The technology industry's dominant moral framework is aggregative. It celebrates the democratization of capability, the collapse of the imagination-to-artifact ratio, the twenty-fold productivity multiplier. The celebration is not dishonest: the aggregate gains are real, and the cycle documents them with the specificity of someone who lived through them. What the aggregative framework cannot see is the structure of who bears the costs. The SaaS Apocalypse was not a destruction of value. It was a transfer: from established software companies to AI platforms and their shareholders, from the displaced to the connected, from the many to the few. The aggregate may net positive. The specific people who net negative are not consoled by the aggregate.
The practical implication of the separateness of persons for the AI transition is that distributive justice cannot be assessed by looking at total welfare or even at median welfare. It must be assessed by looking at the position of the least advantaged: the administrative assistants displaced by AI tools that perform their work competently; the communities organized around industries being restructured; the students whose educational institutions have not adapted; the workers documented in the Berkeley study whose burnout accumulated invisibly while the organization's output metrics improved. These people are separate from the engineers in Trivandrum whose capabilities have been multiplied. Their costs are not offset by those engineers' gains. Justice requires that the institutions be designed to address both.
Martha Nussbaum's extension of Rawlsian justice through the capabilities approach sharpens this further. What matters is not only what people have but what they are able to do and be. The AI transition expands some capabilities enormously—the capability to create, to build, to access information that was previously gated by institutional barriers. It threatens others: the capability to develop embodied expertise through productive friction, the capability to maintain relationships against the encroachment of productive compulsion, the capability to exercise genuine autonomy rather than the simulacrum of autonomy that consists of choosing among options pre-selected by an algorithm. Primary goods are not fungible. You cannot compensate a person for the loss of their professional identity by giving them money. The separateness of persons is also the separateness of what matters to each person.
Rawls introduced the separateness of persons in Section 5 of A Theory of Justice as the foundational objection to classical utilitarianism. The utilitarian tradition, from Bentham through Mill and Sidgwick, justified social arrangements by maximizing the sum (or average) of individual welfare. Rawls argued that this aggregation commits an error analogous to the error of assuming that one person's sacrifice can be justified by the gains it produces for the same person over time: “It does not take seriously the distinction between persons.” When society imposes costs on some to benefit others, the same logic that makes interpersonal sacrifice wrong at the individual level makes aggregate sacrifice wrong at the social level.
The critique was sharpened by Robert Nozick, who approached the separateness of persons from a libertarian rather than an egalitarian direction: if persons are separate, no redistribution is justified without their consent, regardless of how much it improves the aggregate. Rawls accepted the separateness while rejecting Nozick's conclusion: the separateness of persons demands not that we leave each person with what the market happens to deliver, but that we design institutions that no rational person—imagining they might be anyone—would have reason to reject. The veil of ignorance is the procedural mechanism for ensuring that institutional design honors the separateness of persons.
The non-fungibility of persons. Benefits to one person cannot compensate losses to another, because persons are not interchangeable units of welfare. The aggregate may improve while specific individuals are made worse off, and justice requires that the specific individuals be taken as seriously as the aggregate. This is the foundational objection to any policy analysis that justifies distributional harms by appeal to net positive outcomes.
The separateness of costs. The AI transition distributes costs along identifiable lines: the displaced professional bears identity dissolution; the burnout worker bears cognitive depletion; the community organized around displaced industries bears economic collapse; the student in an unadapted educational system bears credential devaluation. These costs are borne by specific people in specific circumstances. They are not offset by the gains of specific other people in specific other circumstances. They require specific institutional responses designed to address them.
Non-fungibility of primary goods. Rawls lists the primary goods—income, wealth, opportunities, powers, the social bases of self-respect—as the basic resources that every rational person wants. But the separateness of persons implies that these goods are not fully fungible with each other. The loss of professional identity, which is a social basis of self-respect, cannot be compensated by income transfers alone. The loss of attentional capacity, which is a prerequisite for exercising the opportunities that AI tools nominally provide, cannot be compensated by better tools. Justice requires addressing each dimension of disadvantage in its own terms.