NOTE-260: The Capability Approach

Contribution

This is the standard philosophical reference for the capability approach as a whole: Sen, Nussbaum and the literature after them.

• Framework, not theory (§1.2). Its organising thesis is Robeyns’s (2017). The capability approach is an open, underspecified framework whose only core commitment is to conceptualise well-being and well-being freedom in terms of functionings and capabilities. Nussbaum’s partial theory of justice, Sen’s comparative theory of justice and “development as freedom”, capabilitarian econometrics and so on are theories built on it by adding further choices.
• What it catalogues.
  • The core concepts (§2): functionings and capabilities, real freedom, conversion factors, means versus ends, diversity and agency, and the interpersonal-comparison metric.
  • The specification choices (§3): purpose, functionings or capabilities, selection and aggregation, and whether capabilities are binary or incremental.
  • What a capability theory of justice would still need (§4). It lists eight items.
• What the 2025 text adds.
  • Non-Western philosophies (§5): Māori, Ubuntu, Buen Vivir.
  • Applications (§6): development, health, disability, environment, animals, education and AI ethics.
  • Robustness of conversion factors (§§2.3, 3.4), following Byskov, Kramm & Östlund 2024.

Key insight

The capability approach answers one question, “in what space should advantage be compared?”, and leaves open everything a full theory of justice or policy would need. Many critiques of “the capability approach” are really critiques of one theory built on it, Nussbaum’s or Sen’s. Many attributions to “the approach” are really optional add-ons, such as sufficiency, egalitarianism, an obligatory list or deliberative democracy. The entry’s most useful move is to keep asking which layer a claim belongs to.

Assumptions

What the entry takes as settled, with its authorities:

• Core definitions.
  • Functionings are achieved “doings and beings”.
  • Capabilities are “the real, or substantive, opportunity” to achieve them.
  • Capability set is the set of capabilities one can choose from.
  • Basic capabilities has two senses: Nussbaum’s innate equipment (2000, 84), or Sen’s freedom to do basic things needed to escape serious deprivation (1992, 45 fn. 19). In Sen’s sense their role is “deciding on a cut-off point for the purpose of assessing poverty and deprivation” (Sen 1987b, 109; §2.1).
• Value-neutral definitions (§2.1). The authors’ own position, following Byskov 2020 and Robeyns 2017, is that capabilities and functionings should be defined value-neutrally at the framework level, so bad capabilities (to kill, to pollute) and neutral ones count as capabilities. Selecting the valuable ones is then a theory-level step. The entry notes that this is a debated position. It is not Sen’s “reason to value” formulation.
• Real freedom means opportunity freedom (§2.2). Capabilities are “freedoms conceived as real, or substantive, opportunities”. The authorities are Sen 1985a, 3–4; 1985d (the Dewey Lectures), 201; 2002 ch. 20; and Kaufman 2006a. Such freedoms are “corrected for potential impediments” (Byskov et al. 2024).
• Ethical, not methodological, individualism (§2.3). The approach is individualist in the ethical sense only. This is Robeyns 2017 §4.6’s answer to the collectivist critique.

Key results

What the entry reports and argues, by section.

• §1.1 Origins.

  • Lineage. Sen 1974, 1979a and 1979b criticise the “limited informational bases” of utilitarianism and resourcism. Aristotle, Smith and Marx are named as antecedents (Nussbaum; Sen 1993a; Sen 1999: 14, 24).
  • The disability example. It is retold from “Equality of What?”. The retelling is garbled (see corrections).
  • Lowered expectations. Sen documents that deprived people “lower their expectations as a coping mechanism” (Kynch & Sen 1983). This is the adaptive-preference point, although the entry never uses that term.

• §1.2 Framework or theory. The approach supports three normative exercises: assessing individual well-being, evaluating social arrangements, and designing policy.

  • Nussbaum’s five principles. Nussbaum 2011a gives five shared principles: each person as an end, choice and freedom over achievement, value pluralism, concern with entrenched injustice, and an urgent task for government.
  • Robeyns’s reply. Robeyns 2016 answers that only “each person as an end” and value pluralism are essential.

• §2.3 Conversion factors.

