Free will and punishment

Do we deserve to be punished when we break the law? Many people would say yes, and there is some philosophical justification for this view. Kant, for example, argued for a retributivist theory of punishment. He reasoned that people in general are rational agents who choose their own actions, and should therefore be held morally responsible for them. With reference to the first formulation of his moral imperative, Kant argued that, by acting badly, criminals bring the punishment upon themselves – since by punishing them we are acting towards them as they are acting towards others. Moreover, he took this one stage further, by suggesting that the severity and type of punishment should fit the crime (the Biblical lex talons). For this reason, he argued that capital punishment was justified for (at least some) cases of murder. Kant felt that the sole justification for punishment was that somebody had behaved criminally. Whilst he accepted the utilitarian arguments that punishing people might increase overall societal happiness, deter other would be criminals, and possibly help to rehabilitate existing criminals, he nevertheless believed that these reasons could never be used as the justification for punishment. At best, they would be merely positive by-products of punishment. To do otherwise, he reasoned, would be to treat people as means to an end, rather than as ends in themselves (i.e. punishing the criminal would be the means to achieve the end of deterring other would be criminals, and punishing criminals would be the means to achieve the end of improving overall societal happiness). Kant argued that we should never do this.

However, Kant’s concept of retributive punishment presupposes that those being punished are rational agents who are able to freely choose their actions and make their own decisions, and are thus morally responsible for them. For this reason, Kant held that animals, babies, or the severely mentally impaired shouldn’t be punished for their actions. But, are we responsible for our actions in the way that Kant’s theory of retributive justice requires? Do we freely choose our actions? Could we have chosen otherwise? In other words, do we have free will? If not, then this theory of retributivist punishment does not apply to us, as it is unreasonable to hold people morally responsible for the consequences of actions that they were constrained to make. For example, if I was forcibly pushed out of a window, and landed on a pedestrian, then I could hardly be held morally responsible for any injury I caused them, as it was out of my control. Whether we are able to morally justify retributive punishment or not has clear implications for the way that we judge and punish criminals. As the possession of free will seems to be essential to assigning moral responsibility for actions, I will examine it in some detail.

Intuitively it does seem that we do (in general) possess free will, are rational agents, and should therefore be held morally responsible for our actions and decisions (on Kant’s theory). However, the existence or otherwise of free will is open to a great deal of metaphysical debate, and this calls into question our moral responsibility. This metaphysical debate first splits into those who accept determinism and those who don’t. Determinism is the theory that, given the state of the physical universe at any specific time, together with the laws of nature, the state of the physical universe at any other time (in the future or in the past) is uniquely determined. In particular, if we assume that our cognitive activity is entirely dependent upon physical processes within our physical brains (I will consider alternatives later), then our future cognitive activity (which governs our future decisions, actions etc.) is already uniquely determined by the state of the universe now, together with the laws of nature. So, on determinism, free will seems to be an illusion, as every one of my future decisions and actions is already uniquely determined. But, is determinism true?

According to classical theories of physics (e.g. Newtonian dynamics), the universe is, in principle, perfectly deterministic. However, on the subatomic level, there is some debate as to whether quantum events are deterministic or not. If indeterminacy is inherent in quantum events, as current theories suggest, then this would clearly rule out determinism (although, it should be noted that the effect of a certain amount of quantum indeterminacy might be limited at the macro level). However, it could be that events at the subatomic level are actually perfectly deterministic, but that the provisional nature of our current quantum mechanical theories doesn’t allow us to determine this. Nevertheless, even if there is some indeterminacy inherent in the quantum world, this might be irrelevant as far as moral responsibility is concerned – as we shall see a little later. So, determinism is at least a plausible theory that has some evidential support. Perhaps it is the argument to the best explanation.

Note that fatalism is not the same as determinism, and is not entailed by it. Determinism does not say that the future will be the same whatever course of action we choose, in the way that fatalism does. Rather, it says that there is only one possible future, and all of our future actions are already determined by the current state of the universe together with the laws of nature.

Those who reject determinism are known as libertarians. Some libertarians have supernatural beliefs that cause them to reject determinism, and others not. In the former camp are those who believe that we possess a non-physical mind or soul that somehow enters into the causal chain in the physical world leading to physical actions. According to supernatural libertarian theories, whilst the physical universe might be deterministic, the non-physical mind or soul can override the physical and allow us to retain free will. However, the proponents of this theory have yet to give any satisfactory explanation as to how a non-physical entity can have a causal effect upon a physical one (e.g. the physical brain). They have yet to produce any good empirical evidence that such non-physical objects actually do cause such effects (even if we don’t know how they do it). Moreover, they have yet to produce any good empirical evidence or compelling arguments that such non-physical entities exist at all. Thus, such theories lack plausibility, have little evidential scope and power, and require the positing of various ad-hoc entities and mechanisms. We could invent an infinity of such theories, and have no reliable way to discriminate between them in terms of likely truth. This is a very unreliable way of getting at the truth.

In the latter camp are those libertarians who reject determinism for naturalistic reasons e.g. by appeal to quantum indeterminacy to create some ‘elbow-room’ for free will. Whilst such theories at least have the benefit of plausibility, they would appear to achieve nothing more than allowing that some events in the brain are random (due to the random nature of some quantum events). However, this wouldn’t give us genuine free will. One could no more be held morally responsible for the consequences of random events in the brain than for the results of actions constrained by a lack of free will. So, whilst some of these non-supernatural libertarian theories are feasible, it is not at all clear how they can entail the free will that is necessary for moral responsibility.

