GREENE, COHEN, AND WHAT A NEUROSCIENCE CHALLENGE IS NOT

The article that is perhaps most frequently taken to have presented a neuroscience challenge to criminal justice is a 2004 article by Joshua Greene and Jonathan Cohen.Footnote ² Since it does not, it is worth setting the record straight, and, felicitously, doing so will be instructive in showing what a neuroscience challenge must actually look like. Hence, it is to this article that we shall initially turn.

Greene and Cohen's argument consists of a series of both empirical and theoretical claims. Among the former, they claim that, first, contemporary criminal justice is largely retributivist, which is to say that it is structured around ‘the intuitive idea that we legitimately punish to give people what they deserve based on their past actions’.Footnote ³ Call this empirical premise EP1.

Second, systems of criminal justice inevitably adjust to the prevailing moral beliefs of the societies in which they exist: ‘The legitimacy of the law itself depends on its adequately reflecting the moral intuitions and commitments of society. If neuroscience can change those intuitions, then neuroscience can change the law.’Footnote ⁴ (EP2)

Third, folk psychology is fundamentally, even if not always reflectively, libertarian: ‘What people really want to know is if the accused, as opposed to something else, is responsible for the crime, where that “something else” could be the accused's brain, genes or environment.’Footnote ⁵ (EP3)

Fourth, neuroscience shows human decision-making to be mechanical: ‘It is one thing to deny that human decision-making is purely mechanical when your opponent offers only a general, philosophical argument. It is quite another to hold your ground when your opponent can make detailed predictions about how these mechanical processes work, complete with images of the brain structures involved and equations that describe their function.’Footnote ⁶ (EP4)

Fifth, dissemination of the knowledge produced by neuroscience through society is capable of affecting folk psychology: ‘as more and more scientific facts come in, providing increasingly vivid illustrations of what the human mind is really like, more and more people will develop moral intuitions that are at odds with our current social practices’.Footnote ⁷ (EP5)

In addition to these empirical premises, Greene and Cohen also make two substantial theoretical claims about criminal justice ethics and free will respectively. The first is that retributivism as a theory imposes certain conditions: ‘The retributivist approach . . . is plausibly regarded as requiring free will’ because ‘[r]etributivists want to know whether the defendant truly deserves to be punished’.Footnote ⁸ Call this theoretical premise TP1.

The second is that the libertarian account of desert-generating free will is incompatible with the form of neurological mechanics demonstrated by neuroscience, since: ‘if determinism (or determinism plus quantum mechanics) is true, then no actions are truly free because forces beyond our control are always sufficient to determine behaviour’.Footnote ⁹ (TP2)

Beyond these seven premises, Greene and Cohen make a number of additional points, including a brief defence of consequentialist criminal justice ethics as an alternative to retributivism, a psychological explanation of why folk-psychology is libertarian, and a critique of free will compatibilism. These we can set aside for present purposes, as they are not premises in the central argument. That argument, rather, can be summarized as follows:

1. (1) The current criminal justice system is retributivist, i.e. justifies punishment by and is justified only if offenders deserve to be punished. (EP1)

2. (2) The retributivist desert-justification of punishment requires that offenders possess free will. (TP1)

3. (3) The justification of the current criminal justice system requires that offenders possess free will (from 1 and 2).

4. (4) Current folk psychology subscribes to libertarian free will. (EP3)

5. (5) Libertarian free will is incompatible with human action being the effect of a mechanical neurological system. (TP2)

6. (6) Current folk psychology is incompatible with human action being the effect of a mechanical neurological system (from 4 and 5).

7. (7) Neuroscience shows human action to be the effect of a mechanical neurological system. (EP4)

8. (8) Neuroscience shows that current folk-psychology is mistaken (from 6 and 7).

9. (9) Familiarity with neuroscientific results is capable of changing folk-psychology. (EP5)

10. (10) Dissemination of neuroscientific knowledge will make the public revise current folk-psychological beliefs (from 8 and 9).

11. (11) The dissemination of neuroscientific knowledge will lead the public to revise a belief (libertarian free will) that the justification of the current criminal justice system requires (from 3 and 10).

12. (12) The legitimacy of a criminal justice system depends on its fitting with society's moral beliefs. (EP2)

This allows Greene and Cohen to conclude that:

This is a good argument: the conclusion is interesting, the argument as I have reconstructed it is valid, and the premises are plausible. This is not to say, of course, that it is impregnable. The empirical claim with which I would be most likely to quibble is that folk-psychology is as amenable to revision by disseminated familiarity with neuroscience (EP5) as Greene and Cohen assume.Footnote ¹¹ One can also question whether folk-psychology is genuinely libertarian (EP3) – the developing research on this topic is at best complex and divided.Footnote ¹² And both the claims that the existing criminal justice system is fundamentally retributivist (EP1) and that neuroscience has shown human decision-making to be fundamentally mechanical (EP4) will encounter critics too.Footnote ¹³

But these are minor points, and there are more serious problems for the argument conceived as a neuroscience challenge to criminal justice. The second-largest problem the argument faces is its reliance on the two theoretical claims, both of which are underdeveloped in the article, overlook important features of the concepts and theories they concern, and are in any case controversial claims in two debates so complex that it would be unreasonable to expect Greene and Cohen to make substantial advances in support of them as a sideshow in a short and otherwise occupied article. We will consider these troubles and their implications in more detail shortly.

