1 See Philips, Michael ‘Is Kant's Practical Reasons Practical?’ Journal of Value Inquiry 15 (1981), 95–108.CrossRefGoogle Scholar

2 Some might deny that restrictions on liberty for this reason are properly called punishments. This might be said by those who hold strong retributivist theories of the definition of punishment (i.e. theories according to which moral ill-desert is a necessary condition for the use of ‘punishment’). There is a considerable literature on the problem of defining ‘punishment’ (see, e.g. Primora, Igor ‘Is Retributivism Analytic?’ Philosophy 56 [1981], 203–11CrossRefGoogle Scholar; and Day, J.P. ‘Retributive Punishment,’ Mind 87 [1978), 498–516CrossRefGoogle Scholar). My own rather unexciting view is that the state punishes someone when it imposes a burden on her on the basis of a belief that she has violated a law. This allows us to speak of punishing the innocent in the way that we ordinarily do (‘My God! We've punished the wrong man!’). And at the same time it allows us to distinguish between punishments and other restrictions on liberty a state might impose (e.g. quarantines, curfews, preventative detention). Nothing that I say in this paper, however, turns on this way of speaking.

3 My approach to this question differs radically from the more standard approaches in the literature. Most such approaches understand the notion of a justified belief in relation to some model ideal reasoning or reliable belief formation (see, e.g. Bonjour, Laurance ‘Extemalist Theories of Justification,’ Midwest Studies 5 (1980), 53–73CrossRefGoogle Scholar; and Goldman, Alvin ‘What is Justified Belief?’ in George 5. Pappas, ed., Justification and Knowledge [Dordrecht: Reidel 1979]1–24Google Scholar). Hilary Komblith argues against these approaches in ‘Justified Belief and Epistemically Responsible Action.’ Philosophical Review 92 (1983), 33-48. In any case, he argues, they are inadequate to the purposes of an ethic of belief. Komblith, however, does not go nearly far enough. On his view an epistemically responsible agent is one who desires to have true beliefs, who thus desires to have his beliefs produced by processes that produce true beliefs and whose actions are guided by these desires. There is no recognition on his account that different degrees and types of validation may be appropriate to different categories of belief by different agents in different contexts. I recognize that I have not given conclusive arguments in support of my own approrach here but I believe that my examples make that approach at least plausible.

4 It is an implication of this view that we may be justified in adopting a policy that we can be certain will treat some person unfairly or unjustly. And this seems to conflict with deontic views of the inviolability of persons. It seems to me, however, that either this deontic standard is too restrictive or that this way of understanding it is mistaken. We know with moral certainty that any general policy or practice is bound to treat some persons unfairly or unjustly. If we took this to be a conclusive argument against implementing a policy or having an institution we could not conduct social life at all. Consider punishment. We can be morally certain that if we have a criminal justice system, many people will be trated unfairly or unjustly by agents of the state. Not only will police officers violate the rights of citizens, but some persons will be convicted and imprisoned for crimes they did not commit. If we took it to be a decisive objection to the institution of punishment that innocents are punished in this way, our dictum could not be ‘better ten guilty persons go free than one innocent person be convicted,’ but, rather, ‘better every guilty person go free … ’. Note also that the more we rewrite rules of evidence, codes of trial procedure and so forth to increase the likelihood of punishing the guilty the more we also increase the likelihood of punishing the innocent. Still, when the problem of crime is serious enough, we may be justified in doing this. It is worth emphasizing, moreover, that we need not defend this decision on utilitarian grounds (i.e. on the ground that it maximizes the satisfaction of preferences). We may also defend it on the grounds that we are preventing rights violations or immoral acts. Indeed, if we share some of the intuitions of the strong retributivist we could say that punishment of some innocents is a price we must pay to discharge our obligation to punish the guilty. Such arguments are of course consequential, but the consequences in question are moral, not utilitarian. The idea here is simply that we may be morally justified to do something that would otherwise be considered wrong in order to bring about some morally superior future. It seems to me clear that there are at least some important cases in which we are entitled to do this (e.g. punishment). If this is a departure from the principle of the inviolability of persons, then that principle is too restrictive.