NOTE-267: Personal Autonomy

Contribution

The entry frames personal autonomy as a puzzle in the metaphysics of agency. Every agent necessarily treats her own practical judgments as authoritative, because acting requires it. Yet her authority can be “a mere formality” (§1). An account of autonomy is therefore an account of which motives are “external” to an agent who is both governor and governed. The entry surveys four families of answer (§2). It argues that none of them captures minimal self-government, the threshold for being accountable (§3). It then sets agent-causal incompatibilism against the neo-Kantian “practical point of view” (§4), and concludes that the self-relation “is extremely difficult to pin down” (§5). It is an argument, not a neutral survey: its thesis is that ideal self-government and threshold self-government come apart, and that the threshold keeps collapsing into the conditions of agency itself.

Key insight

Theories of autonomy spell out ideals that agents realise by degrees: coherence, responsiveness to reasons, good reasoning, or freedom from determination. None of these is the minimal condition under which a motive’s power is the agent’s own. Coherence fails in both directions. A manipulated endorsement is coherent but not autonomous. A weak-willed act, or one against one’s plans or values, is incoherent and still self-governed. Any attempt to locate identification in some attitude invites the question of what makes that attitude the agent’s, which is the regress.

Assumptions

The premises and authorities the entry works from:

• Agency entails self-authority. In forming an intention one must treat one’s own judgment as authoritative, so the authority is “inalienable” (§1). This is asserted, and the whole entry is built on it.
• Autonomy and accountability share a threshold. The minimal conditions of self-governing action are the conditions an agent must satisfy to be accountable for the motivating force of her states (n. 16). The entry assumes “that only autonomous actions can be attributed to those who perform them” in a special sense (n. 22). It sets aside the Strawsonian view that responsibility has no metaphysical condition (nn. 16–17).
• Paradigm threats are fixed points. Brainwashing and addiction are “the favorite examples” (§1). The unwilling addict who is not even minimally self-governing is taken as data (§3).
• Weakness of will is genuine (§3). The argument that coherence is not necessary is conditional on this (“if weakness of will is a genuine phenomenon”).

Key results

What the entry argues, section by section.

• §1 The problem. An agent defers to herself whenever she acts. So how can a motive’s power fail to be “a manifestation of her power to act”? There is no second, independently identifiable standpoint inside one psyche, unlike the political case of an army dictating to a president.
• §2 Four families.

  • Coherentist. An agent governs her action iff her motive coheres with a mental state representing her point of view. That state is variously identified with:

    • “her highest-order desires regarding which of her first-order desires moves her to act” (Frankfurt 1988c, i.e. the 1971 paper);
    • her evaluative judgments about which actions are most worth performing (Watson 1975; n. 2 says Watson “explicitly allies himself with Plato”, whereas hierarchical accounts are Humean);
    • harmony with long-term plans (Bratman 1979, 2007);
    • “caring” (Frankfurt 1988f, 1999d; Jaworska; Shoemaker 2003);
    • character (R. Dworkin);
    • integration (Arpaly & Schroeder).

    These accounts also tie synchronic unity to diachronic unity. Strict coherentism is “doubly internalist”: an addict who has no objection to her addiction is autonomous, and neither the origin nor the content of the higher-order attitudes matters. Strict coherentists hold that one can act autonomously from desires “so integral to one’s identity that one could not possibly will to resist them” (n. 8, Frankfurt’s “volitional necessity”).

  • Reasons-responsive (Fischer & Ravizza, Nelkin, Wolf): one does not govern oneself unless one’s motives are responsive to a sufficiently wide range of reasons (n. 4 gives Wolf’s “the True and the Good” and Fischer & Ravizza’s “regular reasons-receptivity”).

  • Responsiveness-to-reasoning (Christman 1991, 1993; Mele 1993, 1995): self-government is the capacity to evaluate one’s motives in light of one’s other attitudes and to adjust them. This family carries a weak externalism, because reasoning is norm-governed. It also invites historical conditions against indoctrination. Relational accounts (Benson; Mackenzie & Stoljar; Anderson & Honneth; n. 10: Oshana; Anderson 2003) add that others can block the self-respect or self-trust a point of view of one’s own needs. Such damage can also come without anyone else’s action.

  • Incompatibilist (Pereboom; Kane 1996; van Inwagen 1983): if our motives are determined by causes independent of us, we do not govern them. Velleman’s theoretical-reasoning view of intention is noted as a variant (n. 12).