Amongst the determinists, there are those that believe that determinism is incompatible with free will, and those who believe that the two are actually compatible (i.e. not mutually exclusive). The former are known as hard determinists, and the latter as compatibilists. The compatibilist argues that we retain free will so long as we are not actually forced to make a particular choice (by having a gun to one’s head, for example). They further argue that possession of free will under this definition entails moral responsibility. Hard determinists argue that compatibilists are wrong to base their definition of free will upon the everyday usage, rather than on causal necessity, in order to give them a basis for moral responsibility. They argue that we might feel that we have free will, but that determinism entails that this is an illusion. The hard determinist also notes that only certain constraints would remove free will in the sense that the compatibilist employs it. Being physically restrained or hypnotised might do it, but having a gun to one’s head would not, since one can still freely make the choice to be shot.

More importantly, the hard determinist would argue, why does the compatibilist think that certain causes are compatible with free will (determining ones), and others not (constraining ones)? Is this distinction not arbitrary? For example, imagine that someone is hypnotised to always choose a particular flavour of ice cream over others, but is unaware that they are hypnotised (this example is based upon one given by Stephen Law). As far as they are concerned, they are always freely choosing chocolate rather than anything else but, in reality, they are under the control of the hypnotist. The compatibilist would probably agree that this person no longer has free will, as they cannot make any other choice but chocolate (or whatever). But, what is the relevant difference between this situation and the one under determinism? In both cases the choice is constrained, as no other choice is actually possible, but the person still has an illusion of free will. Moreover, the onus is on the compatibilist to explain why we retain moral responsibility under one type of constraint (the deterministic type), but not under another (the everyday sort). Why should a person be morally responsible for an action that they were powerless to avoid?

For example, on determinism, it might be an absolute certainty (from the state of the universe now, together with the laws of nature) that a particular person will be born tomorrow, and that this person will go on to commit a murder thirty years later. This future is the only one possible. However, as this future is uniquely determined already, how can we be justified in eventually holding this person morally responsible for this action, even though it is something that they cannot avoid doing (any more than we can hold morally responsible the person who is pushed out of a window and lands on a pedestrian)? We can further motivate this concept by comparing it with the past – which is similarly invariant on determinism. Imagine it is a fact that somebody committed a murder yesterday. Further imagine that we can time-travel into the past to witness this murder. Would we say at the moment of committing the act of murder that the person involved has the free will to choose not to do it – despite the fact that we know that the murder will be committed (as we have come from the future, where it is a recorded fact that this murder was committed)? Since we already know the outcome, we know that the person involved must commit this act. It is hard to see how we could defend the argument that the person has the free will to choose not to commit the murder, as free will seems to entail that one could choose to act differently.

Perhaps we can defend the idea that free will doesn’t require choice? One thought experiment that is used to motivate compatibilism (courtesy of Harry Frankfurt) asks us to imagine a person who intends to rob a bank. Unbeknownst to this person, a bank-robbery loving demon is watching him, and will step in to force him to rob the bank should he choose to call off the robbery. However, the demon’s intervention is not required, as the person does not change his mind. In this case, it seems intuitively that the person freely chose to rob the bank, even though no other choice was actually possible (because of the demon). Moreover, from his personal perspective, he retained a feeling of free will. From this we are supposed to conclude that free will does not require that we actually have more than one possible choice – free will is compatible with this constraint. However, the hard determinist could argue that the situation under determinism is closer to the hypnotism thought experiment above, rather than to the bank robbery one – which is a subtle but important difference. Frankfurt’s thought experiment, they would argue, leads our intuition astray, as it makes us feel that the robber initially has a perfectly free choice, and that the demon will only need to enforce one particular choice if the robber makes the wrong one. However, on determinism, there is never a time when the choice is free – as the choice is already determined before the person makes their choice (in fact, from the beginning of the universe). They are constrained just as the hypnotised person is. So the case that free will doesn’t require choice is not made.

The compatibilist says we are free if our actions come from our character and desires without coercion. However, on determinism, our characters and desires are caused by forces outside our control – we did not choose our characters or our desires. This is a serious problem for the compatibilist. This objection actually applies to the supernatural type of libertarianism as well. In that case, our actions come from our character and desires, as well as our soul (either acting independently, or influencing our character and desires). The only difference from the compatibilist case is the introduction into the equation of a supernatural soul. However, the design and properties of our soul were not within our control either – they were designed and given by God (or some other supernatural agency). Even if our soul changes according to our (supposedly free-willed) life choices, then the rules governing any such changes were created by God (or whatever). So, this is out of our control as well, and free will becomes incoherent on this theory too.

So, where does each of these metaphysical possibilities leave us with regard to moral responsibility? Of the varieties of libertarianism, only the supernatural variety would seem to allow any room for genuine free will (and that’s open to debate, as the theist’s concept of free will may be incoherent). So, on this theory moral responsibility might exist (notwithstanding the implausibility of the theory itself).

