Agency

Type: Node / Situated, Target-Relative Capacity Profile

Placement note: Agency is the keystone of Division III's second concept: not who can be a bearer of action (Agent), but what a bearer's situated capacity to act actually is, regarding a specific matter, in a specific context, at a specific time. Division III is becoming an agentive-relational grammar: who can act (Agent, Agency), how they participate (Role, and Participation once admitted), and how persons acknowledge one another (Interpersonal Recognition, once admitted), with Dignity constraining treatment regardless of Agency, Participation, Role, or Recognition.

Working Definition: Agency is the situated capacity of an Agent to form, revise, sustain, refrain from, refuse, enact, and communicate intentions regarding a specific matter, attributable to that Agent, within a stated context and at a stated time. It is assessed dimension by dimension and target by target, not as a single scalar: an agent's capacity to enact one matter can be diminished while their capacity to enact a different matter, in the same context and at the same moment, remains intact.

Function in the Grammar: Agent supplies the bearer: the primitive from which actions, claims, and commitments originate. Intention supplies the orientation: what an agent aims at, regarding some matter. Action supplies the doing: what an agent performs. None of these names a further, separate fact this grammar needs: whether, and how, an agent's own capacity to move between orientation and doing, regarding this specific matter, is currently available to them. An agent who intended to act and did not may have abandoned the intention, been obstructed, or been fully capable and simply chosen otherwise; or their capacity to sustain or enact that intention, regarding that matter specifically, may itself have been diminished, by exhaustion, compulsion, illness, trauma, or circumstance, without their ceasing to be an Agent, and often without their Intention having changed at all. Agency names that capacity and keeps it visible as its own fact, separate from whether an Action ultimately occurred.

Agency is relative to a matter, not only to a context, and that relativity extends all the way up. The same agent, in the same context, at the same moment, can have a diminished capacity to enact one intention and an intact capacity to enact a different one. A single scalar per dimension cannot express this; collapsing across targets is a second, subtler version of the flattening this entry exists to prevent, alongside collapsing across dimensions. It also means an aggregate judgment like "constrained" cannot honestly describe an entire context: a diminished capacity regarding one matter says nothing, by itself, about a different matter assessed in the same context at the same time, including a Commitment regarding that different matter. Every capacity judgment this entry makes is therefore stated regarding a specific matter, alongside its context and time; see Formal Grammar Representation.

Capacity is also not uniform across dimensions. A person recovering from a craving may retain full capacity to refrain from acting on it and to communicate what is happening, while their capacity to sustain or enact, regarding the specific intention to call a sponsor, is diminished for the evening. A person recovering from surgery may retain full capacity across every dimension regarding the decision to consent to or decline treatment, while their bodily capability for direct physical tasks is separately and severely limited, a fact about capability and environmental access, not about Agency; see Distinctions.

Recruiting another's help is not agency-free, and it does not bypass enactment. Texting a peer to ask for a check-in is itself an Action, drawing on the agent's own capacity to enact. Ordinarily this is a personal-mode exercise, the agent acting alone to send the text, distinct from a proxy-mode exercise in which someone else acts on the agent's behalf, and regarding a different matter, the text itself, than the one that prompted it, the sponsor call. A person who cannot readily enact one matter unaided may still readily enact an adjacent matter that recruits assistance toward it; that is agency exercised through an alternative action, not agency exercised without any action at all.

The chain this entry supports is: an Agent, a stated matter, a Context, a time, and a capacity profile across dimensions together constitute a situated Agency Claim regarding that matter. That Claim, like any other, may be supported by Evidence under an applicable Warrant. A supported Agency Claim, together with whatever Standard and Procedure the case actually invokes, may bear on a further consequence, a Commitment's validity, a consent's validity, effect, or currency, or a Decision, but no step in that chain licenses the next automatically.