  • Definition. A conversion factor is “the degree in which a person can transform a resource into a functioning” (Sen 1992, 19–21, 26–30, 37–38).
  • The three types. Personal factors are metabolism, physical condition, sex, skills and intelligence. Social factors are policies, norms, discrimination, and hierarchies of class, gender, race and caste. Environmental factors are climate, pollution and infrastructure.
  • Scope in Sen. On the “most plausible reading” of Sen 1985a, conversion concerns tangible resources. Sen 1997 extends it to human capital.
  • Robustness. The same capability is less substantive when it is exclusive (siblings sharing one bike) or when its conversion factors may perish (a war zone).
  • Collective capabilities. The collective-capabilities critique is reported (Evans, Gore, Stewart, Ibrahim), with Robeyns’s reply.

• §2.4 Means and ends. Capabilities are the ends of interpersonal comparison. Resources are means, and good proxies only if everyone converts them equally.

  • Two advantages of starting from ends. Means stay instrumentally valued. And the question of which means matter is left open, so the answer can be institutions or norms rather than income.

• §2.5 Diversity and agency. Human diversity enters in three ways: through plural evaluative spaces, through conversion factors, and through agency.

  • The four-way distinction. Sen distinguishes well-being freedom from agency freedom, and each from its achieved counterpart, well-being achievement and agency achievement (Sen 1985d, the Dewey Lectures).
  • Agency freedom. It covers freedoms valued although they do not promote one’s well-being. The illustration, giving up a career to fight for global justice, is the entry’s own and is not attributed to Sen.

• §2.6 The metric.

  • Which space for which comparison. Functionings are the space for comparing achieved well-being. Capabilities are the space for comparing “well-being freedom” (Sen 1992: 40).
  • Sets, not options. The holistic point is that what matters is which combinations are jointly open. The example is a poor single parent who can hold a job or care for her children, but not both.

• §3.2 Functionings or capabilities. The entry gives four considerations.

  • Anti-paternalism (Sen and Nussbaum’s main argument).
  • Responsibility-sensitivity.
  • Conditional capabilities that depend on others’ choices (Basu 1987).
  • People who cannot choose, such as infants and people with dementia.
  • A bridging concept. Sen’s “refined functioning” (1987b: 36–7) bridges the two: “‘fasting’ as a functioning is not just starving; it is choosing to starve when one does have other options” (Sen 1992: 52).
  • Measurement. Functionings are easier to measure (Sen 1992: 52–3).

• §3.3 Selection and aggregation. This is the central Sen–Nussbaum divide.

  • Nussbaum. She holds a ten-item list (life; bodily health; bodily integrity; senses, imagination and thought; emotions; practical reason; affiliation; other species; play; control over one’s environment; Nussbaum 2006: 76–78). She justifies it by dignity, grounds it in overlapping consensus, and treats its items as incommensurable, with a threshold for each.
  • Sen. He “consistently and explicitly refuses” a canonical list (Sen 2005a: 158) and appeals to democratic process and public reasoning. But he “has never explained in detail how such a selection could and should be done”. His formula is capabilities “we have reason to value” (1987b; 1992; 2002; 2009). Reasonableness works as an objective check through scrutiny and public reasoning, because people “may be mistaken about their own well-being” (Sen 1985a; 1985c; 1987a).
  • Three readings of “reason to value” (Khader & Kosko 2019): procedural-autonomy, process (individual reflection or public deliberation), and perfectionist.
  • Selection methods. Byskov counts 14 methods in four families: ad hoc, foundational, procedural, and mixed.
  • Weighting. Proposals include weights set by a social-choice procedure (Chakraborty 1996) and life-satisfaction weights (Schokkaert 2007). Aggregation need not be additive, because capabilities can be complements.

• §3.4 Binary or incremental. On the binary view, a capability is had only if jointly sufficient conditions hold. On the incremental view, it is had to the degree that conditions are met (Byskov et al. 2024).

• §4 Justice. Reading the approach as a theory of justice “is mistaken”. The entry gives eight requirements a capability theory of justice must still meet:

  1. a justification;
  2. functionings or capabilities as the metric;
  3. selection, quantification and aggregation;
  4. a comparison with other metrics, namely Dworkinian resources and Rawls’s primary goods;
  5. a distributive rule;
  6. who bears responsibility and duties;
  7. procedural and democratic justice;
  8. intergenerational justice.