• §3 Against each as a minimal condition.
  • (a) Manipulation. A brainwashed agent can have coherent endorsements that are even highly reasons-responsive. Historical supplements have trouble distinguishing autonomy-conferring reasoning from autonomy-undermining reasoning “without implicitly appealing to the very phenomenon one is trying to explain”.
  • (b) Coherence is not necessary. The weak-willed agent “identifies with” her motives in whatever way accountability requires. Frankfurt’s adoption example (n. 14) shows that someone can defy her own contemporaneous normative verdict without defying anything deep about herself. Human reflection can distance itself “from every aspect of their own psyches—even their rational reflections”. So identification “cannot be cashed out in terms of higher-order attitudes of approval and disapproval”. The same holds for plans.
  • (c) The regress. If identification is something the agent secures, then we must ask under what conditions she identifies with the motives that move her to identify. To stop the regress there must be an attitude from which “no agent can possibly be alienated”. No highest-order desire, evaluative judgment, plan or combination qualifies. The only candidate is “the desire to be a self-governing agent” (n. 15: Korsgaard for, Velleman’s reservations). It fails as a criterion because it would be causally at work even when the agent fails to govern herself.
  • (d) Reasons-responsiveness is not necessary. Othello authors his act even if jealousy made him unable to track the reasons. Denying this requires incompatibilism.
  • (e) Threshold collapse. Coherentist and reasoning accounts make the unwilling addict a “passive bystander”. That assimilates her to the Tourette’s case, where the behaviour is not action at all. So they fail to explain intentional action that is not accountable. Buss 2012 (n. 18) offers a fifth, normative account, keyed to “identity as a representative human being”.
• §4 Agents as causes. Agent causation (Chisholm, Clarke, O’Connor, R. Taylor) is obscure. The neo-Kantian reply (Korsgaard, Bok; n. 21: McKay, Hampshire, Velleman’s “epistemic freedom”, Sartre) is that from the practical point of view no fact can settle what we have reason to do, so our decisions are “up to us” even under determinism.
• §5. The self-relation remains hard to pin down. Note 22 lists the attributability / answerability / accountability distinctions since Watson 1996 and Shoemaker 2012.

What the entry attributes to the paywalled originals

All attributions are the entry’s own; section or note numbers are given.

• Frankfurt 1971 (“Freedom of the Will and the Concept of a Person”, cited as 1988c, reprint pp. 11–25).
  • §2: a point of view constituted by “highest-order desires regarding which of her first-order desires moves her to act”.
  • Nothing else. There are no second-order volitions, wantons, persons, freedom of will vs. action, or 1971 identification language.
• Frankfurt 1976 (“Identification and Externality”, 1988e). Listed in the bibliography, never cited.
• Frankfurt 1987 (“Identification and Wholeheartedness”, 1988g). Never cited by key. Note 1 says Frankfurt answered regress worries “first” with “decisive commitments”, with no citation.
• Frankfurt 1992 (“The Faintest Passion”, 1999b). Note 1: “more recently”, “satisfaction” with one’s motives, “an absence of ‘restlessness or resistance’ in response to their motivating force”. No page is given.
• Frankfurt 1994 (“Autonomy, Necessity, and Love”, 1999c). Listed in the bibliography, never cited. Love appears only in note 14, via the 2002 replies: a parent who judged it undesirable to seek his children’s well-being would continue to love them, and “this discrepancy… would not show that I had become alienated from the desire”.
• Other Frankfurt (not paywalled originals in this batch’s list, but relevant).
  • Caring: 1982 “The Importance of What We Care About” (1988f) and “On Caring” (1999d), §2.
  • Volitional necessity: note 8. The experience of being unable to will otherwise is compatible with self-government and is “the best indication of where one stands”.
  • The 2002 replies: note 14. Importance “may simply be a brute fact, which is not derived from any assessment”. This is probably the reply to Watson; see corrections.
• Watson 1975 (“Free Agency”).
  • §2: the agent’s point of view is her “(contemporaneous or long-term) evaluative judgments regarding which actions are (most) worth performing”.
  • Note 2: a Platonic rather than Humean alliance.
  • Nothing else. The entry gives no authority objection, no regress objection, and no “valuing vs. desiring” phrasing. Note 14 points to Watson 2002, “Volitional Necessities”, for “a thorough discussion of Frankfurt’s position”.
• Bratman 1996, 2004. Listed only. Bratman is represented by the plan-harmony view, cited to 1979 and 2007, in §2.