Formal Grammar Representation

CAPACITY(
  agent: a,
  regarding: g,
  context: Context(k),
  at: t,
  dimension: D,
  assessment: <dimension-assessment>
)
  -- one targeted, assessed (or explicitly unassessed) capacity
     judgment; g identifies the specific matter, intention, decision,
     or act the judgment is about, not merely the surrounding Context

<capacity-dimension> ::= form | revise | sustain | refrain | refuse
                        | enact | communicate

<dimension-assessment> ::=
    Assessed(level: <capacity-level>, relative_to: r)
  | NotAssessed

<capacity-level> ::= intact | diminished | absent
  -- these three describe the alleged state of the capacity itself.
     NotAssessed is not a fourth level; it describes the epistemic
     state of the record, not a state of the agent, and must be kept
     out of the same enumeration

r ::= prior_functioning | stated_population_baseline
  -- relative_to states only the comparison baseline for the level
     itself. Whether an assessed level suffices for a particular task
     is a separate, further question; see SUFFICIENT_FOR

Profile(a, regarding: g, context: k, at: t, relevant: R) :=
  the set of CAPACITY(a, g, k, t, D, ...) judgments recorded for
  agent a, regarding matter g, in context k, at time t, restricted to
  dimensions D in R
  -- relevant: R states which dimensions actually bear on g; not
     every dimension is relevant to every matter

PORTFOLIO(a, k, t) :=
  the set of all Profile(a, g, k, t, R) recorded for agent a, in
  context k, at time t, across every matter g actually assessed
  -- a Portfolio is a collection of Profiles, not itself a single
     Agency judgment; only a Profile, regarding one matter, carries a
     CONSTRAINED, UNCONSTRAINED, or UNDETERMINED status

CONSTRAINED(a, g, k, t, R) iff
  some D in R has a recorded CAPACITY(a, g, k, t, D, Assessed(level, r))
    with level diminished or absent

UNCONSTRAINED(a, g, k, t, R) iff
  every D in R has a recorded CAPACITY(a, g, k, t, D, Assessed(intact, r))

UNDETERMINED(a, g, k, t, R) iff
  neither CONSTRAINED(a, g, k, t, R) nor UNCONSTRAINED(a, g, k, t, R)
  holds: no D in R is recorded diminished or absent, but some D in R
  is NotAssessed or unrecorded
  -- an incomplete profile must not be treated as an unconstrained
     one; without this third outcome, silence about a dimension could
     pass the CONSTRAINED test and be mistaken for its absence

SUFFICIENT_FOR(profile: Profile(a, g, k, t, R), target: x, standard: s)
  ::= sufficient | insufficient | undetermined
  -- whether an assessed Profile, regarding g, meets the threshold a
     specific Standard s sets for a specific target x. This is the
     entry's own answer to "is this enough capacity for this task,"
     deliberately downstream of, and separate from, the raw
     comparison-basis judgment CAPACITY itself makes

<agency-mode> ::= personal | proxy | collective
  -- Bandura's three modes of human agency

<support-configuration> ::= unaided | assisted

EXERCISES_AGENCY(
  agent: a,
  regarding: g,
  via: Action(v),
  mode: <agency-mode>,
  support: <support-configuration>
)
  -- records that a exercised agency regarding g, through the
     specific Action v, in a given mode and support configuration.
     Mode and support characterize an actual exercise; they are not
     fields on CAPACITY, since a capacity judgment may be made without
     any exercise having occurred at all

Withdrawing a previously given consent is not modeled as its own
relation here. Giving and withdrawing consent are temporally distinct
acts; withdrawal does not erase the historical fact that consent was
once given. Withdrawing draws on whatever capacity dimension the
withdrawal itself requires, typically refrain, refuse, or communicate.