First, however, it is worth underscoring the most important reason why the argument cannot serve as a neuroscience challenge to criminal justice, which concerns not the soundness of the argument but its character. Consider what the argument does and does not say. It says that developments in neuroscience and dissemination of these findings will lead to a shift in public perceptions and undermine support for the current system of criminal justice, in a way that will eventually lead to its replacement by a different system. It does not say, for example, that we ought to replace the current system of criminal justice – it says that we (probably) will. It does not say that retributivist punishment is unjust – it says that the public will increasingly come to believe that it is. In fact, the article takes pains to avoid making any explicit normative claims. Its conclusion is not a moral prescriptive for action, but a sociological prediction. Retributivists might well respond with a dry ‘so what?’ to the promise of shifting beliefs in the general populace. What matters to criminal justice ethics, after all, is whether the balance of reasons supports one theory or the other, not how neuroscientific insights will affect public sentiments.Footnote ¹⁴

Of course, this failure to moralize need not be taken by Greene and Cohen as a failing of their argument, since presumably the argument was not supposed to do so. And indeed the argument is still concerned with a form of challenge to the criminal justice system, it just is not the type of normative argument we typically have in mind as a neuroscience challenge to criminal justice. It is a challenge, rather, in the sense that e.g. populist politics and public support of ever more draconic punishment over the last decades is a challenge to criminal justice.

But there is a second argument, woven into the tapestry of the first, which better fits the type of challenge we have in mind. Consider the following partial argument:

1. (1) The current criminal justice system is retributivist, i.e. justifies punishment by and is justified only if offenders deserve to be punished. (EP1)

2. (2) The retributivist desert-justification of punishment requires that offenders possess free will. (TP1)

3. (3) The current criminal justice system is justified only if offenders possess free will (from 1 and 2).

4. (4) Free will is incompatible with human action being the effect of a mechanical neurological system.Footnote ¹⁵ (TP2*)

5. (5) Neuroscience shows human action to be the effect of a mechanical neurological system. (EP4)

These, of course, are simply some of the premises of Greene and Cohen's argument with the remainder removed. Now let us add the obvious inference:

1. (6) Neuroscience shows that human agents do not have free will (from 4 and 5).

And from these we can derive a different conclusion:

1. (C) Neuroscience shows that the retributivist desert-justification of the current criminal justice system cannot be true.

As I have argued, the above is not the argument that Greene and Cohen actually make. However, the above argument is easy to construct using (some of) the premises in their actual argument, which explains, perhaps, why many have taken Greene and Cohen to be presenting a neuroscience challenge to criminal justice.

THE BASIC STRUCTURE OF THE CHALLENGE

A neuroscience challenge to criminal justice must fundamentally take the following shape:

1. (1) Criminal justice is justified only if condition P obtains

2. (2) Neuroscience shows that condition P does not obtainFootnote ¹⁶

3. (C) Neuroscience shows that criminal justice is unjustified.

Now, obviously, there must be more to the argument than this, if only for the reason that neuroscience is concerned with the workings of the human brain, while criminal justice is the enterprise of holding offenders accountable and imposing sanctions upon them for their wrongdoing. Any version of the challenge will need supplementary premises to bridge the gap between the two domains, and we can take our inspiration for these from the version developed on the basis of Greene and Cohen's argument above.

Consider first the question of why criminal justice might stand in need of justification, in the way that the first premise above holds. The most obvious answer is that criminal justice involves punishment, and punishment involves causing harm. In general, the idea that causing harm is bad, a pro tanto (at least) moral wrong-making feature, is so uncontroversial as to be hardly worth consideration. As abolitionists have long noted, legal punishment, as the institutionalized, routine infliction of severe suffering, therefore raises a serious challenge of justification, the aptly named ‘problem of punishment’.Footnote ¹⁷

This suggestion, of course, is open to the challenge of whether criminal justice need really involve punishment.Footnote ¹⁸ One response would be to argue for particular conceptions of criminal justice and punishment. Pursuing this response, however, would bog us down in complex and somewhat tedious conceptual analysis, so I will adopt a short-cut.Footnote ¹⁹ Let us say that a system of criminal justice in anything like its present shape will involve knowingly inflicting harm upon offenders in response to their wrongdoing.Footnote ²⁰ Most such systems will presumably inflict harm through imposing punishment, but since it is the knowing infliction of harm (as a response to wrongdoing) which raises the challenge of justification, we do not need to insist on punishment being involved, and can avoid the need to argue or stipulate a particular definition.

