Harm, Benefit, and Consent

A patient refuses an operation that would save their life. They understand the diagnosis, they understand that they will die without the surgery, and they decline. The surgeon operates anyway. The patient survives, recovers fully, and lives another twenty years.

Every legal system in the developed world treats what the surgeon did as a serious wrong, and most people's reaction agrees with the law. Yet the patient is alive, healthy, and better off by every measure that would have been used to decide the case had they been unconscious.

Something was done to them that no good outcome repairs. Working out what that something is requires three ideas, and this article takes them in the order they depend on each other.

The comparative idea

Harm and benefit look like the simplest concepts in the subject and they are not. The natural analysis of both is comparative, and stating it exposes the trouble.

The account is intuitive, it is what people mean most of the time, and it does a great deal of work. It explains why breaking someone's arm is a harm and why an unpleasant but life-saving treatment is not, since in the second case the alternative was worse.

It also breaks in four places, and the breaks are not curiosities.

The first is preemption. Suppose someone poisons a traveller's water bottle, and had they not done so a second person waiting behind them would have poisoned it in exactly the same way. The traveller is no worse off than they would have been. On the counterfactual account the poisoner did not harm them, which is plainly wrong.

The second is overdetermination. Two people each independently fire a shot that would have been fatal on its own. Neither made the victim worse off than they would otherwise have been, since the other shot was coming. On the counterfactual account neither harmed anybody.

The third is that some conditions look bad in themselves rather than by comparison. A person born with a painful condition, whose only alternative was never existing at all, is not worse off than they would otherwise have been. Whether they have been harmed is a question the comparative account answers no and many people's judgement answers yes, and this is one of the entrances to a problem that the advanced tier is largely about.

The fourth is the practical one. The account requires knowing what would otherwise have happened, and the alternative is a matter of dispute in most real cases. Whether a factory harmed a community depends on what the community's situation would have been without it, and the answer to that is contested in every actual case of the kind.

The alternative on offer is a non-comparative account, holding that certain states are bad for a person, full stop, and that causing someone to be in one is harming them whatever the alternative was. It handles the four cases above and gives up something valuable in exchange: it has to produce a list of bad states, and the list will be disputed, whereas the comparative account needed no such list. Neither approach has driven the other out.

Being the author of one's own life

This is a different idea from the autonomy that appears in Kantian moral theory, and the two are confused constantly. Kantian autonomy is the capacity of a rational will to give itself the moral law, and a person exercises it by acting on principles anyone could act on. Personal autonomy is about a particular person's authority over their own life, including the authority to make choices that no rational will could universalise and that most people would consider foolish.

The standard account of what makes a choice a person's own is hierarchical. Harry Frankfurt's version has it that we have desires and also desires about our desires, and that a person acts autonomously when the desire moving them is one they endorse from that higher standpoint. An addict who wants the drug and wants not to want it is not governing themselves when they take it, even though nobody is interfering. Gerald Dworkin developed the idea in terms of a person's settled character rather than a single higher-order desire.

There are two well-known difficulties. One is that endorsement can itself be manufactured, so a person raised to want nothing but service to others may endorse that desire wholeheartedly and still not look self-governing. Purely procedural accounts, which ask only about the structure of the person's attitudes, have no way to say so; substantive accounts add conditions on the content of the person's values and then have to defend those conditions against the charge that they are simply the theorist's own. The other, pressed by writers on relational autonomy, is that self-governance is not a solitary achievement: the self-respect and self-trust that make it possible are developed in relationships, and an account that treats autonomy as an internal structure misses the social conditions it depends on.

Being able to decide this

The four abilities named in that definition are the standard clinical framework, developed by Thomas Grisso and Paul Appelbaum and set out in their Assessing Competence to Consent to Treatment in 1998.

Three features of the concept matter and are regularly missed.

Capacity is decision-relative. It is not a global property of a person but a property of a person facing a specific decision at a specific time. Someone may have capacity to decide what to eat and not to decide about major surgery, and someone whose capacity fluctuates may have it in the morning and not in the evening. A finding of incapacity attaches to a decision, not to a life.