EXERCISES_AGENCY(a, alpha) does not entail BLAMEWORTHY(a, alpha)
  -- shorthand here for "a exercised agency via alpha, under whatever
     mode and support configuration applied." That an Action was
     agentive, including one exercised through a severely constrained
     pathway, bears on whether it is attributable to a; it does not
     by itself settle whether a should be blamed, praised, or held to
     any particular standard for it

CONSTRAINED(a, g, k, t, R) does not by itself entail:
  INVALID(Commitment(c)) for any Commitment c actually made by a
    regarding g
  -- voluntariness, coercion, an applicable Standard, and the specific
     decision's own scope remain separately governed questions. A
     Commitment refused or never made has no such object whose
     validity there is anything to evaluate

The seven capacity dimensions, three agency modes, and two support configurations above are an analytic decomposition this encyclopedia introduces. They are not presented as an exhaustive psychological taxonomy, a clinically validated assessment instrument, or a claim that the dimensions are empirically independent of one another. "Separately assessable" means only that one dimension's level must not be inferred automatically from another's.

Semantic Constraints

The expression CAPACITY(a, g, k, t, D, assessment) is well-formed when the agent, matter, context, time, and dimension are identifiable, and the assessment is either a stated level with its comparison basis or an explicit NotAssessed. Well-formedness does not establish that the assessment is accurate. Asserting a capacity judgment is a Claim, subject to the same epistemic discipline as any other; see Claim and Warrant.

relative_to states only the baseline a level is compared against, the agent's own prior functioning or a stated population baseline, not whether that level is sufficient for any particular task. Whether an assessed Profile meets a task's demands is a separate question, SUFFICIENT_FOR, governed by whatever Standard the case actually invokes. Folding task sufficiency into relative_to would let a Standard enter the raw capacity assessment before this entry's own discipline says it should.

CONSTRAINED, UNCONSTRAINED, and UNDETERMINED are relative to a stated matter and a stated relevant-dimension set R, not to an agent's entire context. A Profile diminished regarding one matter does not make a different Profile, regarding a different matter in the same context, constrained; only a PORTFOLIO, the collection of an agent's Profiles across matters, spans multiple targets, and a Portfolio does not itself carry a CONSTRAINED status. These three outcomes are also distinct from each other: an incomplete Profile is UNDETERMINED, not UNCONSTRAINED, and must not be treated as evidence that no constraint exists.

A diminished or absent value for one (g, D) pair does not entail diminishment for any other. communicate regarding one matter may be intact while enact regarding that same matter is absent; enact regarding one matter may be diminished while enact regarding a different matter, in the same context and at the same time, remains intact. Sustain and refrain name different capacities though both concern not-yet-enacting: sustain is the capacity to continue holding an intention over an interval without abandoning or forgetting it; refrain is the capacity to intentionally not perform an available, desired, or impulse-prompted act. Refraining does not presuppose that the underlying impulse ever became a formed Intention; declining a craving is refraining whether or not the craving was ever FORMS-ed as one.

CAPACITY does not itself determine whether an act performed, refrained from, refused, or requested under a given assessment is valid, binding, or should be treated as voluntary. A constrained Profile is evidence relevant to such a determination, not the determination itself; the applicable Standard, Procedure, or authorized Decision must still be identified and applied. This entry does not supply that further machinery, and does not attempt to.

Formal Pattern

Target-relative divergence within one context:
  CAPACITY(a, regarding: Intention(x), k, t, enact,
    Assessed(diminished, relative_to: prior_functioning))
  CAPACITY(a, regarding: Action(y), k, t, enact,
    Assessed(intact, relative_to: prior_functioning))

  THEREFORE a's enact capacity regarding x and regarding y are
  different facts, not one fact reported twice; a's successful
  performance of y does not contradict a's diminished enact
  regarding x, and does not by itself establish that x's own
  diminished assessment was mistaken

Refraining as agentive act, not absence:
  FORMS(a, Intention(i), at: t1)
  CAPACITY(a, regarding: i, k, t2, refrain,
    Assessed(intact, relative_to: prior_functioning)), where t1 < t2
    -- a continues to hold i; a has not enacted it and has not
       revised or abandoned it either

  THEREFORE a's non-enactment of i at t2 is itself the exercise of an
  intact capacity, not evidence that a lacks agency generally or that
  i was never really held