Consider next the first premise of the implicit Greene and Cohen challenge, which holds that the current system justifies punishment by claiming that offenders deserve to be punished. Let us set aside the empirical question of whether this is an accurate description of existing systems of criminal justice. A more interesting question is whether it best captures the kinds of justification that the neuroscience challenge potentially affects. I believe it does not.

The most basic distinction in criminal justice ethics is between consequentialist and non-consequentialist theories. The former justify inflicting harm on offenders by the benefits created, and in so doing arguably (and notoriously) have the implication that it can in at least some circumstances be justified for a criminal justice system to knowingly harm innocents.Footnote ²¹ This implication alone is considered by many sufficiently morally repugnant to reject consequentialist justifications in favour of theories that cannot justify harming innocents.

The retributivism that Greene and Cohen target is one attempt to do so. In a broad sense retributivism is any theory of criminal justice that takes considerations of a person's desert to be essential to justifying the knowing infliction of harm.Footnote ²² However, we need to recognize a further distinction between theories that take desert to influence reasons to harm offenders and those that take desert only to influence reasons against causing harm. We should distinguish, that is, between:

and

Paradigmatic retributivism combines the two, but among the two it is the negative thesis that allows retributivism to avoid justifying harm to the innocent, and it is this claim which constitutes (a version of) the necessary premise in the neuroscience challenge.Footnote ²⁵

We can broaden the premise, though. Consider that desert requires moral responsibility, i.e. what Fred Feldman has dubbed:

If we adopt this thesis it allows us to include retributivist theories under a version of the challenge with a much wider scope, one that extends to any theory of criminal justice that affirms the criminal responsibility precondition-thesis:

This condition is, as I previously noted, assumed by virtually every non-consequentialist theory, since it, in combination with the plausible general principle that an action or practice is unjustified if there is an undefeated reason against it, is what guarantees that it cannot be justified for the criminal justice system to knowingly harm innocents.Footnote ²⁸ It also shows us where and why theories of criminal justice connect to a neuroscience challenge. It is moral responsibility that is a necessary condition, and it is accordingly the possibility of moral responsibility that neuroscience must disprove.

Let me sum up the challenge as we have developed it so far:

1. (1) Criminal justice in anything like its present shape involves knowingly inflicting harm upon offenders in response to their wrongdoing.

2. (2) There is at least one reason against a system of criminal justice knowingly harming a person, which is ordinarily defeated only by reasons grounded in that person's moral responsibility for doing wrong.

3. (3) So there is at least one reason against a system of criminal justice in anything like its present shape, which is ordinarily defeated only by reasons grounded in that person's moral responsibility for doing wrong (from 1 and 2).

4. (4) An action or practice is unjustified if there is an undefeated reason against it.

5. (5) So criminal justice in anything like its present shape is justified only if the persons whom the system knowingly harms are morally responsible for doing wrong (from 3 and 4).

This concludes our analysis of the first half of the challenge. To pursue the second half we will begin by taking a look at Stephen Morse's argument that that there can be no challenge.

MORSE, JUSTICE AND JUSTIFICATIONS

Stephen Morse has presented an argument in a series of articles that neuroscience does not and indeed cannot pose a challenge to criminal justice.Footnote ²⁹ Importantly, Morse does so while granting a number of the premises that the challenge relies on. Thus, Morse accepts the problem of punishment, the need for a non-consequentialist justification, and the desert-responsibility thesis:

Morse's overall argument against the challenge is then the following: there are two potential neuroscience challenges to the justification of criminal justice – the challenge to the causal efficacy of mental states, and the determinist challenge to agential free will and responsibility – but neither succeeds, and therefore neuroscience does not threaten criminal justice.Footnote ³¹ Let us begin with the challenge to the causal efficacy of mental states. Here Morse argues that:

1. (1) The justification of criminal punishment requires that the law's folk-psychological conceptions of personhood, agency, and responsibility are at least moderately accurate, particularly as regards the causal efficacy of agents’ mental states, but

2. (2) The neuroscience challenge to the causal efficacy of mental states is unpersuasive,

3. (C1) so criminal justice is not threatened by the causal efficacy challenge (from 1 and 2).

Next, with regard to the determinist challenge, Morse argues that:

1. (3) The justification of criminal punishment requires that agents can act freely and be responsible,

2. (4) agents can act freely and be responsible if either the libertarian theory of the conditions of free will and non-determinism are true, or compatibilism and determinism are true,

3. (5) compatibilism and determinism are true (and libertarianism and non-determinism false),

4. (6) so agents can act freely and be responsible (from 4 and 5),

5. (C2) so criminal justice is not threatened by the determinist challenge (from 3 and 6).

Finally, this allows him to complete his overall argument:

1. (7) The challenge to the causal efficacy of mental states and the determinist challenge to agential free will and responsibility are the only challenges to criminal justice raised by neuroscience,

2. (8) criminal justice is threatened by neither the challenge to the causal efficacy of mental states nor the determinist challenge (C1 and C2),

3. (C) so criminal justice is not threatened by neuroscience (from 7 and 8).

