A foundational principle of liberal political thought holds that the only legitimate reason to restrict a person's liberty against their will is to prevent harm to someone else -- never merely to prevent harm to themselves. Stated as a general rule, the principle sounds clean and easy to apply. It sits uneasily, however, beside the ordinary content of nearly every legal system in the world, each of which restricts a considerable range of choices that appear to harm no one but the person making them: mandatory seatbelt and helmet laws, restrictions on recreational drug use, cooling-off periods that delay a consumer from finalising a large purchase. If the harm principle were actually followed as stated, most of this ordinary regulation would have no legitimate basis at all -- and yet it persists, largely unchallenged, across societies that otherwise profess considerable commitment to individual liberty.

Consider the seatbelt law directly. A driver who chooses not to wear one appears, at first glance, to risk only their own life, a purely self-regarding choice the harm principle would seem to place entirely beyond the state's legitimate reach. Defenders of the law respond that the choice is not, in fact, purely self-regarding at all: an unbelted driver injured in a crash draws on a shared insurance pool, occupies emergency medical resources other patients might otherwise have used sooner, and imposes costs on a public healthcare system funded by people who never consented to subsidise that particular driver's risk. This is a real cost, but it is also a strategy capable of justifying nearly any restriction on personal risk-taking whatsoever, since almost every genuinely self-regarding choice can be redescribed, with sufficient effort, as imposing some diffuse cost on someone else somewhere in the system.

A more honest paternalist justification does not lean on this indirect-harm argument at all, and instead distinguishes two genuinely different reasons a state might restrict a self-regarding choice. Soft paternalism intervenes only to ensure that a choice is genuinely informed and voluntary -- a mandatory cooling-off period after a high-pressure sales pitch, for instance, does not override what a buyer actually wants, but simply protects the conditions under which a want can be said to be authentically theirs rather than a product of momentary pressure. Hard paternalism does something considerably more intrusive: it overrides a choice the state accepts was fully informed and genuinely voluntary, because the state has independently judged the underlying values behind that choice to be mistaken -- banning a competent, well-informed adult from bungee jumping regardless of how carefully that adult weighed the risk is hard paternalism in its purest form.

This distinction matters because soft paternalism sits comfortably within a liberty-respecting framework in a way hard paternalism never quite manages to. Soft paternalism does not actually substitute the state's judgment for the person's own true preference; it protects the very conditions -- adequate information, freedom from coercion, absence of panic or fleeting impulse -- under which a choice can genuinely be called that person's own. Hard paternalism does the opposite: it overrides a preference the state itself concedes was genuinely, freely held, replacing the person's own judgment about their own life with the state's judgment about what that person's life should contain. Liberal political theory has never found a comfortable way to justify this second move without simply abandoning the harm principle it otherwise claims to hold.

Most real-world laws that get labelled "paternalistic" are, on inspection, an uneasy mixture of soft-paternalist measures that genuinely protect informed, voluntary choice and hard-paternalist restrictions dressed in the borrowed language of indirect harm to others. A more honest test for sorting the two asks a single question: would a fully informed adult, free of panic, coercion, or momentary impulse, still form a genuine wish to make the restricted choice? If the answer is yes, restricting that choice is hard paternalism, and needs a justification the harm principle alone cannot supply. If the answer is no -- if the apparent choice was never really free or informed to begin with -- then correcting the gap is closer to respecting a person's autonomy than to overriding it, whatever the resulting law happens to be called.