Capacity is about process, not outcome. The four abilities concern how a person reaches a decision, not what they decide. This matters because the temptation runs entirely the other way: when a patient makes a choice that looks disastrous, the immediate impulse is to doubt their capacity, and if the standard were the wisdom of the outcome, capacity assessment would become a mechanism for overriding any decision the assessor disliked.

Capacity may be risk-relative, and this is disputed. On the asymmetry view, the threshold should be higher for decisions with worse expected consequences, so that more is required to establish capacity to refuse a low-risk life-saving treatment than to accept it. The argument for it is that the cost of an error differs between the two. The argument against it is that it makes a person's capacity depend on the merits of their choice through the back door, which is what the process standard was designed to prevent.

What agreement does

The four conditions in the definition are the standard elements: disclosure, understanding, voluntariness, and capacity. Each can fail independently, and a failure in any one of them means that the consent obtained does not do the work consent is supposed to do.

The institution is recent and it was built in response to atrocity. The Nuremberg Code, part of the judgment delivered in August 1947 after the trial of physicians who had experimented on prisoners, opened by declaring the voluntary consent of the human subject absolutely essential. The World Medical Association's Declaration of Helsinki, adopted in 1964, developed the requirement for clinical research and has been revised repeatedly since.

The philosophically interesting fact about consent is that it changes what may be done. An incision made without consent is a battery; the same incision with consent is surgery. The physical act is identical and its moral character is not. Consent does this by waiving a claim the person holds against interference, which is why it is the person whose body it is who has to give it, and why the consent of someone else who loves them will not serve.

This is what the surgeon in the opening case failed to obtain, and why the good outcome does not repair it. The wrong was not done to the patient's health, which was improved. It was done to their standing as the person entitled to decide, and no improvement in health is the same kind of thing.

Overriding a person for their own sake

Three things are needed for an act to be paternalistic in this sense: it restricts the person, it is done without their agreement, and the reason offered is their own good. An intervention justified by harm to others is not paternalism whatever else it is, and this distinction does most of the work in practice, because many restrictions defended as protecting people are in fact defended by the costs their choices impose on everyone else.

Mill's On Liberty of 1859 is the founding statement of the anti-paternalist position, and his own example is more nuanced than his reputation. A person is about to cross a bridge that an observer knows to be unsafe, and there is no time to warn them. Mill holds that stopping them is legitimate, because we do not know that they want to fall in, and liberty consists in doing what one desires. If they know the bridge is unsafe and choose to cross anyway, the case changes entirely.

That example marks the standard division. Soft paternalism intervenes only where the choice is not fully voluntary or not fully informed, and its aim is to protect the person's own decision from being made in the dark. Hard paternalism overrides choices that are voluntary and informed, on the ground that they are bad ones. Very few people reject soft paternalism, and hard paternalism is where the argument lives. The related contrast between weak and strong paternalism turns on whether the intervention corrects a mistake about the means to the person's own ends or substitutes different ends.

The case against hard paternalism has two strands. One is Kantian: overriding a competent person for their own good treats them as an object of management rather than as someone entitled to run their own life, and the insult is not cancelled by the benefit. The other is Mill's, and is broadly empirical: a person's own mode of living their life is likely to be better for them than one imposed, not because their choices are wise but because it is theirs, and because the people doing the imposing are usually worse informed about what suits them than they are.

The case for it also has two strands. One is that a temporary restriction can protect autonomy over a life, since some choices foreclose all subsequent choosing. The other is contractual: an intervention is defensible if the person would have agreed to it had they been fully informed and thinking clearly, which is why almost nobody objects to being stopped from stepping in front of a bus.

Next

The cases here have all involved someone acting on a person: operating, restraining, informing, withholding. A large family of practical questions turns instead on whether one did anything at all, and on what one meant by it. Whether letting a patient die is different from killing them, and whether a harm foreseen is different from a harm intended, are the next article's subject.