There are many parts of the argument that seem persuasive, but also parts that appear problematic. Consider the first challenge, to the efficacy of conscious intentions. Here Morse distinguishes between two fundamentally different accounts of action, according to one of which actions are materially caused by the prior physical state of the world, with, paraphrasing Laplace's apocryphal quip, no need for the hypothesis of mental states. The other, ‘folk-psychological view’, ascribes to the reasons of an agent a causal role in her actions, in that the actions result (at least in part) from the intentions formed on the basis of agents’ beliefs and desires.

Between these two accounts, Morse claims, law is firmly wedded to the latter: ‘The legal view is… that people are creatures who are capable of acting from and consistently with their reasons for action and who are generally capable of minimal rationality according to mostly conventional, socially constructed standards of rationality.’Footnote ³² Why is it important that law must prefer this account? Morse gives two reasons. The first is that: ‘People are meant to and can only use these rules as potential reasons for action as they deliberate about what they should do…. If people were not capable of understanding and then using legal rules as premises in deliberation, law would be powerless to affect human behavior.’Footnote ³³ This is a peculiar argument. Even if true, all it would show is that law could not affect human behaviour in the way that Morse believes is one of law's defining features. It would not, as Morse admits, show that ‘law’ could not affect human behaviour. Worse, there is no obvious reason why we should prefer one way of affecting human behaviour to the other, and even if there was, it would be wishful thinking to believe that since it would be better if both laws and persons functioned in this way, then that must be the way they are. Morse also gestures at a second reason, which is that: ‘bodily movements in the absence of consciousness are not actions’.Footnote ³⁴ This is arguably either too broad or too narrow; standard action theory requires not consciousness but intentionality under some form, which is at once more demanding than consciousness conceived as simply the presence of some form of mental state, and less demanding than consciousness conceived as focused attention and awareness.Footnote ³⁵ But even on a charitable reading it returns us to the question of why this in turn is important. The answer (presumably; Morse does not follow the argument this far) is that it would be morally wrong to treat persons the way the criminal justice system treats offenders if their assumed wrongdoing were not in fact their actions. This follows directly from the criminal responsibility precondition-thesis if we assume, plausibly, that persons are morally responsible, if for anything, only for their actions.Footnote ³⁶ If this is the proper understanding of the way that this first challenge works, then moral responsibility turns out to be at the heart of it again.

However, the state of philosophy of mind is such, Morse claims, that we do not and will not in the foreseeable future have persuasive reason to abandon ‘our view of ourselves as creatures with causally efficacious mental states’.Footnote ³⁷ Morse thus denies the psychological claims that intentions are causally inefficacious, including those based on the much-discussed Libet studies.Footnote ³⁸

Likewise, Morse recognizes the potential threat of determinism:

Ultimately, Morse accepts determinism, and consequently denies libertarianism. The choice, he claims, is between hard determinism's rejection of free will and responsibility, and compatibilism.

Hard determinism, however, is unacceptable:

Morse's solution is to thus adopt compatibilism. Compatibilism, however, is not a unified theory, but rather a family of fairly heterogeneous theories whose only necessarily shared feature is accepting that free will and moral responsibility are compatible with determinism; unlike libertarians therefore, although compatibilists typically provide an account of the necessary conditions of free will, they are not committed to affirming that ordinary agents can meet these conditions. Furthermore, filling out the account of how the compatibilist thesis can be true can be done in very different ways. The next questions are therefore which version of compatibilism Morse endorses, and whether that version can serve the purpose required of it as a premise in his argument. Importantly, the compatibilist premise in Morse's argument is not conditional: he claims not merely that if compatibilism is true then agents have free will and moral responsibility (that claim is part of premise 4), but that some specific version of compatibilism actually is true, and therefore agents have free will and moral responsibility. Given the complexity of the issue and the contentious state of the philosophical literature, arguing this claim would be no small feat.Footnote ⁴¹

Morse is clear that the account of responsibility and agency he has in mind is a reasons-responsive form of compatibilism:

The idea here is that responsibility (and, perhaps, the relevant sense of free will) requires agent-control of actions, and that agents exercise the requisite control when they act for reasons, that is, when agents respond to reasons, and their response affects (to a sufficient degree) the actions they subsequently perform.Footnote ⁴³ This is an influential type of compatibilism, with prominent defenders in the contemporary literature.Footnote ⁴⁴

His account is also capacitarian, in the sense that it is an agent's having the ability to respond to reasons that makes the agent responsible, rather than her exercising this ability:

Ultimately, Morse claims that what is implicitly his preferred version of compatibilism can give three answers to the challenge of determinism:

Can this account of compatibilism support Morse's argument against the possibility of a neuroscience challenge? I think not. Note first that what Morse appears to have in mind may not be a form of compatibilism proper at all. His claim that his suggested form of compatibilism ‘need not conform to any ultimate metaphysical facts about genuine or “ultimate” responsibility’ suggests as much, and the claim is repeated and elaborated elsewhere:

Possibly Morse does not mean to suggest that this form of compatibilism justifies the practices of the criminal justice system – after all, all he claims above is that it is ‘most consistent with the legal view of responsibility’ and ‘normatively desirable’, and it is not clear that meeting these criteria entails the ability to justify legal practices of holding agents responsible. Taken literally, however, what Morse seems to be endorsing is the type of responsibility theory advocated by J. J. C. Smart.Footnote ⁴⁸ The idea here is that no matter what the metaphysics of agential responsibility may turn out to be, we have reasons to adopt, support and institutionalize certain practices of holding agents responsible, e.g., in Smart's version, because those practices promote welfare. Possibly Morse intends that we should adopt this understanding not merely as our theory of when to hold persons morally responsible, but also as our theory of when persons are morally responsible. Among theorists who admit that this flies in the face of our folk conception of moral responsibility, which requires much stronger metaphysical underpinnings, this approach is dubbed revisionism.Footnote ⁴⁹

The obvious problem for a deontological revisionist compatibilist is that the approach has abandoned the criminal responsibility precondition-thesis. It no longer makes the metaphysical claims about which persons are and are not morally responsible upon which the thesis relies, and as such cannot distinguish between innocents and offenders in the traditional sense, but only between those we ought to hold responsible, that is, between those we ought to treat as innocent and those we ought to treat as offenders. Not only is this perfectly acceptable to consequentialists, as noted above its most prominent historical defenders, such as J. J. C. Smart, have proposed it as the consequentialist theory of responsibility. Of course, this does not mean that Morse is required to adopt a consequentialist version of revisionist compatibilism, but it does mean that as regards the moral constraint against knowingly harming the (genuinely) innocent, he has placed himself in the same boat as consequentialists. This both raises the spectre of the traditional deontological challenges to consequentialist criminal justice and means that the response is not properly speaking a response to the challenge, since it avoids rather than shows a way to satisfy the criminal responsibility precondition premise on which the challenge rests.

Suppose we charitably read Morse's comments above as offering not a replacement but a supplemental argument for adopting compatibilism. The trouble then becomes that the classical challenges to compatibilism, avoiding which seems to have at least partly motivated Morse's shift of focus above, re-emerge. On this reading, what Morse needs is an argument that not only compatibilism but his preferred form of compatibilism is the best account of the metaphysics of human agency, and that this account in turn is not only consistent with but justifies the practices of the criminal justice system in anything like the shape they currently take.

Morse does offer sketches of such an argument:

There are several things to say about this. First, while the claim that compatibilism is the dominant position among philosophers of responsibility is true, this can only qualify as an argument for compatibilism if no better evidence is available.Footnote ⁵¹ But obviously it is, since we can assess the same evidence on which these philosophers base their beliefs. In that light, it would be an ad populum-fallacy to claim that compatibilism must be true because many persons believe it. Second, it is not clear why incompatibilism could not be proven right. This would have to be either because the standard of proof Morse has in mind is so demanding as to preclude proving any position right, including Morse's preferred compatibilism, or because there was some special difficulty faced only by the proponent of incompatibilism. I cannot think what the latter would be, and Morse gives no indication of it, so he must have in mind the former. But this cannot give us reason to adopt any position, that is, cannot constitute the argument for compatibilism Morse needs. All it says is that there is some standard of proof we cannot meet, so we must choose our beliefs by some other, less demanding standard, which brings us back to the demand for an argument for Morse's compatibilism. Similar difficulty afflicts his observation that the dispute cannot be resolved by science, and is ‘unlikely to be resolved by logic’. If this is just a general observation then it is beside the point, but if it is meant to constitute an argument for compatibilism, it fails, since even if it was true (which one might reasonably contest), it could not be an argument for any particular position. Finally, it is not entirely clear what Morse takes the implications of compatibilism's alleged superiority in explaining our ‘responsibility practices’ to be, but if he intends it to constitute an argument in favour of compatibilism he would need both to support the claim (why would e.g. libertarianism not be at least an equally good account?) and to clarify how the argument can avoid being a fallacy of appeal to consequences. The latter challenge arises because the convenience of some proposition P being true, say the fact that if true it would provide a justificatory explanation of some favoured practice of ours, does not constitute a reason to think that P is true. It is not easy to see how Morse could do either.

In conclusion, Morse's argument helpfully illustrates both the type of argument proponents of the challenge need to offer, and several possible ways of responding. It also shows, by virtue of the difficulties it encounters, the complexities involved in both setting out and refuting the challenge. We can generalize the lessons drawn from this analysis of Morse's argument, and in the next section I shall do so.

THE SKETCHY PREMISES AND THE BURDEN OF PROOF

Recall that I suggested the second part of the general structure of a neuroscience challenge is a series of premises that jointly entail that neuroscience shows that a condition necessary for the justification of criminal justice does not obtain. And I have suggested that the condition at stake is moral responsibility, since this is required by theories that subscribe to the criminal responsibility precondition-thesis. As Morse's examples show, neuroscience does not, and arguably given its domain of study cannot, challenge moral responsibility directly. It challenges moral responsibility, if at all, by undermining one of the conditions necessary for it.