Of the theories of free will that admit determinism, only one might be compatible with free will, and hence moral responsibility – compatibilism. However, in order to rescue free will, the compatibilist has to explain why it is that free will might be removed by some physical constraint (handcuffs, hypnotism etc.), but not by determinism (without equivocating between relevantly different concepts of free will). Furthermore, they must explain why the concept of free will is not incoherent. Hard determinism does not allow for genuine free will. The non-supernatural variety of libertarianism allows for a certain amount of randomness, but this doesn’t seem to entail that we have genuine free will.

So, where do we stand when we return to Kant’s theory of retributive justice? Kant argued that it should apply to rational agents who are free to choose their actions. Under these circumstances, such agents are morally responsible for their actions and, if necessary, should be punished accordingly. However, if we are not free to choose our actions because such actions are uniquely determined by the state of the universe, together with the laws of nature, then retributive justice should not be applicable. As we have seen, hard determinism dictates that we do not have genuine free will so, on this theory, retributive justice would be inappropriate. Furthermore, whilst compatibilists and those who subscribe to some non-supernatural variety of libertarianism wish to allow for free will, it can be argued that they have so far failed to successfully make their case – at least in a form that would entail moral responsibility. So, on both of these theories, retributive justice would also seem to be inappropriate. The only theory of free will that I have examined that might allow for genuine free will is the supernatural type of libertarianism (and this theory is implausible for a variety of reasons).

If we reject free will, and hence moral responsibility, then it would seem that retributive justice (at least as formulated by Kant) is untenable. In this case, should we still be in favour of punishing those guilty of crimes? Perhaps there is a way of reconciling determinism or libertarianism with free will, and this retributive justice but, regardless, we can still justify punishments, including the removal of freedom, on Utilitarian grounds. If a criminal is removed from society, then they are no longer able to commit such crimes – which tends to improve the overall happiness and well-being of the rest of society. Furthermore, such punishments plausibly act as a deterrent against the committing other crimes (if you think not, then try to imagine a society in which there was no legal consequence for law-breaking). One problem with the utilitarian approach, as opposed to the Kantian one, is that it doesn't inherently entail that we punish the criminal himself, rather than some innocent scapegoat (it also does not entail that we match the severity of the punishment to the severity of the crime). If this latter approach resulted in an increase in societal happiness, then why shouldn't we do this on utilitarianism? However, the utilitarian can counter this objection. Firstly, it would not prevent the real criminals from committing more crimes, as they would still be at large - so any increased societal happiness would probably be short-lived. Secondly, the arbitrary punishment and removal of liberty that this strategy entails would indirectly lead to less happiness and societal well-being, as people would justifiably fear that it might happen to them. So, on a more sophisticated version of utilitarianism, I think that we can actually derive the rule that only the real criminal should be punished.

Whilst the end results of punishment on retributive justice and on utilitarian grounds might have similarities, it is likely that they would diverge in many ways. Under hard determinism, we could no longer justify punishment in terms of being ‘an eye for an eye’ (although such an act might arguably psychologically benefit the victims of crime, which could form a utilitarian justification for retribution in particular cases). Instead, it would perhaps need to be justified on utilitarian grounds as maximising the overall happiness of the rest of society. By contrast, if we were to accept that we do have genuine free will, and we do not reject retributive justice for some other reason, then we could easily justify beating violent criminals, and executing murderers, for example (with no thought to any benefits from deterrence and societal happiness). Without retributive justice, such justifications become much more difficult (but perhaps not impossible). Our justice systems include elements of retributive and utilitarian theories of punishment. If we do have genuine free will then, arguably, retributive justice could be given much more focus. Conversely, if we do not possess genuine free will, then any elements of retributive justice should arguably be removed altogether from the legal system. In this latter case, we can still say that somebody is bad, but allow that this was totally outside of their control. In a sense, they are just unlucky. Nevertheless, as part of our implicit contract with the state, we agree to hand over much of the responsibility for our protection, on the understanding that the state will take this role. That carrying this out will involve restricting the liberty of people who might have had no choice but to behave as they do is unfortunate but unavoidable, as it is better than any of the other options. If the state did not do this, then the inevitable resulting rises in crime would increase the total sum of misery, and people would inevitably start to take matters of justice into their own hands, probably leading to the eventual breakdown of the state.

As human beings, we are drawn to retributive justice, as it was adaptive behaviour when enforcing social contracts. However, we need to consider if such behaviour is still appropriate. If determinism is true, as I believe it to be, then perhaps we could still attempt to rescue retributive justice by appeal to the Social Contract theory of morality, although there would be some serious obstacles to overcome. In any case, we should care about our conclusions, as the nature of our justice system affects us all in some way or another, whether directly or indirectly.

Addendum

What follows is a slightly different take on the above issues. Where it differs with the foregoing, then it should be considered to supercede it. In particular, I now believe that a broadly compatibilist position on free will is the most defensible one.

Introduction

In this post I will discuss the metaphysical concept of free will by analysing four particular theories of free will. In the course of this I will explain what these theories propose and what differentiates them; examine each theory and determine what would be the effects upon me and upon society in relation to the criminal justice system if this theory was true; and demonstrate how these effects follow from the tenets of these theories.

Definitions

Firstly, I need to explain some terms that will be used throughout this essay.

Free Will

A minimal definition of free will is the capacity of rational agents to freely choose their actions. However, there are two semantically different concepts of what it is for an action to be freely chosen.