  On specific points:

  • Rawls. Rawlsians (Freeman, Kelly, Pogge, Richardson) argue that capability theorists overstate Sen’s critique, and “not all of these critiques have been sufficiently rebutted”.
  • Consequentialism. Sen’s target is utility-maximisation, not outcome-focused morality as such (Sen 1979c; Qizilbash 2022). Others read the approach as non-consequentialist.
  • Distributive rule. Sen “has remained uncommitted to one single distributive rule”. “Equality of What?” asked only “if equality, then of what?”.
  • Democracy. Sen does not give a full account of democratic justice (Srinivasan 2007).

• §§5–6. A survey of the literature on non-Western philosophies and of applications. It includes Nussbaum’s extension to animals (2011b, 2023), against the Rawlsian view that animals are owed compassion but not justice.

Claims

id	claim	strength	support
C1	The capability approach is a framework whose core is conceptualising well-being in terms of functionings and capabilities, not a theory of well-being or justice	moderate	informal argument (§1.2), resting on Robeyns 2017 and Qizilbash 2012; Nussbaum 2011a’s narrower characterisation is reported as the rival view
C2	Sen distinguishes well-being freedom and achievement from agency freedom and achievement	moderate (as a report)	citation of Sen 1985d (Dewey Lectures), not quoted; the illustrative example is the authors’ own (§2.5)
C3	Sen refuses “one pre-determined canonical list of capabilities, chosen by theorists without any general social discussion or public reasoning”	strong (as a report)	direct quotation, Sen 2005a p. 158 (§3.3); not checked against the paywalled source
C4	Sen has never explained in detail how selection by public reasoning should be done	moderate	the authors’ reading of Sen’s corpus (§3.3), echoed by Srinivasan 2007 on democracy (§4)
C5	Capabilities should be defined value-neutrally at the framework level	weak–moderate	the authors’ own position (Byskov 2020; Robeyns 2017), argued by counter-examples such as a domestic-violence study (§2.1); it departs from Sen’s “reason to value”
C6	Sen 1979 argues utilitarianism would favour a disabled “pleasure wizard” over a hard-to-please able-bodied person	incorrect as stated	§§1.1, 2.6; contradicted by Sen 1979 pp. 203–204 and 217–218 (see corrections)
C7	Sen is uncommitted to any single distributive rule	moderate	textual reading of Sen 1979a and 2009 (§4); consistent with Sen 1979 pp. 219–220 (“partial guide”; the index “can be used … in many different ways”)
C8	Rawlsian replies (Rawls accommodates special needs) have not all been rebutted	weak	assertion, with citations to Freeman 2006, Kelly 2010, Pogge 2002 and Richardson 2006 (§4); no argument given
C9	Sen’s critique of utilitarianism targets utility-maximisation, not consequentialism	weak–moderate	citation of Sen 1979c and Qizilbash 2022 (§4); the entry reports it as contested

Concepts

• functioning: an achieved being or doing. Functionings are “constitutive of a person’s being” (Sen 1992: 39).
• capability: the real or substantive opportunity to achieve a functioning, corrected for impediments. The entry defines it value-neutrally.
• capability set: the set of capabilities (combinations of potential functionings) one can choose from.
• basic capabilities: (a) Nussbaum’s innate potentials, or (b) in Sen’s sense, the freedom to do the basic things needed to avoid serious deprivation, used for poverty cut-offs.
• conversion factor: the degree to which a person can turn a resource into a functioning. The types are personal, social and environmental.
• well-being freedom and achievement; agency freedom and achievement: Sen’s two-by-two, from the Dewey Lectures (reported, §2.5).
• refined functioning: a functioning that registers the alternatives that were available (Sen 1987b).
• capabilitarianism: Robeyns’s (2016) term for the family of capability theories.
• binary and incremental capability: Byskov et al. 2024’s two views on the threshold for having a capability (§3.4).