This in turn raises the issue upon which Morse's second refutation stranded, for what are the conditions of moral responsibility? To complete the challenge one needs both to argue that there is some condition which is required for moral responsibility and to show that neuroscience has demonstrated that this condition does not obtain. That is, one needs to complete an argument that:

1. (1) A person is morally responsible iff she meets conditions (P₁, P₂,. . . P_n).

2. (2) Neuroscience shows that at least one of the conditions (P₁, P₂, . . . P_n) does not obtain.

3. (C) So neuroscience shows that persons are not morally responsible (from 1 and 2).

To establish a case against moral responsibility, neuroscience must show that there is some candidate condition for moral responsibility that can be experimentally demonstrated to be unattainable by ordinary agents and which is definitely required for moral responsibility, regardless of other disagreements.Footnote ⁵² The structure of the neuroscience challenge to criminal justice, then, is the following:

1. (1) Criminal justice in anything like its present shape involves knowingly inflicting harm upon offenders in response to their wrongdoing.

2. (2) There is at least one reason against a system of criminal justice knowingly harming a person, which is ordinarily defeated only by reasons grounded in that person's moral responsibility for doing wrong.

3. (3) So there is at least one reason against a system of criminal justice knowingly harming an offender, which is ordinarily defeated only by reasons grounded in that person's moral responsibility for doing wrong (from 1 and 2).

4. (4) An action or practice is unjustified if there is an undefeated, not overridden, not outweighed reason against it.

5. (5) So criminal justice in anything like its present shape is justified only if the persons whom the system knowingly harms are morally responsible for doing wrong (from 3 and 4).

6. (6) A person is morally responsible iff she meets conditions (P₁, P₂, . . . P_n).

7. (7) Neuroscience shows that at least one of the conditions (P₁, P₂, . . . P_n) does not obtain.

8. (8) So neuroscience shows that persons are not morally responsible (from 6 and 7).

9. (C) So neuroscience shows that criminal justice, in anything like its present shape, is unjustified.

Can proponents of the challenge fill out the required details and support these premises? The obvious place to start might be with premise 7, since this is where neuroscience finally enters the picture. In fact, the past decade has seen a resurgence of debate on this point, with complex arguments both in favour and against the view that results from neuroscience undermine free will and moral responsibility, including perhaps most prominently successors to the famous Libet-studies.Footnote ⁵³

This, however, would be a mistake. We need to identify which neuroscientific results could potentially undermine moral responsibility (premise 6) before we can investigate whether neuroscience has in fact produced one of these results, and it is doubtful that we can do so. We saw this in passing in the discussion of Morse's theory of moral responsibility above, but the problem here will be readily apparent to anyone familiar with the debate. Contemporary theories of free will and moral responsibility defend a wide range of diverse conditions, many of which can be modified and combined to provide a bewildering array of possible complete theories.Footnote ⁵⁴ Importantly, the problem is not just that there are rival theories, but that the number, diversity and strength of the rival theories imply that a neuroscientific undermining of any one condition is unlikely to undermine moral responsibility according to a majority of plausible theories and prominent thinkers in the field.

Given the lack of theoretical consensus about which conditions are necessary and jointly sufficient for moral responsibility, it appears that there is therefore presently no clear way to complete a neuroscience challenge to criminal justice, i.e. to support premise 8.Footnote ⁵⁵ On the other hand, it is clear that neuroscience is at least potentially capable of challenging criminal justice. The abundance of insights into human cognition and agency it has generated has arguably already fundamentally changed some of our ideas about what it means to be human, and is likely to continue to do so. Consider as illustrations the way ego depletion-theory has emphasized the vulnerability of agents to continued exertions of self-control and thereby the influence of the environment,Footnote ⁵⁶ and the way dual-process models of decision-making demonstrate our reliance on unreflective and systematically biased cognitive heuristics, and call into question our ability generally to make decisions rationally and reflectively.Footnote ⁵⁷

Let me emphasize two points about the situation of neuroscience vis-à-vis moral responsibility as I here present it. First, the claim here is merely that no matter what the right answer to the question of the jointly necessary and sufficient conditions for moral responsibility turns out to be, it appears inevitable that one or more of the conditions will be or rely on an empirically falsifiable claim about human psychology. This means that it is exceedingly likely that neuroscience will be in a position to potentially undermine criminal justice. Whether it will actually be able to do so will depend on a number of factors, including whether the relevant condition(s) in fact obtain, since neuroscience cannot prove false a true claim about human psychology, and the difficulty of experimentally investigating the condition compared to the developmental state of neuroscience at the time. The second point is that a candidate-theory's fit with ideas about human agency widely assumed in folk-psychology cannot be taken to inoculate it against the potential threat from neuroscience, since neuroscience is capable of contradicting folk-psychology, and has an established track-record of doing so. The challenge may therefore not be currently completeable, but pace Morse it is at least open in the sense that we cannot assume that the correct theory of moral responsibility does not contain a condition which neuroscience will demonstrate to be false.