1. On the first concept, an action is freely chosen if and only if one could have chosen otherwise in the very same circumstances (where ‘circumstances’ means the state of the universe immediately prior to the choice of action).

2. On the second concept, an action is freely chosen so long as this choice is caused by one’s desires and is uncoerced (i.e. it is voluntary).

Moral Responsibility

An agent is said to be morally responsible for an event or state of affairs only if they are the appropriate recipient of moral praise or moral blame for that event or state of affairs

Retributive Punishment

Retributive punishment is a theory of justice that holds that a proportionate punishment is a morally acceptable response to a crime, regardless of whether the punishment causes any tangible benefits (e.g. rehabilitating the criminal, deterring others from committing the crime, improving overall societal happiness etc.) Kant provided a philosophical justification for retributive punishment.

Utilitarian Punishment


A utilitarian punishment is one that aims to maximise (or at least increase) the overall utility. The utility has been defined in this context as happiness or pleasure. Others have defined it as the satisfaction of preferences. According to this analysis, increasing the overall utility is a necessary and sufficient condition for a punishment to be considered a morally acceptable response to a crime.

Determinism (causal)

Determinism is the philosophical thesis that every event in the universe is the inevitable result of antecedent causes. An alternative wording is that, on determinism, given the state of the universe at any specific time (where the ‘state of the universe’ subsumes the so-called laws of nature), the state of the universe at any other time (in the future or in the past) is uniquely determined (which doesn’t entail that we can actually predict this past or future state of the universe).

Fatalism


Fatalism is a philosophical doctrine emphasizing the subjugation of all events or actions to fate or inevitable predetermination.

Hard determinism

Hard determinism is the conjunction of determinism with the first of the definitions of free will.

Determinist compatibilism


This is the conjunction of determinism with the second of the definitions of free will.

Libertarianism

This is the theory that claims that some actions are exempt from the causal laws. It is split into Supernatural Libertarianism and Natural Libertarianism. The former posits that a non-physical mind or soul that overrides physical causality; and the latter that there is some level of physical indeterminacy present in the universe (for example, due to random quantum events), so that casual determinacy is not universal.

Justify

To justify is to demonstrate or prove to be just, right, or valid; or to show to be reasonable, or to provide adequate ground for.

Morally acceptable

An action is said to be morally acceptable if it is not in accordance with the ‘bad’ as determined by some mutually agreed moral system. Therefore, it might either be neutral, or in accordance with the ‘good’ as determined by that system.

Rational agent

A rational agent is one that is able to reason based upon knowledge of its environment in order to freely choose actions that will maximise the chances of it achieving its goals


Four Metaphysical Theories of Free Will


1) Hard determinism

As previously mentioned, this is the conjunction of determinism with the first of the definitions of free will. This first conception of free will is not compatible with determinism, for those ‘circumstances’ in the definition of free will determine everything that happens afterwards, including the choice to perform that particular act and the subsequent action. In other words, on hard determinism (basic argument for hard determinism):

P1: If everything we do is the inevitable result of antecedent causes, then we do not have free will (using the first definition of free will)
P2: Everything we do is the inevitable result of antecedent causes (from the definition of determinism
C: Therefore, we do not have free will

[NB. the premises of this argument will be referred to again in subsequent sections]

I will go down one more level to derive P1, as it is not self-evidently true (I could go further than this, but I think that this is sufficient for these purposes):

P3: If everything we do is the inevitable result of antecedent causes, then we could not have chosen to act otherwise in the very same circumstances (as our choice is causally determined)
P4: If we could not have chosen to act otherwise in the very same circumstances, then we do not have free will (from the first definition of free will)
C: Therefore, if everything we do is the inevitable result of antecedent causes, then we do not have free will

So, hard determinism, by accepting both determinism and the first definition of free will, entails that we do not possess free will.


2) Determinist compatibilism

This theory accepts determinism, but rejects the first definition of free will in favour of the second (and therefore rejects P1 from the basic argument for hard determinism). So, on compatibilism, as one can choose an action based upon one’s desire and free from coercion, and that choice of action still be the inevitable result of antecedent causes, compatibilism and the second definition of free will are compatible. In other words, on determinist compatibilism:

P1: If some thing does not coerce us into choosing a certain action, then this thing is compatible with our free will (from the second definition of free will)
P2: Determinism does not coerce us into choosing a certain action
C: Therefore, determinism is compatible with our free will

Whilst determinism entails that all our choices are the inevitable result of antecedent causes, we are nevertheless not coerced in the usual sense of the word (which is how the compatibilist uses the word).
So, as determinist compatibilism uses the second definition of free will, the acceptance of determinism does not prevent us from possessing free will i.e. free will and determinism are compatible. This contrasts with hard determinism, in which free will and determinism are not compatible.


