When and why should we punish? Though easy to state, this question is difficult to answer. Numerous philosophers and legal thinkers have attempted to answer this question, and their answers have lead to a variety of models of punishment. The two most common models are those of utilitarianism and retributivism. In many ways these two models seem to stand in direct opposition to one another.
In this paper I will first describe the utilitarian and retributivist models of punishment in some detail. I will consider the strengths and weaknesses of both models. Finally, I will address motivations both for accepting and rejecting these models.
While I find both accounts of punishment compelling, I do not believe that either one alone is immune from serious problems. I do not, however, think this means that both models should be entirely abandoned. Instead, I think that several interesting and important models arise from attempting to reconcile the differences between utilitarianism and retributivism. In the final part of this paper I will posit several such ``synthesized'' models and explore them in some detail.
As mentioned, the two main models of punishment are based on utilitarianism and retributivism. The difference between these two can be roughly described as a difference between punishment with an eye toward the future and punishment for the sake of the past. That is, in brief, utilitarianism is a consequentialist theory and, as such, is concerned with future consequences of punishment. In contrast, retributivism sees punishment as the direct and deserved response to already committed crimes (SSA, 666).
As utilitarianism is a form of consequentialism, the utilitarian must look to the consequences or outcome of possible punishment to determine when punishment should be applied. As utilitarianism aims at maximizing utility (or happiness), it follows that punishment should be applied when it leads to an improved situation. Even though punishment decreases the happiness of the person being punished (by depriving him of rights, money, time, etc.) utilitarians claim that punishment, and/or the threat of punishment, often increases the happiness of the society as a whole (MC, 118).
According to the utilitarian, punishment can be beneficial in any or all of the following ways: general deterrence, special deterrence, rehabilitation, and incarceration (SSA, 669). General deterrence refers to the belief that the threat of punishment (as presented by punishing offenders) has a deterrent effect on other members of the society. Although this is difficult to show empirically, there does seem to be some truth to the claim that rational beings who wish to avoid punishment and perceive a threat of punishment as associated with committing crimes will have some motivation to avoid committing crimes. The argument that punishment deters generally relies on a perceived threat of punishment; for punishment to deter generally prospective offenders must believe that they are likely to be punished if they break the law. Punishment, however, may also be specially deterrent; that is, one who has been punished for a previous offense may be deterred from committing future crimes. For him, the threat of punishment is real.
Even if one does not believe that punishment deters, punishment seems to maximize happiness in other ways. It is certainly possible that one who is punished can receive moral education and be rehabilitated. Or he may be in some way asked to make reparations or restitution for his crime. Someone who has been locked up is (hopefully) unable to commit further crimes. Punishment in the form of incarceration, then, benefits society as a whole because happiness is maximized when individuals are protected from becoming future victims of a known past offender.
In summary, the utilitarian believes that people should be punished when, and only when, it maximizes happiness to do so. While happiness is the goal of utilitarians, it may be pursued in different ways. Additionally, non-utilitarian consequentialists consider other consequences than happiness. Act consequentialists focus on the utility of specific acts of punishment. In contrast, rule consequentialism is concerned with the utility of the system of principles which can be extrapolated from particular acts.
Retributivism, on the other hand, offers a model that is not about consequences at all. Instead, retributivism is about punishing offenders because they deserve punishment. Kant words this argument in terms of guilt. According to Kant and other retributivists, the guilty deserve punishment; punishment is their just desert. Punishment is proportional to guilt. People are punished because they are guilty, and for no other reason. The retributivist does not concern himself with consequences; in fact, he finds it offensive that one might let the consequence interfere with deserved retribution. Kant, in particular, argues against consequentialism because he believes that it is never right to use individuals, even criminals, as means to an end. Retributivism seems to be based on fairness - punishing people when (and as much as) it is deserved (SSA, 691).
Retributivism can actually refer to any or all of three different claims: minimal, permissive, and maximal retributivism. According to minimal retributivism, it is always wrong to punish someone who is not guilty. This, I believe, is the most mild and commonly accepted form of retributivism. Permissive retributivism claims that it is never wrong to punish someone who is guilty. Maximal retributivists take retributivism further, claiming that not only is punishment of the guilty permissible, but it is required; it is wrong not to punish the guilty (SSA, 693).
The question remains, of course, how does one determine how much punishment is deserved. It seems that the weight of the punishment must correspond in some fairly direct way to the weight of the crime. I will not attempt to determine the ``badness'' of various crimes and punishments. Nor do I intend to explain how one could know the consequences of a particular action. Such questions are beyond the scope of this paper, and I will assume both that there is some way to determine roughly how much punishment is deserved or at least to rank the weights of crimes and punishments and that we can have a reasonable idea what the outcome of particular actions will be.
I will turn now to an evaluation of retributivist and consequentialist (in particular, utilitarian) models of punishment. According to Kant, people should be punished because they are guilty (or to rephrase the retributivist argument without reference to guilt, because certain actions are deserving of punishment). The consequences of punishment are irrelevant to determining when and how much to punish. Kant believes that even if society were about to be dissolved (i.e., there were no future with which we could be concerned), the guilty must still be punished (Kant, 703).
