The Capability Approach

In the AI Story

The approach emerged from Sen's dissatisfaction with the dominant traditions of welfare economics. Utilitarianism measured welfare by preference satisfaction, which Sen demonstrated could be distorted by adaptive preferences — the mechanism by which the deprived cease to desire what they have been denied. Resourcist approaches measured welfare by the distribution of primary goods, which Sen argued failed to account for the variable capacity of different people to convert the same resources into genuine well-being. The capability approach cut through both impasses by insisting that welfare lives not in what people have or what they report feeling, but in what they are substantively free to do and to be.

The operational core of the framework is the distinction between functionings and capabilities. A functioning is an achieved state of being or doing — being educated, being nourished, being creatively engaged. A capability is the real freedom to achieve a functioning, whether or not the functioning is exercised. A person who could pursue creative work but chooses rest has a larger capability set than a person who rests because creative work is unavailable to her. The freedom inheres in the choice, not in the outcome.

The framework's analytical power comes from its attention to conversion factors — the personal, social, and environmental conditions that determine whether a resource translates into a capability. A bicycle is a resource; mobility is the capability. Whether the bicycle becomes mobility depends on whether the person can ride, whether roads exist, whether cultural conditions permit riding. The same resource produces radically different capability expansions in different institutional contexts. AI tools follow the same logic.

The capability approach directly informed the United Nations Human Development Index and continues to shape global policy debates on inequality. Its application to AI represents the most rigorous available instrument for evaluating whether the technology's celebrated power actually converts into expanded human freedom — or whether, as with many previous general-purpose technologies, the power concentrates among those who already possessed the conversion factors to benefit from it.

Origin

Sen developed the approach through a series of Tanner Lectures at Stanford in 1979 ('Equality of What?'), refined it through Development as Freedom (1999), and extended it in The Idea of Justice (2009). Martha Nussbaum's parallel development specified a list of central capabilities; Sen deliberately left the selection open to democratic deliberation.

Key Ideas

Means versus ends. Income, productivity, and tool access are means; capabilities are the ends. Means are valuable only insofar as they convert into capabilities.

The evaluative space matters. What a society measures determines what it manages. Choose the wrong space and you will optimize for the wrong thing.

Freedom as constitutive. Substantive freedom is not a consequence of development — it is development's definition.

Democratic selection of capabilities. Which capabilities matter most is a question for public reasoning, not expert determination.

Debates & Critiques

Critics have argued that the approach's openness — its refusal to specify a definitive list of capabilities — makes it operationally underdetermined. Nussbaum's central capabilities list addresses this by providing a specific enumeration, but Sen has defended the openness as a feature, insisting that the selection of relevant capabilities belongs to democratic deliberation rather than philosophical decree.