3) Supernatural libertarianism

This theory rejects P2 from the basic argument for hard determinism, but may accept P1 (depending upon the exact kind of supernatural libertarianism). Using this definition does not rule out free will though (as with hard determinism), as it posits that we possess non-physical minds or souls that can override physical causality – meaning that not everything we do is the inevitable result of antecedent causes. This non-physical mind or soul is often supposed to have been given to us by God. More formally, on such a variety of supernatural libertarianism:

P1: If [the standard Christian] God exists then he would desire us to have free will (so as to not be mere automata, and to be capable of choosing to do good or bad), and would be able to create us in any way that he desired (being omnipotent and omniscient he can do anything that is logically possible, and free will is assumed to be logically possible)
P2: [the standard Christian] God exists
C: Therefore, God created us with free will

Moreover, God is supposed to have realised his desire for us to have free will by equipping us with a non-physical mind or soul that can override physical causality and allow us to make completely free choices. So, supernatural libertarianism approaches the subject from a different angle – making our free will one of God’s design aims when creating humans.

4) Natural libertarianism

As with supernatural libertarianism, natural libertarianism also rejects P2, but may accept P1. By positing some quantum indeterminacy in the universe, including in our brain, it says that not everything we do is the inevitable result of antecedent causes. So, natural libertarianism:

P1: If not every event in the universe (including those in our brain) is the inevitable result of antecedent causes, then there is some ‘elbow-room’ that may allow us to have free will
P2: Due to quantum indeterminacy, not every event in the universe (including in our brain) is the inevitable result of antecedent causes
C: Therefore, then there is some ‘elbow-room’ that may allow us to have free will

By utilising the concept of the randomness of quantum indeterminacy, natural libertarianism is seeking to create some ‘elbow-room’ to allow for free will. As it stands, this argument doesn’t establish that we do have free will, merely that free will is not ruled out for deterministic reasons.


Implications of each theory for the punishment of criminals

I will now consider whether the truth of the previously described metaphysical theories of free has any implications for the punishment of criminals. In order to do that I will first need to need to briefly describe four main moral justifications given for punishment, and then explain what moral responsibility is and how it relates to punishment, as the subsequent discussion is closely bound up with these concepts.


Four justifications for punishment

Arguments that seek to justify punishing those who break the law typically proceed along one or more of the following lines:

1. Retribution - proportionate punishment is a morally acceptable response to a crime, regardless of whether the punishment causes any tangible benefits

2. Deterrence – punishment is a morally acceptable response to a crime if such punishment acts as an effective deterrent against the committing of similar crimes in the future

3. Incapacitation – punishment is a morally acceptable response to a crime if such punishment prevents the offender from committing a similar offence (e.g. by incarcerating them)

4. Rehabilitation – punishment is a morally acceptable response to a crime if it includes work to reform and rehabilitate the wrongdoer so that they will not re-offend.

Note that numbers 2, 3, and 4 are (at least partly) utilitarian arguments for punishment, as they seek to maximise the overall utility.

Moral responsibility

I will discuss this in rather more detail, as I think that the implications of each theory for the punishment of criminals ultimately turn on part of the concept of moral responsibility.

Whatever type(s) of punishment is proposed for some criminal act, both retributive and utilitarian theories of justice presuppose that the person being punished is morally responsible for the event or state of affairs in question. That is, the person being punished is the appropriate recipient of moral blame for that event or state of affairs. Even though the moral frameworks, crimes, and punishments vary between these theories of justice (e.g. some may appeal to divine command, others to utilitarian arguments; some include punishment for ‘crimes’ such as blasphemy and heresy, and others not; some apply mainly retributive methods, and others not etc.), it is nevertheless acknowledged that the punishing of those not considered to be the appropriate recipient of moral blame is unreasonable and morally unacceptable. So, moral responsibility for some event or state of affairs is regarded as a necessary condition for punishment. It may not be a sufficient condition, however. Perhaps the act in question is not deemed to be sufficiently serious to warrant punishment, for example. Of course, cases exist of people being punished when they are not the appropriate recipient of moral blame, but these are seen as being problematic exceptions that are not morally acceptable. For example, justice systems (at least in the US and the UK) go to great lengths to avoid such punishments.

I would like now to formalise the concept of moral responsibility a little by listing what I think are its three necessary and sufficient conditions. An agent is morally responsible for an act if the following necessary and sufficient conditions are met:

1. That the agent committed the act

2. The act is held to be in some sense wrong

3. The agent had no relevant and sufficient excuse for committing the act (e.g. it was not a mistake, an accident, due to coercion, due to ignorance, due to some mental problem etc)

A justification for these conditions being a necessary condition for a morally acceptable punishment is rather lengthy, so I have moved it to the appendix, where I also give a little digression on the evolutionary roots of a punishment/reward system. For now I will just take these conditions to be plausibly true.


Implications on each metaphysical theory of free will

So, if we now take my conditions for moral responsibility to be necessary prerequisites for a morally acceptable punishment, are there any implications on each metaphysical theory of free will? I will examine each in turn, and attempt to draw out any such implications. As I will show, the implications of each theory turn largely on the third of my conditions for moral responsibility.

1) Hard determinism

As previously discussed, hard determinism entails that we do not possess free will (where free will in this case is the first of my definitions). Now, when it comes to assigning moral responsibility for some event or state of affairs on hard determinism, the first two of my moral responsibility conditions are not contentious. However, when considering the third condition, the hard determinist would argue that the agent who committed the act in question does in fact have a sufficient excuse for committing it, as they did not possess free will, and this lack of free will means that they could choose no other action. Hence, on hard determinism, condition 3 of my moral responsibility definition is not met, and the agent cannot be held morally responsible for that act – so any punishment for that act would be morally unacceptable.

