Suppose every choice has causes. Nobody chose their starting brain, childhood, temperament, or the exact pressures acting on them when they caused harm. What, then, could justice possibly mean?
Table of contents
On this page
What Retribution Is Trying to Protect
The disputed idea is often called basic desert moral responsibility. A person has basic desert if they deserve blame or punishment merely because of what they knowingly did, independently of any future good that blaming or punishing them might achieve.
The difficult test is not an accident caused by confusion or coercion. Imagine somebody who understands the harm they are causing, plans it carefully, repeats it, and shows no remorse. Even if they will never harm anyone again, many people think justice still requires suffering in return. Anything less can seem like treating the original harm as unimportant.
That is the strongest case for retribution. It should not be reduced to bloodlust. As the Stanford Encyclopedia of Philosophy’s account of retributive justice explains, retributivism ordinarily combines deserved proportionate punishment with strict prohibitions against punishing the innocent or punishing the guilty excessively. It promises to recognise the wrong, respect the victim, and prevent social utility from turning a person into a disposable means.
Those protections matter. A system concerned only with outcomes could be tempted to punish an innocent person to reassure the public, or impose a grotesque sentence because it might deter others. Retributivism says no. Guilt and proportionality constrain the state.
Compatibilists offer another serious defence. They need not believe in the tiny, uncaused captain I rejected earlier in this series. They can locate responsibility in capacities such as understanding reasons, controlling conduct, and acting from one’s own values without coercion. On many compatibilist accounts, that kind of reasons-responsive agency is enough for moral responsibility even if the agent has a complete causal history.
I agree that these capacities matter. Intention is not the same as accident. Understanding is not the same as confusion. A person who responds to reasons can be asked to explain themselves and can change in response to consequences.
I just do not see how any of that makes their suffering good in itself. Their capacity to understand reasons also had causes. It can justify addressing them as an agent, but it cannot make them the ultimate author of the machinery doing the understanding.
Justice Is a Bundle
Part of the confusion is that we use justice to name several different things. Retribution is only one of them.
Accountability means establishing what happened and requiring the person who caused harm to answer for it. They can be asked for reasons, confronted with evidence, and expected to comply with measures that protect others. None of this requires believing they created themselves.
Censure means saying clearly that the act was wrong. Causal explanation does not turn assault into weather. A public judgement can affirm the norm that was violated without declaring the person an ultimately self-created monster. The distinction between censure and deliberately burdensome punishment is already important in philosophical accounts of legal punishment.
Restitution concerns what can be returned or repaired. Stolen property can be restored. Financial losses can be compensated. A person who caused damage can reasonably be required to contribute towards fixing it. The aim is to repair the victim’s position where possible, not to manufacture an equivalent amount of pain elsewhere.
Victim recognition is different again. Some harms cannot be repaired. Justice can still name what happened, believe the person who was harmed, acknowledge the wrong publicly, provide practical support, and give them a meaningful voice. No victim should have to forgive, reconcile, or meet the person who hurt them to make a non-retributive theory look compassionate.
Protection can require boundaries, supervision, removal from a role, or physical separation. If somebody presents a serious danger, keeping them away from potential victims may be necessary. Caused danger is still danger.
Deterrence also survives. A person can respond to laws, expectations, and consequences because those things become new causes. That does not mean harsher is always better. Daniel Nagin’s review of deterrence research finds stronger support for the deterrent effect of the perceived certainty of apprehension than for greater punishment severity. The evidence is uneven, so deterrence should justify only measures that have a credible preventive basis, not every punishment imposed in its name.
Rehabilitation becomes more important, not less. Education, treatment, stable housing, useful work, social support, and opportunities to practise different responses can alter the conditions from which later choices emerge. Rehabilitation is not a guarantee. It must not replace measures needed for victims’ safety or require victims to participate. It is simply more rational than treating character as an unchangeable moral essence.
Restorative justice can combine accountability, recognition, and repair. A 2013 Campbell systematic review of ten randomised experiments found a modest average reduction in detected repeat offending and higher victim satisfaction from face-to-face conferences. That finding applies to cases in which both victims and offenders consented before random assignment. It does not establish that restorative justice is suitable for every offence, relationship, or person. It must never become a cheaper way to pressure a victim into contact, forgiveness, or an unsafe agreement.
Structural prevention asks what repeatedly produces harm before an individual case reaches a courtroom. Trauma, deprivation, social isolation, poor treatment access, bad institutional incentives, and opportunities for abuse can all matter. Changing those conditions does not excuse the person who acted. It tries to produce fewer victims.