The above conclusion might be taken to be trivially parsimonious. Perhaps there is an argument based on neuroscience against criminal justice, but given theoretical disagreements about the conditions of moral responsibility the details have yet to be filled out, so we cannot presently say with certainty. This in turn might suggest that we can all get back to the business of traditional criminal justice ethics, and leave neuroscience and its implications for a rainy day.

In fact, I believe this would badly misinterpret the situation. Criminal justice, after all, involves knowingly causing harm. We tend to think of this particular harm as unproblematic, but this, it seems to me, is because we unthinkingly assume that there is sufficient justification for it. Few people, if pressed, would presumably defend criminal justice by denying the problem of punishment and arguing that causing foreseeable and avoidable harm is in general morally unproblematic. And this in turn suggests that the onus is on the proponent of criminal justice (in anything like its present shape).Footnote ⁵⁸

We can summarize this concern as the precautionary argument against criminal justice:

1. (1) If φ’ing will cause harm, then we should not φ unless we have (sufficient) reason for thinking that φ’ing is justified in spite of causing this harm (precautionary principle for causing harm),

2. (2) criminal justice (in anything like its current shape) will knowingly cause some persons harm (problem of punishment),

3. (3) so unless we have (sufficient) reason to think that criminal justice is justified in spite of knowingly causing harm, then we should not employ criminal justice (in anything like its current shape) (from 1 and 2); and

4. (4) criminal justice is justified in spite of knowingly causing harm only if offenders are morally responsible for offending (criminal responsibility precondition-thesis),

5. (5) we do not have (sufficient) reason to think that offenders are morally responsible in the (e.g. desert-generating) sense that would justify knowingly causing harm (the open challenge from neuroscience to criminal justice),

6. (6) so we do not have (sufficient) reason to think that criminal justice is justified in spite of knowingly causing harm (from 4 and 5).

7. (C) Thus, we should not employ criminal justice (in anything like its current shape) (from 3 and 6).

I have already discussed some of the premises in the above (2, 4 and 5), and others are sub-conclusions that follow logically (3 and 6), but I should say something about premise 1 and a little more about the role of premise 5 in this argument.

The first premise (precautionary principle for causing harm) is one that would properly require a more elaborate explication and defence than I can provide here, but let me at least sketch two reasons why I believe it to be plausible. First, the principle is strongly intuitive. It seems right that an agent who invokes a defeasible justification for performing an act that she knows will cause harm must have good reasons to think that the justification actually obtains. Cases involving such situations will tend to support the intuition. Consider for example the rigorous scrutiny to which we subject new medical procedures or pharmaceutical products in order to determine the type and severity of side-effects, but also, and crucially, to ascertain that they produce a benefit which justifies their use.Footnote ⁵⁹ A second obvious example even closer to the issue at stake is the high standard of proof liberal systems of criminal justice impose on the determination of guilt e.g. by adopting the presumption of innocence and the BARD-threshold, in Blackstone's famous formulation because ‘the law holds it better that ten guilty persons escape, than that one innocent party suffer’.Footnote ⁶⁰

Second, I intend the principle to be a type of mid-level principle, neither fundamental nor specific. The benefit of such principles is that they can be acceptable to and justified in a specific situation by a wide range of moral theories. Consider for comparison Peter Singer's celebrated principle that ‘if it is in our power to prevent something very bad from happening, without thereby sacrificing anything else morally significant, we ought, morally, to do it’.Footnote ⁶¹ Obviously, as a utilitarian Singer cannot believe this to be a fundamental principle, nor is it a principle specific to the situation (poverty relief through individual charity) he discusses. Rather, it is a broad but not fundamental principle designed to be acceptable to and justified in a number of situations (including the one Singer applies it to) by a wide range of moral theories. Similarly, the precautionary principle, it seems to me, is one that all prominent theories of the justification of criminal justice should accept as applicable to and justified in that context.

Next, and arguably most importantly, it is worth saying a little more about the threshold of sufficiency at stake in premises 1 and 5, specifically whether it is plausible, as premise 5 holds, that the standard of sufficiency in an intuitively acceptable version of premise 1 is not met in the light of the open challenge from neuroscience. Let me make two observations, a confrontational and a concessionary argument, in that order.

Note first that it might be tempting to say that arguments for the justification of criminal justice that require moral responsibility are similarly uncompleteable, and that hence we cannot have sufficient reason for thinking that the criminal justice system's knowingly causing harm is justified. The first half of this claim follows immediately from the fact that such arguments are similarly dependent on being able to specify the necessary and jointly sufficient conditions for an agent being morally responsible in order to support the claim that ordinary agents are so responsible. However, the inference to the claim that we do not have sufficient reason risks setting the bar too high. In practice we will always face doubt and uncertainty, so holding an action that has anything but completely certain justifications to be impermissible is to argue for the complete paralysis of all realistic agents.