Claims

id	claim	strength	support
C1	Every agent necessarily treats her own judgment about how to act as authoritative	assertion	§1; stated as following from what it is to initiate action; no citation
C2	Coherentist accounts are “doubly internalist”: origin and content of higher-order attitudes are irrelevant, so a willing addict is autonomous	informal argument (a characterization)	§2; attributed to the family as a whole, no page citations to Frankfurt
C3	Coherence is not sufficient for minimal autonomy, because endorsements can be manipulated	informal argument	§3; the brainwashing possibility is asserted, no case worked out
C4	Historical supplements cannot distinguish good from bad reasoning histories without circularity	assertion	§3, one sentence (“very difficult”); no argument or citation
C5	Coherence is not necessary: weak-willed and out-of-character action is still self-governed	informal argument	§3; conditional on weakness of will being genuine; Frankfurt’s adoption case (n. 14)
C6	Identification cannot be cashed out as higher-order approval, because reflection can distance itself from any attitude	informal argument	§3, from “human experience”; no citation
C7	The regress can stop only at an attitude no agent can be alienated from, and the only candidate (the desire to be self-governing) cannot discriminate success from failure	informal argument	§3; n. 15–16; it is the entry’s own argument, not attributed to Watson or anyone
C8	Frankfurt answered regress worries first with “decisive commitments”, then with “satisfaction”, the absence of “restlessness or resistance”	second-hand report	n. 1, citing Frankfurt 1999b; no page
C9	Watson 1975 locates the agent’s standpoint in evaluative judgments and is Platonic, not Humean	second-hand report	§2, n. 2; no page
C10	Reasons-responsiveness is not necessary for authorship (Othello) unless incompatibilism is accepted	informal argument	§3, one case
C11	Coherentist and reasoning accounts cannot explain unaccountable intentional action (they assimilate addict to Tourette’s)	informal argument	§3

Concepts

• Coherentist account. The agent governs her action iff her motive “coheres with (is in harmony with) some mental state that represents her point of view on the action” (§2). This is broader than usual. It includes Frankfurt’s hierarchy, Watson’s evaluative judgments, Bratman’s plans, caring and character. The IEP entry uses “coherentist” for Ekstrom and Bratman as opposed to hierarchical accounts.
• Strict coherentism / doubly internalist. Neither the causal history nor the reality-relation of the autonomy-conferring attitudes matters (§2).
• Reasons-responsive vs. responsiveness-to-reasoning. Responsiveness to the reasons there are, which is externalist, versus the capacity to evaluate and revise motives by one’s own reasoning, which is weakly externalist (§2).
• Identification. “Perhaps the most important metaphor in the literature” (§3). It is the relation in which a motive’s power counts as the agent’s own. The entry argues that it cannot be reduced to higher-order approval.
• Volitional necessity. Frankfurt’s term (n. 8) for being unable to will otherwise as a mark of where one stands.
• Satisfaction. Per n. 1, Frankfurt’s later criterion: no “restlessness or resistance” toward one’s motives.
• Minimal vs. ideal self-government. This is the entry’s organising distinction (§3).

Connections

• Within this batch. The IEP “Autonomy” entry (Dryden) is the complement. It states the hierarchy with second-order volitions and gives the regress and ab initio problems under those names. It attributes the regress to Watson 1975, which this entry does not do. It follows Frankfurt through “Identification and Wholeheartedness” but not to satisfaction. The SEP PAP entry (Robb) cites Frankfurt 1971 for the willing addict. It treats identification as a compatibilist account of responsibility (§5.1), the same move this entry calls strict coherentism. Velleman’s “Identification and Identity” (2002, in Contours of Agency, which Buss co-edited) is the detailed treatment of the 1976–1992 sequence that note 1 compresses into one sentence. Frankfurt’s Tanner Lectures are the only place in the batch where his own late view (wholeheartedness, satisfaction, love) can be read first-hand.
• Constitutivism (LIT-117). Note 15’s candidate regress-stopper, “the desire to be a self-governing agent” as the constitutive aim of agency (Korsgaard for, Velleman with reservations), is the constitutivist move that Enoch’s “Agency, Shmagency” attacks. LIT-117 is Deferred and unread, so the link is to its title and abstract only.
• Adaptive preferences and manipulation (LIT-202). Pettigrew’s criterion (space and resources for free agency in forming a preference) is a historical, mechanism-based condition. It is exactly the kind of supplement that §3 says faces a circularity worry. Per NOTE-144 it also uses a Frankfurt-style actualist point: how a preference was actually formed counts, not how it would otherwise have been formed.
• AI and free will (LIT-197, Deferred). From its abstract, it argues that AI undermines free will through manipulation and a lack of reasons-responsiveness. §2 of this entry supplies the distinction the abstract seems to run together: reasons-responsive autonomy versus incompatibilist free will.
• Self-determination theory (LIT-560, LIT-558). In this entry’s classes, SDT’s autonomy is coherentist in its 2006 criterion (reflective endorsement “by the whole self”, LIT-560, p. 1560), historical through internalization (a value taken in under control is introjected, under autonomy support integrated; THEORY-047), and substantive, since some values are said to be “not integrateable” (LIT-558, p. 246). Its integrated regulation is the empirical neighbour of the coherentist “integration” account this entry attributes to Arpaly and Schroeder, and it grades autonomy where that account makes it a threshold (THEORY-054).
• Anthology of the SOTA. No ANTH document mentions autonomy, Frankfurt or free will.