These aims can conflict. Repair may not reduce risk. Incapacitation may protect people while making rehabilitation harder. Censure may be necessary even when reconciliation is impossible. Removing retribution does not give us an automatic formula. It gives us clearer questions.
The People We Are Most Afraid Of
Deliberate cruelty remains deliberate without free will. The person’s intent is part of the causal story, not an illusion. Repeated harm and an absence of remorse may indicate that persuasion alone is unlikely to protect others. They are reasons for stronger protection, not proof of a metaphysical stain.
Some people may remain dangerous despite every reasonable attempt at change. Incapacitation can then be justified by protection. That may involve confinement, sometimes for a long time. The important difference is that confinement is no longer justified by a wish to make the person suffer. Every burden of preventive confinement must answer to safety.
This makes due process more important. Predictions of future harm can expand without the natural stopping point supplied by a completed offence and fixed sentence. Any preventive restriction therefore needs a high evidential threshold, legal representation, an independent decision-maker, written reasons, a right of appeal, the least restrictive effective conditions, and regular review.
Proportionality survives too, but its basis changes. A response should be proportionate to the seriousness and credibility of the threat, and no more burdensome than protection requires. Past conduct is relevant evidence. It is not a blank cheque for whatever treatment fear makes politically convenient.
Risk assessment cannot carry this weight by itself. A systematic review and meta-analysis of 73 samples found that commonly used tools were substantially better at identifying low-risk people than correctly identifying who would later commit violence. The authors warned against using them as sole determinants of detention, sentencing, or release. A score can inform judgement. It cannot replace evidence, rights, and review.
People deprived of liberty also retain their dignity. Article 10 of the International Covenant on Civil and Political Rights requires humane treatment and respect for the inherent dignity of every detained person. Rejecting desert strengthens that demand. If suffering is not the point, neglect and humiliation cannot be smuggled in as unofficial punishment.
None of this asks victims to feel calm or charitable. Anger can register that something precious was violated. A person may need permanent distance and may never want reconciliation. Compassion can be non-permissive: I can understand that your behaviour had causes, refuse to hate you, and still not let you near me.
The Quarantine Model
Derk Pereboom compares preventive detention with quarantine in his discussion of free will scepticism and criminal behaviour. We do not blame a person for carrying a dangerous disease, but we may temporarily restrict their liberty to protect others. The restriction should track the actual threat, use the least burdensome effective means, and include care for the person restricted.
Gregg Caruso develops this into a broader public-health quarantine model. Rather than waiting to punish individual offenders, society should reduce the background conditions that generate violence and other serious harm.
There is much I like here. The model separates protection from vengeance, makes every restriction justify itself, and directs attention upstream. It asks why we tolerated preventable conditions until somebody became dangerous.
The metaphor also worries me. People are not pathogens. Describing them that way can encourage fear, stigma, and indefinite exclusion. If predicted danger is enough, confinement can drift towards punishment for what somebody might do. Public-health institutions are capable of coercion too, and medical language can hide political choices and unequal treatment behind an appearance of technical neutrality.
The model can also obscure agency. A person who knowingly harms somebody is not merely a site where risk appeared. They can understand reasons, answer questions, recognise another person’s claim, and participate in repair. Those capacities may not create basic desert, but justice should still address them.
I would keep the quarantine model’s discipline without making it the whole of justice: prevent serious harm, prefer the least restrictive measure, support rehabilitation, address structural causes, and review every coercive burden. I would keep censure, answerability, restitution, victim recognition, and strict legal rights alongside it.
What Survives
Justice survives without free will because justice was never only about deserved suffering.
It can tell the truth about what happened. It can recognise the person who was harmed. It can require answers and repair. It can set firm boundaries, restrain dangerous people, change incentives, rehabilitate where possible, and alter the conditions that keep producing harm.
Intent, knowledge, coercion, capacity, and remorse still matter. They help explain the act, reveal present risk, and tell us which responses might work. What disappears is one further inference: that because this person, whose action had causes, knowingly did wrong, making them suffer is therefore good in itself.
That is not a softer justice. In some ways it is more demanding. Every consequence must now do real work.
The same is true beyond courts and prisons. If nobody is ultimately self-authored, we still have to choose, deliberate, set boundaries, and live together as people shaped by causes. Part 7 turns to that ordinary task: Doing Our Best in a Caused World.