Profile constrained regarding one matter, silent on another:
  CONSTRAINED(a, g1, k, t, R) holds
  no CAPACITY(a, g2, k, t, D, ...) has been recorded for any D, for a
    distinct matter g2

  THEREFORE Profile(a, g2, k, t, R) is UNDETERMINED, not
  UNCONSTRAINED and not CONSTRAINED. A Commitment made by a regarding
  g2 that same day requires its own Profile, not an inference drawn
  from PORTFOLIO(a, k, t)'s worst-assessed matter

Core Relations

Typical Questions

• Regarding which specific matter, intention, decision, or act is this capacity judgment being made, and is that stated, or only the surrounding context?
• Does capacity vary across dimensions, and does it also vary across different matters within the same context and moment?
• Is a diminished or UNDETERMINED Profile regarding one matter being wrongly generalized to a different matter, including a Commitment the case actually turns on?
• Is a given non-enactment a refraining, a refusal, or an absence of intention altogether, and is the target the agent's own intention or another's request?
• Is an alternative action, requesting assistance, being read as the exercise of intact capacity through a personal, proxy, or assisted-support pathway, or wrongly as evidence of diminished capacity generally?
• Is a diminished or absent assessment being reported relative to the agent's own prior functioning or a population baseline, and has task sufficiency been kept separate as its own, Standard-governed SUFFICIENT_FOR question?
• Is an unassessed dimension or matter being treated as UNDETERMINED, or wrongly folded into UNCONSTRAINED?
• Is a limitation in external means, environment, or opportunity being misattributed to the agent's own Agency?
• Is Agency being conflated with Capability, available means, Power, Authority, Autonomy, or a legal or clinical decisional-capacity finding?

Examples

Each row shows the dominant matter relevant to that domain; a real record may show different levels for the same dimension across different matters within one context, as the Minimum Viable Test Case demonstrates directly. The healthcare row is the sharpest case in the table: an agent can be UNCONSTRAINED across every dimension regarding a specific decision while facing severe, separate limitations in bodily capability and environmental access, and conflating the two would misattribute an external or physical limitation as a deficiency in the person's own Agency.

Distinctions

Agency ≠ Agent. Agent is the ontological primitive, the bearer from which actions, claims, and commitments originate. Agency is the situated, target-, context-, and time-bound capacity attached to that bearer. The distinction is what makes it possible to say a person remains an Agent, and retains Dignity, while their capacity to act regarding a given matter is temporarily constrained, or diminished along some dimensions and not others.

Agency ≠ Capability. Capability, in the sense Power's own entry distinguishes itself from, is technical, mechanistic, or bodily capacity: what an agent's body or skill can do. Agency is broader and reflexive: it includes the capacity to refrain, refuse, and communicate, dimensions capability does not naturally describe. A person can have full bodily capability and lack the agency to consent to using it, or lack the bodily capability entirely, as in the surgical-recovery example, while retaining full agency over the decision itself.

Agency ≠ available means or environmental opportunity. A person may hold intact, UNCONSTRAINED agentive capacity across every dimension regarding a matter while lacking the external means, accessible environment, or permitted opportunity to exercise it: no working phone, no ramp, no one willing to answer. Attributing that absence to the person's own Agency, rather than to their circumstances, is a distinct and consequential error; see the External obstruction failure mode.

Agency ≠ Power. Power is the causal capacity to produce effects within institutional or social fields, a fact about an actor's standing relative to others. Agency is the situated personal capacity to form, revise, sustain, refrain from, refuse, enact, and communicate one's own intentions, a fact about the agent's own relation to their own conduct. An agent can hold substantial institutional Power while experiencing severely constrained Agency, and can retain full Agency while holding no Power at all.

Agency ≠ Authority. Authority neither entails nor is entailed by Agency. Authority is a recognized, institutionally traceable right to bind others; exercising it still requires whatever task-relevant capacities the act itself demands, exactly as any other act does, and an agent may hold Authority while severely constrained, or full Agency while holding none.