But this strikes me as intuitively absurd. We can't undo the past, so why should punishment be so focused on the past? Empirically, it seems that a major motivation for punishment is the belief that punishment is likely to bring about some good in the future. We do not punish merely because there is a guilty party, but because we are trying to avoid or shape certain future consequences. If there is no future to prepare for, no consequences to be enjoyed, what is the good of punishing? Such punishment seems pointless, futile, and needlessly costly. In such circumstances as these, I am unconvinced that punishment should be required. And so I reject maximal retributivism (MC, 118).
Consequentialism, then, seems to offer a model more in alignment with my beliefs. I believe that we should and do look to consequences when considering punishment. I cannot accept any model which admits needless or other such ``unprofitable'' punishment. I do not believe that punishment can be justified if it does more harm than good.
However, the retributivist does make one objection to utilitarianism that leaves me worried. It is compatible with utilitarianism, says the retributivist, that innocent people may be punished (Rawls, 685). For suppose that making someone a scapegoat, using them as an example, brings about a decrease in crime. Then utilitarianism seems to sanction such undeserved punishment. As a less extreme example, utilitarianism allows for someone who is guilty of a crime to be punished more than his guilt and the weight of his crime merit (SSA, 671).
Such a model is unacceptable to the retributivist because it treats people as means to an end (Kant, 702). But it is also objectionable for reasons which are not necessarily retributivist. Utilitarianism appears to inadequately respect individual rights; it leaves individuals with no protection from a tyrannical government. If it is true that utilitarianism allows for scapegoating, I will have to abandon utilitarianism.
My focus in addressing retributivism and utilitarianism has been on two questions: (1) Are there instances when criminals should be punished more leniently than the offense alone merits, or not at all? (2) Are there instances when offenders (or innocents) should be punished more harshly than would be prescribed by the offense alone?
I believe these are extremely important questions, as they highlight the tension between utilitarian and retributivist models of punishment. I have characterized retributivism as a theory that answers both questions negatively. I have rejected retributivism on the grounds that other factors besides the weight of the offense and guilt should be considered in determining punishment. Utilitarianism seems to answer both questions affirmatively, and so I will reject it too.
But must consequentialism answer both questions affirmatively? I believe the answer is no. I believe there are many consequentialist models of punishment which offer protection against the objection raised by question (2). For example, if the goal of a model of punishment is to minimize crime, then a society which punishes innocents may not effectively deter crime. Additionally, I believe that any theory which places a high value on individual rights will not sanction the punishment of innocent people.
It seems that any utilitarian theory that incorporates the belief that the preservation of individual rights leads to happiness would also have to reject the claim that the punishment of an innocent is justified. It maximizes happiness neither to punish one undeserving individual, nor to have a society in which people can be wrongly punished. Then it may be that a society is generally happier in which only those who deserve punishment are punished (SSA 671-672).
It seems plausible, then, that the retributivist has drawn the wrong conclusion in claiming that the utilitarian would allow for excessive punishment, much less the punishing of an innocent. If this is the case, then I am satisfied by utilitarianism. That is, I am satisfied by a consequentialist model which places high value on the retributivist notion of just deserts. It seems that it may be then, that the models of retributivism and consequentialism are more similar than they first appear.
However, if the utilitarian model of punishment does not have this built-in protection of individuals against the majority and can provide examples in which such over-punishment is allowed, I must then reject utilitarianism. This does not mean that, however, that I must reject consequentialism entirely and accept pure retributivism. Nor does it mean that I must abandon punishment. What I would like is a model that allows offenders to sometimes be punished more leniently than their offense alone merits, but never permits punishing more than is deserved.
The answer, I believe, is to combine elements of both retributivism and consequentialism into a single system, a system which provides checks and balances between the two (SSA, 718). I will consider two approaches: one which is primarily retributivist and one which is primarily consequentialist.
I will first propose a system that is primarily retributivist. It is a system that uses retributivism as a rule of thumb, but it also allows for exceptions. In particular, such a system could respect the claim of the minimal retributivist that innocents can never be punished and still have room for consequentialism. In such a system, then, it is possible that people can be used for the sake of future consequences, but they cannot be used only for the sake of consequences. I suggest a model in which individuals are punished only when they do wrong, but beyond that, the kind and degree of punishment is limited by the constraint that the punishment must not provide more harm.
A very similar model can be formulated by starting with utilitarianism, or another consequentialist theory, and providing a retributivist check. I posit the following model: Punishment is never allowed unless it leads to improved consequences which in term lead to increased happiness. Additionally, punishment is never allowed if it punishes more than is deserved.
What is the difference between these models? In the first, punishment is allowed to whatever extent it doesn't worsen the situation. In the second, punishment is only allowed if it improves the situation. In the first, punishment is primarily about past actions, while in the second it is rooted in consequences. This is a subtle distinction, and the models are really quite similar. Both say that punishment must satisfy two requirements: one about the offense itself and one about the consequences of the punishment. In both models deservedness and utility are necessary for punishment, but neither alone should provide a sufficient condition for punishment.
These two models provide accounts in which punishment is allowed only when it both (1) is deserved and (2) does more good than not punishing would. That is, they incorporate principles that I believe are necessary to any just model of punishment. These combination models suggest that utilitarianism and retributivism need not be abandoned entirely to find a more satisfactory theory of punishment. A liberalist model can incorporate features and avoid pitfalls of utilitarinism and retributivism. I believe that models such as the ones I have outlined above are rooted in and superior to both pure utilitarianism and retributivism.