278 Contemporary Incompatibilism: Skeptical Views
just because they have knowingly done wrong, neither do they deserve punish-
ment just because they have knowingly done wrong. The free will skeptic thus
recommends that the retributivist justification for punishment be abandoned.
By contrast, a theory that justifies criminal punishment on the ground that
punishment educates criminals morally is not jeopardized by free will skepticism
per se. However, we lack strong empirical evidence that punishing criminals
reliably results in moral education, and without such evidence it would be
immoral to punish them for this reason. It is generally immoral to harm someone
to realize some good without substantial evidence that the harm will produce the
good. Moreover, even if we had impressive evidence that punishment is effective
in morally educating criminals, non- punitive ways of achieving this result would
be morally preferable, independently of whether criminals are morally respons-
ible in the basic desert sense.
According to deterrence theories, punishing criminals is justified for the
reason that punishing deters future criminal behavior. The two most- discussed
deterrence theories, the utilitarian version and the variant that grounds the right
to punish on the right to harm in self- defense and defense of others, are not
threatened by the skeptical view per se. But they are questionable on other
grounds. The utilitarian version, which specifies that punishment is justified
when and because it maximizes utility, is subject to well- known objections. It
would counsel punishing the innocent when doing so would maximize utility; in
certain situations it would recommend punishment that is unduly severe; and it
would authorize harming people merely as means to the safety of others.
The type of deterrence theory that grounds the right to punish in the right we
have to harm and threaten to harm aggressors in order to defend ourselves and
others against immediate threats, advocated by Daniel Farrell, for example
(1985: 38–60), is also independently objectionable. A threat that one could justi-
fiably make and carry out to protect against an aggressor in a situation in which
law enforcement and criminal justice agencies have no role cannot legitimately
be carried out in a context in which the aggressor is in custody. The minimum
harm required to protect ourselves from someone who is immediately dangerous
in the absence of law enforcement is typically much more severe than the
minimum harm required for protection against a criminal in custody. If our justi-
fication is the right to harm in self- defense, what we can legitimately do to a
criminal in custody to protect ourselves against him is determined by the
minimum required to protect ourselves against him in his actual situation. If one
proposes to harm him more severely, for instance to provide credibility for a
system of threats, the right to harm in self- defense would not supply the requisite
justification, and one would again be in danger of endorsing a position subject to
the use objection.
What is the minimum harm required to protect ourselves from a violent
and dangerous criminal in custody? It seems evident that nothing more severe
would be required than isolating him from those to whom he poses a threat.
Thus it would appear that Farrell’s reasoning cannot justify punishment of crimi-
nals, exactly, supposing that punishment involves the intentional infliction of