Agency ≠ Autonomy. Autonomy, in its most common philosophical sense, is a standing or ideal, self-governance according to one's own reasons, often treated as binary or as a background condition of personhood, though relational-autonomy theorists have argued for more graded accounts closer to what this entry calls Agency. This entry does not take a position in that debate. What it commits to is narrower: Agency, as used here, is a situated, target-, context-, and time-bound profile that can vary across dimensions and matters, whatever the correct general theory of autonomy turns out to be.

Agency ≠ legal or clinical decisional capacity. A legal or clinical capacity determination is a formal finding, made under a specific Standard and Procedure, about whether a person may validly consent to or perform a specific class of acts, a contract, a medical procedure, a will. Agency may supply relevant evidence to such a determination. It does not constitute it, and a constrained Agency assessment does not by itself establish incapacity in the legal or clinical sense; that finding requires the applicable Standard's own criteria and an authorized Decision.

Common Failure Modes

Minimum Viable Test Case

Person_44, six weeks into a recovery program, mid-craving, at t1, in
Context(EveningCravingEpisode) = k:

CAPACITY(Person_44, regarding: Intention(call_sponsor), k, t1, form,
  Assessed(intact, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: Intention(call_sponsor), k, t1, revise,
  Assessed(intact, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: Intention(call_sponsor), k, t1, sustain,
  Assessed(diminished, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: Intention(call_sponsor), k, t1, enact,
  Assessed(diminished, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: Intention(call_sponsor), k, t1, communicate,
  Assessed(intact, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: CravingImpulse_1, k, t1, refrain,
  Assessed(intact, relative_to: prior_functioning))
CAPACITY(Person_44, regarding: Request(offer_to_use), k, t1, refuse,
  Assessed(intact, relative_to: prior_functioning))

FORMS(Person_44, Intention(call_sponsor), at: t1)

Person_44 texts a peer: "I'm struggling to make myself call, can you
check in with me"

EXERCISES_AGENCY(Person_44, regarding: Intention(call_sponsor),
  via: Action(text_peer), mode: personal, support: unaided)
  -- sending the text is itself an unaided, personal-mode exercise;
     it recruits assistance toward call_sponsor without being an
     instance of that assistance itself. Had the peer instead called
     the program on Person_44's behalf, that further act would be a
     proxy-mode exercise regarding a distinct matter again, not a fact
     about Person_44's own enact capacity regarding call_sponsor

Result:
  Regarding Intention(call_sponsor): CONSTRAINED(Person_44,
  Intention(call_sponsor), k, t1, R) holds, R including sustain and
  enact, both diminished. Regarding Action(text_peer), the act
  Person_44 actually performed: no CAPACITY judgment above records
  enact as diminished for that matter, and the successful,
  personal-mode, unaided exercise of it is direct evidence that it
  was not. These are two different Profiles, not one fact reported
  twice. Person_44's refrain regarding CravingImpulse_1 and refuse
  regarding a hypothetical offer to use are recorded intact, and are
  themselves distinct matters from call_sponsor again. Every
  dimension shown above is Assessed; any dimension or matter not
  shown is UNDETERMINED for this agent, context, and time, not
  assumed intact.

Contrast, the non-entailment this entry exists to protect:
  Suppose Person_44 also makes a program commitment that same evening,
  regarding a wholly different matter, monthly attendance, not
  call_sponsor. CONSTRAINED(Person_44, Intention(call_sponsor), k, t1, R)
  holding does not, by itself, make Profile(Person_44,
  monthly_attendance, k, t1, R') anything at all; absent its own
  recorded CAPACITY judgments, that Profile is UNDETERMINED, not
  CONSTRAINED and not UNCONSTRAINED. And even where a Commitment's own
  matter genuinely is CONSTRAINED, that does not by itself make the
  actually-made Commitment invalid or voidable. Whether it is requires
  the program's own Standard for what capacity, regarding what
  matter, a commitment of that kind presupposes, and whether that
  Standard's conditions were met at t1, a separate, further
  determination this entry does not make on its own.