So, on hard determinism:

P1: If one has a relevant and sufficient excuse for committing an act, then one should not be held morally responsible for that act (from the third moral responsibility condition)
P2: An agent who does not possess free will has a relevant and sufficient excuse for committing an act
C: Therefore, an agent who does not possess free will should not be held morally responsible for an act

Corollary: If an agent is not morally responsible for some act, then it should not be punished for it (as moral responsibility is a necessary condition for punishment, as discussed in the appendix)

So, on hard determinism it is not morally acceptable to punish an agent for some event or state of affairs, as the agent can never be morally responsible for that event or state of affairs, and moral responsibility is a necessary condition for morally acceptable punishment.

How about the conjunction of hard determinism and utilitarianism? For a utilitarian, the justification for not punishing an agent who is not morally responsible for some act is itself a utilitarian one (as I discuss in the appendix). However, when we look back at the four types of punishment that I listed earlier (i.e. retribution, deterrence, incapacitation, rehabilitation), then a utilitarian could also argue that it is morally acceptable to apply the last three types of punishment in response to some criminal act if doing this will maximise the utility. So, the utilitarian has an apparent conflict. However, they do have a possible way out of this conflict, as they might be able to determine whether, in the absence of complete moral responsibility (i.e. moral responsibility condition 3 not being satisfied), it will maximise the utility to punish the agent or not. They might, for example, determine that so long as the agent committed the wrong act then, even in the absence of free will, it will still maximise the utility to punish the agent in a way that deters, incapacitates, or rehabilitates.

The retributionist also has a conflict on hard determinism, as the retributive theory of punishment (or, at least Kant’s version of it, which is what I am using here) is unconcerned with the consequences of the punishment, but does require that the person being punished is a free and rational agent. But, if the agent does not have free will, then they can hardly be classed as a free agent (I discuss this in more detail in the appendix). Unlike the utilitarian, the retributionist has no way out of this conflict, and must agree that a retributive punishment is morally unacceptable.

So, on utilitarian punishment we have:

P1: If an agent committed a wrong act without possessing free will, then it would still be morally acceptable to respond to this act by punishing them in such a way that it deters, incapacitates, and rehabilitates so long as doing this will maximise the utility
P2: An agent committed a wrong act without possessing free will
C: Therefore, it would still be morally acceptable to respond to this act by punishing them in such a way that it deters, incapacitates, and rehabilitates so long as doing this will maximise the utility

On retributive punishment we have:

P1: If an agent is not free and rational, then it is not morally acceptable to punish them retributively (from Kant’s theory of retributive justice)
P2: An agent without free will is not free and rational
C: Therefore, it is not morally acceptable to punish them retributively

So, to summarise, the conjunction of hard determinism and utilitarian punishment is superficially incompatible, but we might be able to overcome this incompatibility by determining the course of action that will maximise the utility. The conjunction of hard determinism and (Kant’s theory of) retributive punishment is incompatible (in fact, as an aside, if we refer to Kant’s Categorical Imperatives then the truth of hard determinism would rule out any punishment at all, as a utilitarian punishment would go against the imperative to not use a rational agent purely as a means to an end). Unlike the utilitarian case, (Kant’s theory of) retributive punishment cannot be morally acceptable on hard determinism.

2) Determinist compatibilism

As mentioned earlier, this theory accepts determinism, but rejects the first definition of free will in favour of the second (and therefore rejects P1 from the basic argument for hard determinism).

Again, when it comes to assigning moral responsibility for some event or state of affairs, compatibilism finds the first two of my moral responsibility conditions to not be contentious. However, when considering the third condition, the compatibilist would disagree with the hard determinist, and argue that the truth of determinism does not grant the person committing the act a relevant and sufficient excuse. They would argue that we do in fact possess free will so long as we can choose an action based upon our desires and free from coercion. The fact that this choice is the inevitable result of antecedent causes is irrelevant to free will, as the compatibilist is using a different definition of free will to that used by the hard determinist. Hence, the truth of determinism does not imply a lack of free will, with the resulting lack of moral responsibility. So, as the moral responsibility conditions can be satisfied, there is no conflict between the conjunction of determinist compatibilism and either utilitarian or retributive punishment. So, on determinist compatibilism:

P1: If an agent commits a wrong act and does not have a relevant and sufficient excuse for committing that act, then the agent should be held morally responsible for that act (from the moral responsibility conditions
P2: An agent that commits a wrong act is not given a relevant and sufficient excuse by the truth of determinism alone (as the conjunction of determinism and free will is possible)
C: Therefore, such an agent should be held morally responsible for that act

Or, worded a little differently:

P1: If a rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and without coercion), then the agent should be held morally responsible for that act (from the moral responsibility conditions)
P2: A rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and without coercion) (as free will is possible)
C: Therefore, such an agent should be held morally responsible for that act

So, determinist compatibilism is compatible with both utilitarian and retributive theories of punishment, as the necessary moral responsibility conditions can be met.


3) Supernatural libertarianism

As mentioned earlier, this posits that we possess non-physical minds or souls that can override physical causality – meaning that not everything we do is the inevitable result of antecedent causes. Therefore, supernatural libertarianism proposes that we do possess free will.