Connections

• The other secondary sources in this batch.
  • The IEP entry. The IEP entry “Sen’s Capability Approach” (Wells) covers the same Sen material from a Sen-centred angle. It disagrees with this entry on several points (see the sen5 reading).
  • Robeyns 2017. Robeyns’s Wellbeing, Freedom and Social Justice (2017) is this entry’s source for the framework/theory split, for the ethical-versus-methodological individualism reply, and for the necessary/optional specification scheme. The entry is in effect a digest of that book.
  • Nussbaum 2003. Nussbaum’s “Capabilities as Fundamental Entitlements” (2003) is cited as one of those demanding that morally bad capabilities be excluded (§3.3). It is the other pole of the list debate.
  • Valentini and Zwolinski. Valentini’s critique of Sen (2011) and Zwolinski’s review of The Idea of Justice bear on §4’s report of Sen’s comparative, non-transcendental stance (Sen 2006, 2008, 2009). The entry states that stance and does not assess it.
• The primary texts in this batch. Sen’s “Equality of What?” (1979) is the source the entry paraphrases in §§1.1, 2.6 and 4. It misstates the argument’s structure (corrections). Sen’s Nobel lecture (1998) and “The Impossibility of a Paretian Liberal” (1970) are not cited. The entry’s only social-choice content is the suggestion that weights might be set by a social-choice procedure (§3.3).
• In this record.
  • LIT-202. Pettigrew’s “Consent and the Formation of Preferences” takes up the adaptive-preference theme that the entry mentions only through Kynch & Sen 1983 (§1.1).
  • LIT-146. The SEP “Moral Status of Animals” treats Nussbaum’s capabilities as an objective-list value theory for animals. That meets §6’s report of Nussbaum’s animal-justice extension.

Bearing on the record

This is the definitional anchor for any THEORY or concept document on capabilities in nucleation. Its framework-versus-theory distinction is the right way to file claims: a claim about “the capability approach” should say which layer it belongs to.

What the entry says about Sen’s paywalled texts. For those texts it is a second-hand source. Here is every claim it makes about them, by section:

• “Well-being, Agency and Freedom: The Dewey Lectures 1984” (1985d).
  • §2.2. Capabilities are opportunity freedoms (cited “1985d, 201”, together with 1985a 3–4 and 2002 ch. 20).
  • §2.5. Sen distinguishes well-being freedom from agency freedom, each with an achievement counterpart. Agency freedom covers valued freedoms that need not promote one’s well-being. No quotation is given.
• “Capability and Well-Being” (1993a). It is cited only in §1.1, for tracing the approach to Aristotle, Smith and Marx. There is no definitional claim.
• “Capabilities, Lists and Public Reasons: Continuing the Conversation” (2004a) and “Elements of a Theory of Human Rights” (2004b). Both are cited together, only in §3.3, as critiques that Nussbaum’s list leaves “insufficient scope for democratic deliberation and respect for agency”. Whether 2004b (“Elements”) actually makes that criticism of Nussbaum is unverified. The citation may be imprecise. Nothing in the entry reports the content of Sen’s theory of human rights.
• “Human Rights and Capabilities” (2005a).
  • The quotation. §3.3 quotes it: Sen refuses “one pre-determined canonical list of capabilities, chosen by theorists without any general social discussion or public reasoning” (p. 158).
  • The motive. §3.3 adds that Sen withholds a substantive list to avoid paternalism and the imposition of a conception of the good, “holding that communities should decide for themselves”.
  • What is missing. The paper’s thesis about rights as capabilities, and about the process freedoms that capabilities cannot capture, is not reported anywhere.
• Other paywalled Sen texts cited.
  • Commodities and Capabilities (1985a). Conversion of tangible resources (§2.3), means (§2.4), and people mistaken about their own well-being (§3.3).
  • Inequality Reexamined (1992). Pp. 19–21, 26–30, 36–40, 45 fn. 19 and 52–53, on conversion factors, constitutive functionings, well-being freedom, basic capabilities, fasting and measurability.
  • The Standard of Living (1987b). Pp. 36–37 and 109, on refined functioning and the poverty cut-off.
  • “Utilitarianism and Welfarism” (1979c). §4, on consequentialism.
  • “Rights and Capabilities” (1985b). §4, on metrics.
  • Development as Freedom (1999). Pp. 152–153, on public deliberation (§3.3), and democracy in selection (§4).
• Not in the entry at all. Human rights as a Sen topic, beyond the bibliography; Sen’s social-choice work (Arrow, impossibility, the liberal paradox); and incompleteness or partial orderings.

Nothing here bears on ML practice. The §6 mention of capability-approach work on AI ethics and surveillance (Oosterlaken 2011, 2015) is a pointer to literature, not an instruction, so there is nothing for the Anthology of the SOTA.