The second observation is that, given that we must employ standards somewhere between certainty of truth and certainty of falsehood, the real question is where these should be set. The most theoretically developed way of tackling that issue is expected utility theory, which holds (roughly speaking) that agents should assign probabilities to each relevant condition employing Bayesian reasoning, weight every option according to the sum of the probability-adjusted value of the possible outcomes, and prefer the option with the highest expected value. The threshold probability for the conjunctive condition that any one theory of moral responsibility is true and ordinary agents meet its conditions follows trivially. However, the non-consequentialist theories at stake in the present cannot directly adopt this solution, and it is unclear how non-consequentialist proponents of criminal justice might approach the task of setting such standards in a principled way.Footnote ⁶² The argument will therefore necessarily be somewhat tentative.

A first possible response is confrontational, and says that it is intuitively implausible that the threshold can be as low as appears to be the case for the reasons we have for thinking ordinary agents are morally responsible. Indeed, one proposal in the literature is that we should apply the BARD-standard at this stage in the process of justifying inflicting harm on offenders as well: unless we can be certain beyond a reasonable doubt that persons are morally responsible, we ought not to treat them as if they are.Footnote ⁶³ It seems clear, given both the widespread and considered disagreement on the composition of the correct list of necessary and jointly sufficient conditions for moral responsibility, and the uncertainty of whether ordinary agents as understood in the light of neuroscience will be capable of meeting it, that proponents of criminal justice cannot currently claim to have reason to think that persons are morally responsible beyond a reasonable doubt.

Alternatively, we might approach the problem as one under uncertainty, where we cannot apply expectations because we do not have sufficient information to even assign probabilities to the various conditions or outcomes. In such situations, however, it is widely recognized that we should aim to avoid the worst outcome, and it seems highly plausible that the world in which we employ harmful practices of criminal justice against innocent people because no one is morally responsible is far worse than the alternatives.Footnote ⁶⁴ This world after all is one in which criminal justice systems commit grievous moral wrongs against millions of innocent people every year.Footnote ⁶⁵

The above is unlikely to persuade at least some non-consequentialist proponents of criminal justice, so let me make a second, somewhat more concessionary argument. This rests on the observation that, although obviously we must use such standards, setting them and determining whether or not they have been met is a task for the proponents of criminal justice more so than for the critic. Indeed, we typically think that it behoves those who knowingly cause harm not only to provide justification, but also to show that such justification has met a reasonable standard, not that the harm-causing action is permissible unless and until a critic can show that the justification fails to meet such a standard. Thus, even if some believe that it remains permissible to engage in the practices of criminal justice while doubts about its justification prevail, surely non-consequentialist proponents of criminal justice should expend all reasonable effort in the attempt to dispel them.

This idea can be summarized as the trilemma of alternative justification, abolitionism or reasonable efforts:

1. (1) If it is unclear whether an agent has sufficient reason to think that she has justification for knowingly causing harm, she must (a) provide a different justification, (b) avoid causing such harm, or (c) expend all reasonable effort to ascertain that P justifies knowingly causing avoidable harm (the trilemma).

2. (2) It is unclear whether we have sufficient reason to think offenders are morally responsible in the sense that would justify knowingly causing them harm (the open challenge from neuroscience to criminal justice).

3. (3) We must either avoid knowingly causing harm in the criminal justice system (i.e. become abolitionists), provide a different justification, or expend all reasonable effort to show that offenders are morally responsible in the sense that would justify knowingly causing harm (from 1 and 2).

4. (4) Proponents of criminal justice are not abolitionists (by definition).

5. (C) Proponents of criminal justice who do not provide a different justification must expend all reasonable efforts to ascertain that offenders are morally responsible in the sense that would justify knowingly causing harm (from 3 and 4).Footnote ⁶⁶

Meeting this requirement would be no mean feat. Settling the issue of the moral responsibility justification for criminal justice requires tackling some of the most complex and controversial foundational questions in normative ethics, philosophy of mind and theories of free will and moral responsibility. It involves filling in the blanks I discussed in the last section, including deciding what the best account of moral responsibility is (and whether it is good enough), determining whether this account justifies knowingly causing harm to offenders, and ascertaining that on our best understanding of human agency, informed by neuroscience, this kind of moral responsibility is available to ordinary human agents.

While the first of the challenges here presented suggests merely that these proponents of criminal justice should accept a moratorium on punishment so long as its justification is doubtful, the second argues that the burden of proof falls on those who take a questionable justification to make harmful acts permissible. I do not begrudge non-consequentialist proponents of criminal justice this labour – these are fascinating issues, the exploration of which is certainly worthwhile – but I do not much envy their chances of completing it; in fact, I suspect that non-consequentialist proponents of criminal justice will find the burden staggeringly difficult to lift. I wish them the very best of luck.Footnote ⁶⁷