Intellectual Lineage

Agency draws on Amartya Sen's account of agency freedom within the capability approach (Sen, "Well-Being, Agency and Freedom," The Dewey Lectures 1984, The Journal of Philosophy, 1985; Development as Freedom, 1999). Sen's own explicit multi-dimensional structure is the vector of functionings and capabilities that make up a person's well-being freedom, and his work supports separating an agent's own goals from their resulting well-being; his treatment of agency freedom specifically is argued in less explicitly multi-dimensional terms in his own texts. This entry extends Sen's general capability-approach commitment to plural, non-reducible dimensions to agency itself, rather than claiming Sen already presents agency as a formal profile in exactly this entry's sense.

It also draws on Albert Bandura's account of human agency (Bandura, "Toward a Psychology of Human Agency," Perspectives on Psychological Science, 2006), which supports this entry in two ways. First, Bandura treats forethought and self-regulation, including the capacity to restrain or withhold action, as agentive in exactly the same sense as producing it, the source of this entry's insistence that refraining and refusing are agentive acts, not the absence of one. Second, Bandura's own distinction among personal, proxy, and collective modes of agency is adopted directly for this entry's <agency-mode>. The further <support-configuration> axis, unaided or assisted, is this encyclopedia's own extension, not Bandura's, added to let an ordinarily personal-mode act, texting a peer, be distinguished from a proxy-mode act, someone else acting on the agent's behalf, without collapsing the two.

Judith Herman's clinical account of trauma and recovery (Trauma and Recovery, 1992; also published as "Recovery from Psychological Trauma," 1998) describes trauma through disempowerment and disconnection from others, and recovery through empowerment, restored relationship, and the survivor becoming the author and arbiter of their own recovery. That account, Herman's own vocabulary is empowerment and authorship, not a formal treatment of "agency" in this entry's specific sense, is the reason this concept was found missing from the encyclopedia's institutional grammar in the first place. This entry does not adopt a specific formal apparatus from Herman, only the diagnosis of what was missing. Herman's own emphasis on relationship is also why this entry declines to claim that Interpersonal Recognition has no bearing on an agent's exercised capacity, see Core Relations, only that no formal entailment between the two is defined here.

Cross-References

Required

• Agent — the ontological bearer Agency is a situated capacity profile of
• Intention — what several of Agency's dimensions, form, revise, sustain, refrain, act on
• Action — one possible outcome of Agency, not its only expression; refraining produces no Action, while refusing and requesting assistance are Actions in their own right
• Context — situates a capacity judgment; regarding further individuates it to a specific matter within that context

Consequential

• Commitment — may be made, refrained from, or refused under a constrained profile; Agency alone does not determine an actually-made Commitment's validity
• Standard — may specify what capacity, regarding what matter, a given decision presupposes; SUFFICIENT_FOR is where that threshold is applied
• Procedure — may require a capacity assessment before a governed act proceeds
• Decision — may rely on a supported Agency Claim without the Claim itself constituting the Decision
• Dignity — constrains treatment regardless of any Agency assessment, intact, constrained, or undetermined
• Role — may specify what capacities a class of acts presupposes, without itself granting or removing Agency

Evidential

• Power — institutional and causal capacity, distinct from Agency's situated, personal, and reflexive capacity; Power's own Distinctions section gestures at this entry's territory without naming it
• Authority — neither entails nor is entailed by Agency; a recognized right to bind others, not a fact about one's own capacity
• Claim — asserting a capacity judgment is itself a Claim, carrying the same epistemic discipline as any other
• Warrant — the licensing principle by which evidence would support a specific asserted capacity judgment, if the judgment itself is contested

Participation (needed) — may draw upon Agency and can occur under severely constrained Agency; a further, relational concept about involvement in something beyond the self

Interpersonal Recognition (needed) — a distinct relation this entry does not formally compose with; whether being recognized by another affects an agent's exercised capacity is a further question this entry leaves open, not denies