Limitations

• A particular angle. The entry is written from inside the Robeyns–Byskov research programme. Its framework/theory scheme, its value-neutral definition and its binary/incremental robustness analysis are the authors’ own positions. It presents them as the reference view, with rivals noted.
• A thin account of Sen. On Sen in particular it is thin: little on agency beyond one paragraph, nothing on social choice, nothing on incompleteness, and nothing on human rights. Its paraphrase of Sen 1979 is wrong in detail.
• Critiques named, not stated. The internal critiques (Pogge, Anderson, the Rawlsians) are named with citations but not argued. The reader cannot evaluate them from the entry.
• No quotations from the Dewey Lectures. Every Dewey-Lectures claim is a paraphrase.

Open questions

• Does Sen’s “reason to value” survive the value-neutral redefinition? The entry endorses value neutrality at the framework level. Sen’s own texts define capabilities through what one has reason to value. Settling this would take Sen’s 1993 and 2004 texts, which are paywalled.
• Which of the three readings is Sen’s? Is his public-reasoning stance procedural, process-based or perfectionist (Khader & Kosko)? Only the primary texts can settle it.
• Do the Rawlsian replies succeed? The entry says the replies “have not all been sufficiently rebutted”. Whether they succeed needs Pogge 2002 and Freeman 2006 read directly.

Corrections to the seeded skim

• none (there was no seed or dossier)
• Authorship differs by revision. The scoping report gives “I. Robeyns & M. F. Byskov” with no qualification. That is correct only for the 2020 and 2025 revisions. The 2011 first publication, and the 2016 revision, are by Robeyns alone (SEP archive editions, sum2011 and win2016). Cite it as Robeyns & Byskov, revision of 17 Apr 2025.
• The scope report says the SEP covers Pogge “×5” as a critic. The entry does not present Pogge’s argument. Pogge 2002 is cited three times: as someone who says justice must demarcate relevant from irrelevant and bad capabilities (§3.3); as one of those showing the Rawls–capability comparison is complicated (§4); and as one of the Rawlsians who say Rawls does accommodate special needs (§4). The other two hits are the bibliography (Pogge 2002 and Pogge 2008). His argument itself is never stated. For Pogge, Robeyns 2017 ch. 4 has to carry the weight.
• The entry misreports Sen 1979.
  • The pleasure wizard. §1.1 says “The disabled person may be a pleasure wizard who gains a lot of utility from very little input”, and §2.6 repeats that Sen says the disabled person may be a ‘pleasure-wizard’. In “Equality of What?” the pleasure-wizard is the other person, B. The cripple A “gets half the utility that the pleasure-wizard person B does from any given level of income” (p. 203).
  • Two arguments run together. §1.1 attributes to utilitarianism an argument that Sen aims at total-utility equality and leximin. That argument is the contented cripple with a “jolly disposition” or “low aspiration level”, who then has no total-utility claim (pp. 217–218). Sen’s argument against utilitarianism is different: sum-maximisation gives the less efficient utility-converter less income (pp. 203–204).
  • “Doubly worse off”. Sen uses the phrase for the utilitarian outcome: less utility from the same income, and less income (p. 203). §1.1 reuses it for a person with equal resources and extra disability expenses.
  • What survives. The conclusion the entry draws, that neither resources nor utility track the disabled person’s claim, is Sen’s (pp. 217–218). The route it gives to that conclusion is not his.
• Quotations and page references I could check are accurate. The §4 quotations from Sen 1979a pp. 215–216 (“seems to take little note of the diversity of human beings”, “very different needs”, “overlooking very widespread and real differences”) are verbatim on those pages. The bibliography’s page range “pp. 197–220” is the text proper; the Tanner offprint runs 195–220 with front matter.
• The entry never mentions incompleteness or partial orderings. Sen already raises them in 1979. His indexing of basic capability bundles yields a “partial ordering” that “must be supplemented by certain established conventions of relative importance”, with culture-dependent weights (p. 219). The entry discusses weighting and aggregation (§3.3) but not incomplete rankings, as Sen’s or as anyone’s.
• Minor bibliography defects.
  • Missing or garbled entries. The text cites Qizilbash 2022 twice (§4); it is absent from the bibliography. The Robeyns “Nancy Fraser” entry has no year (the text calls it 2003a).
  • Year mismatches. Hailwood is cited as 2021 in §6 but listed as 2012. Fukuda-Parr & Kumar are cited as 2003 in §6 but listed as 2009 (only the Handbook) and 2003 (as editors of ul Haq).
  • Typos. Singer’s title is misspelled “The Life You Can Safe”.