Bearing on the record

• No THEORY in the record bears on this. If the record acquires a THEORY on hierarchical agency, this entry is the source for the objections: C3, C5, C6 and C7, especially the claim that identification cannot be higher-order approval. It is not a source for what Frankfurt 1971 says.
• No ML-practice content. The entry says nothing about machines. The one distant bearing is conceptual: “autonomy” as used of AI agents is usually the political or operational sense this entry sets aside in §1. Nothing belongs in the Anthology.
• For filing. Tags: agency first (self-government, identification); then metaphysics (§4 determinism, agent causation, incompatibilism); then ethics (accountability, moral responsibility, nn. 16–17, 22). identity is not justified. Identity appears only in passing: diachronic unity in §2, and Buss 2012’s “identity as a representative human being” in note 18.

Note of 2026-10-16: THEORY-029, THEORY-040 and THEORY-120 now cite LIT-296 as a source.

Limitations

• Frankfurt’s 1971 scheme is compressed into one clause, and his three decades of revision into one footnote. A reader cannot recover second-order volitions, wantons, persons, freedom of will vs. action, decisive identification or wholeheartedness from this entry.
• Watson’s actual objection to Frankfurt is absent. Watson appears as a rival coherentist, not as a critic.
• The §3 arguments are informal and rest on contested intuitions (weakness of will; the adoption case; Othello). The historical-account circularity charge (C4) is one sentence.
• Bratman 1996 and 2004 and Frankfurt 1976, 1987 and 1994 are listed but not discussed. Two note-14 citations appear mis-keyed (see corrections).
• The 2018 revision’s division of labour between Buss and Westlund is not stated. Westlund’s own relational work (2003) is listed but not discussed. The relational material is thin: one paragraph in §2 and note 10.

Open questions

• Is there an attitude from which no agent can be alienated? The entry’s negative answer (C7) is what Frankfurt’s “satisfaction” and Bratman’s self-governing policies are meant to dispute. Neither reply is evaluated here.
• Can a historical condition on autonomy avoid the circularity worry (C4)? Christman’s counterfactual non-repudiation test (n. 6) is the candidate. The entry does not assess it.
• Do minimal autonomy and accountability share a threshold (n. 16), or does the attributability / answerability / accountability split (n. 22) dissolve the §3 puzzle?

Corrections to the seeded skim

• none (there was no seed or dossier)
• Authorship differs by revision. The scoping report’s “Sarah Buss & Andrea Westlund, first published 2002-05-28; substantive revision 2018-02-15” is correct for the 2018 revision only. The 2002 first version and the 2013 revision are by Buss alone. Cite it as Buss & Westlund, revision of 15 Feb 2018.
• The scoping report overstates the Frankfurt/Watson content. It reports “Frankfurt” ×18 and “Watson” ×15 and calls the entry the “best synthetic overview of hierarchical, identification-based autonomy”, “including Watson’s objection and the regress”. Most of those hits are bibliography entries. In the body, Frankfurt is cited three times (§2: 1988c, 1988f, 1999d) and in the notes about six times. Watson 1975 is cited once in §2, plus note 2. The entry never presents Watson as objecting to Frankfurt. It files Watson 1975 as a rival coherentist account, on which the agent’s point of view is her evaluative judgments. The regress in §3 is the entry’s own argument and cites nobody. The words “wanton”, “person” (as Frankfurt’s technical term), “second-order volition” and “freedom of the will” (as distinct from freedom of action) do not occur.
• Mis-keyed citations in note 14. The text cites “Frankfurt 2002b … (163)” and “(161)” for the adoption-mother example and the “brute fact” claim, but the bibliography’s 2002b is “Reply to Michael Bratman”, pp. 86–90. Pages 161 and 163 fall inside 2002d, “Reply to Gary Watson” (pp. 160–64), so 2002d is probably meant. The same note cites “Frankfurt 2002, 223”, and there is no undated-letter 2002 item. Page 223 falls inside 2002f, “Reply to Richard Moran” (pp. 218–25). Both identifications are inferences from page ranges and are unverified.
• Cited but unused, or used but uncited. Bratman 1996 (“Identification, Decision, and Treating as a Reason”) and 2004 (“Three Theories of Self-Governance”) are in the bibliography but are never cited in the text or notes. The same is true of Frankfurt 1976 (“Identification and Externality”, 1988e), 1987 (“Identification and Wholeheartedness”, 1988g) and 1994 (“Autonomy, Necessity, and Love”, 1999c). Note 8’s claim that Frankfurt “coins the term ‘volitional necessity’” carries no citation. (The SEP PAP entry, read in this batch, cites the term to Frankfurt 1982, “The Importance of What We Care About”.)