Again, supernatural libertarianism takes no issue with the first two of my moral responsibility conditions (although its definition of a ‘wrong’ act may be different to that under non-supernatural theories if it is based upon divine command). However, with regard to the third condition, as determinism is rejected then it cannot be cited as a reason for us to not possess free will, and hence as a relevant and sufficient excuse for committing some act. So, on supernatural libertarianism, we can possess free will under both of its definitions, and so moral responsibility is possible.

So, on supernatural libertarianism:

P1: If an agent commits a wrong act and does not have a relevant and sufficient excuse for committing that act, then the agent should be held morally responsible for that act (from the moral responsibility conditions)
P2: An agent that commits a wrong act is not given a relevant and sufficient excuse by citing determinism (as determinism is false)
C: Therefore, such an agent should be held morally responsible for that act

Or, worded a little differently:

P1: If a rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and without coercion), then the agent should be held morally responsible for that act (from the moral responsibility conditions)
P2: A rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and without coercion) (as free will is possible)
C: Therefore, such an agent should be held morally responsible for that act

So, supernatural libertarianism is compatible with both utilitarian and retributive theories of punishment, as the necessary moral responsibility conditions can be met.


4) Natural libertarianism

As mentioned earlier, natural libertarianism posits some quantum indeterminacy in the universe, including in our brain, and says that not everything we do is the inevitable result of antecedent causes. So, on natural libertarianism, free will is held to be possible.

Natural libertarianism takes no issue with the first two of my moral responsibility conditions. However, with regard to the third condition, as determinism is rejected then it cannot be cited as a reason for us to not possess free will, and hence as a relevant and sufficient excuse for committing some act. So, on supernatural libertarianism, we can possess free will under both of its definitions, and so moral responsibility is possible.

So, on natural libertarianism:

P1: If an agent commits a wrong act and does not have a relevant and sufficient excuse for committing that act, then the agent should be held morally responsible for that act (from the moral responsibility conditions)
P2: An agent that commits a wrong act is not given a relevant and sufficient excuse by citing determinism (as determinism is false)
C: Therefore, such an agent should be held morally responsible for that act

Or, worded a little differently:

P1: If a rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and
without coercion), then the agent should be held morally responsible for that act (from the moral responsibility conditions)
P2: A rational agent can knowingly and freely choose to commit a wrong act (based upon their desires, and without coercion) (as free will is possible)
C: Therefore, such an agent should be held morally responsible for that act

So, natural libertarianism is compatible with both utilitarian and retributive theories of punishment, as the necessary moral responsibility conditions can be met.


Summary

So, of our four metaphysical theories of free will, only hard determinism is incompatible with some punishments, as it deduces that free will is impossible, and thus determines that an agent cannot be held morally responsible for some wrong act. Specifically, it cannot be reconciled with the Kantian theory of retributive punishment, but may in theory be reconciled with utilitarian punishment. The other three metaphysical theories are all compatible with both retributive and utilitarian punishment, as they allow for the existence of free will, meaning that moral responsibility is possible.


Appendix

Justification of my Moral Responsibility conditions

In this appendix I will give the justification for the conditions for a morally acceptable punishment. Here they are again:

1. That the agent committed the act

2. That the act is held to be in some sense wrong

3. That the agent had no relevant and sufficient excuse for committing the act (e.g. if the act was a mistake, an accident, due to coercion, due to ignorance, due to some mental problem etc.)

How would we go about trying to justify punishing only the person who committed the act in question, as opposed to just punishing some innocent scapegoat? And, how about the other conditions? Well, we might analyse these conditions using one or more moral theories of punishment, as I will now do. Here I will restrict the justification to just utilitarian and retributive theories, although I could extend this to additional moral theories (Social Contract, Virtue etc.). I will not seek to justify the moral theories of punishment themselves, but just use them to help justify the conditions above.

The first of these conditions is usually just assumed to be self-evident. However, it is actually possible to formulate a utilitarian argument for punishment that avoids this condition. Utilitarianism’s rationale is to achieve future social benefits (by maximising the utility) so, for example, punishing an innocent scapegoat, might in itself act as a deterrent, or might (at least temporarily) increase overall societal happiness (if, for example, people believe that a murderer has finally been apprehended). Nevertheless, this is generally seen as a problem for utilitarianism, and one that should be overcome by using a more sophisticated formulation of utilitarianism, not a problem for the concept that a morally acceptable punishment requires that one should have committed the act.

On a more sophisticated utilitarian analysis of punishment, we can say the following:

P1: If a form of punishment will lead to a decrease in overall utility, then it is not a morally acceptable response to a crime (from the definition)
P2: Knowingly punishing innocent people will lead to a decrease in the overall utility (as people will fear the possibility of being punished when they have committed no crime; and also because people will reason that they may as well just commit crimes anyway if there is no direct link between being guilty and being punished - which will in itself lead to a decrease in overall utility)
C: Therefore, knowingly punishing innocent people is not a morally acceptable response to a crime

So, the utilitarian analysis supports the first of my conditions listed above.

