) to punish.  Thus, negative retributivism (only punish the guilty) is a key to justifying the rule of law.