On a (Kantian) retributive analysis, we may say the following:

P1: If a rational agent chooses to treat people in a certain way, then they decree that in their judgement this is the way that people are to be treated (from the first formulation of Kant’s moral imperative)
P2: A criminal (who is a rational agent) chooses to treat people in a certain way
C: Therefore, a criminal (who is a rational agent) decrees that in their judgement this is the way that people are to be treated

Furthermore:

P1: If we treat a rational agent in a relevantly similar way to the way they decree that people should be treated, then we are respecting their judgment as rational agents [as we are treating them in the way that they decree people should be treated]
P2: We treat a criminal (who is a rational agent) in a relevantly similar way to the way they decree that people should be treated (by punishing them proportionally to their crime)
C: Therefore, when we treat a criminal (who is a rational agent) in a relevantly similar way to the way they decree that people should be treated, we are respecting their judgment as rational agents

So, on a (Kantian) retributive analysis, we are justified in punishing a criminal who is a rational agent and whose action(s) has caused harm to others, for we are merely treating them as they decree that others should be treated. However, Kant’s analysis of retributive punishment clearly presupposes that we are punishing the guilty party, or else we wouldn’t be treating them as they treat others. So, again, the first of my conditions is supported.

It would be possible to do a similar analysis of punishment on other moral theories (Social Contract, Virtue etc.), but I think that we already have reasonable grounds for supposing that the first condition is valid, so I will not pursue it any further (although I think that the same conclusion would be reached on other theories).

How about the second condition – that the act must in some sense be wrong? On a utilitarian analysis, we might say the following:

P1: If a form of punishment will lead to a decrease in overall utility, then it is not a morally acceptable response to an act (from the definition of a utilitarian punishment)
P2: Knowingly punishing people for acts that are deemed to be morally acceptable will lead to a decrease in the overall utility (as people will tend to perform fewer such morally acceptable acts and, conversely, will perform more morally unacceptable acts if these are not punished) [this assumes that morally acceptable acts tend to increase the overall utility and vice versa]
C: Therefore, knowingly punishing people for acts that are deemed to be morally acceptable is not a morally acceptable response to an act

On a (Kantian) retributive analysis, punishment is a justifiable response for the committing of an act that causes harm to others, as we are treating the criminal in the way they decree we should treat others. However, this does exclude punishment for ‘victimless crimes’, as nobody is actually harmed (other than perhaps the person committing the act). So, to the extent that a ‘wrong’ act causes harm to others, then retributive theory holds that punishment is reserved for acts that are in some sense wrong. So, on the (Kantian) retributive analysis:

P1: If a rational agent freely chooses to act in a certain way, then we should treat them in a relevantly similar way
P2: A rational agent who freely chooses to act wrongly causes harm to others
C: Therefore, we should treat a rational agent who freely chooses to act wrongly in a relevantly similar way by causing them harm (i.e. by punishing them)

So, again I think that we have reasonable grounds for accepting the second of my conditions for a morally acceptable form of punishment.

This brings us to the third condition. On a utilitarian analysis, we can say:

P1: If a form of punishment will lead to a decrease in overall utility, then it is not a utilitarian punishment (from the definition)
P2: Punishing people who have a good excuse for the criminal act will lead to a decrease in the overall utility (e.g. if the act was a mistake, an accident, due to coercion, due to ignorance, due to some mental problem etc.)
C: Therefore, punishing people who have a good excuse for the criminal act is not a utilitarian punishment

Now, P2 is certainly not self-evident as it stands. However, I think that we might seek to justify it as follows:

P3: If people are punished for wrong acts committed when they are not rational agents, were not aware of the consequences of their actions, were not aware that the act was wrong, committed the act accidentally, or were not free to choose (as opposed to being rational, informed, and able to choose freely), then this will lead to a decrease in the overall utility [as any of us might conceivably commit a wrong act under such circumstances, and knowing that we would still be punished would lead to a decrease in overall societal happiness]

P4: People who commit a wrong act when they are not rational agents, were not aware of the consequences of their actions, were not aware that the act was wrong, committed the act accidentally, or who were not free to choose have a good excuse for that act

C: Therefore, punishing people who have a good excuse for a wrong act will lead to a decrease in the overall utility

The type of things that I have in mind when I say that people might have a good excuse for committing a wrong act are, for example: running over a child who darts out in front of the car you are driving in a careful and lawful manner; being forced at gunpoint to open a safe; poisoning somebody because, unknown to you, the bottle of water you give them has been tampered with; being clinically insane etc. By contrast, if you made a free, rational, and informed choice to commit a wrong act, then you would not have a good excuse.

On a (Kantian) retributive analysis, we reach the same conclusion:

P1: If a rational agent does not make a free, informed, and intentional choice to treat people in a wrong way, then they do not decree that in their judgement this is the way that people are to be treated
P2: A rational agent who has a good excuse for a wrong act does not make a free, informed, and intentional choice to treat people in a that way
C: Therefore, a rational agent who has a good excuse for a wrong act does not decree that in their judgement this is the way that people are to be treated

However, as it is a condition of (Kantian) retributive punishment that the rational agent committing the act is in effect decreeing that this is how we should treat others, then the lack of any such implied decree renders the punishment inappropriate.

So, I have provided reasonable grounds for thinking that my three conditions are necessary, as the absence of any one of them makes punishment morally unacceptable on both utilitarian and retributive theories of punishment. However, I have not yet shown that these conditions are sufficient. I will not attempt that here, due to lack of time and space, but this is not critical as the metaphysical discussion turns on my condition 3. So, even if other conditions are needed, condition 3 is the one of particular relevance to this discussion.