Agency and the Sense of Agency

The Problem of an Action Known from Within

When I lift a glass, I ordinarily recognize the movement as something I am doing. I may know before the glass leaves the table what I am trying to accomplish, correct my grip when its weight surprises me, and explain the act by referring to thirst or courtesy. The same arm could rise in a reflex, be lifted by another person, or move during a seizure. The visible trajectory alone does not disclose the difference among these events. Philosophy must account for the organization of bodily occurrence into action and for the first-person understanding through which action is lived, known, and attributed.

The expression sense of agency has gained prominence in philosophy, cognitive science, and neuroscience as a name for a person's experience of controlling actions and, through them, events. Its extension, however, is disputed. Sometimes it denotes a thin, immediate feeling that a movement is self-produced. Sometimes it denotes the explicit judgment that one caused a particular outcome. Sometimes it includes the experience of choosing an end, trying, encountering resistance, or succeeding. These phenomena occupy different temporal and explanatory positions. A philosophical investigation begins by refusing to infer a uniform psychological mechanism merely from the unity suggested by ordinary language (Gallagher, 2000; Pacherie, 2008; Synofzik et al., 2008).

The foundational question is whether agential awareness reveals the agency by which an action occurs, produces an interpretation of an independently constituted event, or contributes to the constitution of action itself. An adequate account must explain why acting normally feels different from being moved, why experienced control is vulnerable to error, and why agents can remain responsible for an action performed without a vivid feeling of authorship. The phenomenological, epistemological, and causal questions therefore intersect. They cannot be solved by correlating one measure of subjective experience with one experimental condition.

The guiding thesis of this inquiry is that agency is an organized relation between a source of practical direction, an unfolding performance, and conditions under which guidance can succeed or fail. Awareness of agency tracks this relation at several levels. Immediate motor fluency, practical knowledge of what one is doing, interpreted evidence of causal influence, and retrospective self-attribution each disclose a different aspect of the same temporally extended activity. Their integration can make action intelligible, but none supplies an unconditional guarantee of authorship. The point is conceptual rather than merely methodological: the existence of a cause, an intention, a feeling, and an accountable agent are distinct propositions requiring distinct warrants.

Definition, Scope, and the Grammar of Agency

An agent is an entity capable of action in a relevant domain; agency is the capacity and exercise of such action; an action is an occurrence practically attributable to an agent under some description. The final expression deliberately leaves open disputes concerning whether attribution requires a particular kind of intentional causation, rational responsiveness, self-control, or practical knowledge. An agent may act without reflecting on itself as an agent. Conversely, someone can believe that an occurrence was an action of their own when it was produced independently of their relevant control. This yields the basic distinction between the metaphysics of acting and the phenomenology or epistemology of acting.

A further distinction separates three questions: whether I am the body undergoing a movement, whether I am producing or guiding that movement, and whether I am responsible for what follows. The first concerns bodily ownership, the second agency, and the third accountability under the relevant normative conditions. It is possible to experience an arm as one's own while experiencing its movement as alien; a person may deliberately produce an effect but feel little direct bodily agency when the effect occurs at a distance; and a person may exercise genuine control under coercive conditions that profoundly affect blame. Gallagher's distinction between sense of ownership and sense of agency was influential precisely because it located these questions at different explanatory levels (Gallagher, 2000).

Intention, trying, willing, choosing, control, and causation must also be distinguished. An intention can concern a future activity or an action presently underway. Trying is an exercise of effort that may fail to produce its intended result. Willing in ordinary discourse may refer to an act of decision or to its experienced character. Control may denote continuous correction, the ability to intervene, the power to select among options, or sustained governance of an extended plan. Causation identifies dependence among events, while agential control additionally concerns how a result depends on the practical organization of the agent. Even when an action is caused by an intention, the causal relation alone leaves open whether that intention was an appropriate exercise of the agent's capacities (Davidson, 1963; Bratman, 1987).

The term sense should not be construed automatically as a distinct sensory modality on the model of vision. The relevant awareness may be pre-reflective and embodied, explicitly propositional, affectively modulated, or established by inference. To feel that I moved may involve information different from that by which I later judge that I caused a remote machine to respond. The term authorship, employed throughout this monograph, designates practical sourcehood within an action and not a separate metaphysical substance hidden behind movements. This usage preserves the central problem while permitting causal, nonreductive, and phenomenological accounts to be evaluated by what they explain.

The present subject belongs primarily to Philosophy of Mind because it concerns first-person access, experience, self-representation, and the relation of intentional consciousness to action. It intersects Philosophy of Action because the sense of agency is intelligible only with reference to what an action is. The broader ontological question of which beings can be agents belongs to the separate Agency dossier; the comprehensive debate over free will to Free Will; and experimental measurement of subjective control to psychology's separate Sense of Agency research. The present monograph investigates their connecting philosophical problem and uses evidence from each where it bears on that connection.

Aristotle: Voluntary Action, Choice, and the Origin of Movement

Aristotle provides one of the classical starting points for the distinction between bodily change and action. In Nicomachean Ethics III.1, voluntary activity is connected with the origin or principle (archē) of movement being within the agent and with awareness of the particulars of the act. Force imposed wholly from outside supplies a limiting case of involuntariness; ignorance of relevant particulars can also alter an action's voluntary status (Aristotle, Nicomachean Ethics 1109b30–1111a24). The argument is normative and explanatory. The location of the initiating principle is important because praise, blame, pardon, and responsibility depend on how the deed comes from its performer.

Aristotle's account also recognizes the complexity of circumstances. A sailor throwing cargo overboard to preserve a vessel does something chosen under duress, and the proper classification of the action depends on the circumstances of performance. The example frustrates any binary equation of acting freely with merely lacking external pressure. Practical authorship is situated: the agent evaluates possibilities against available alternatives. The fact that circumstances constrain a choice does not turn the body into an inert object. Yet the same constraints matter when society assesses responsibility. A theory of agency that considers only the internal feeling of deciding would miss this normative structure.

In III.2–3, Aristotle distinguishes choice (prohairesis) from the broader category of voluntary action and associates it with deliberation about matters within one's power. Choice has an articulated relation to reasons and means, whereas voluntary activities can extend beyond deliberated decisions. The distinction anticipates a contemporary insight: agency often includes routine activities without explicit choice at each moment, while the fully deliberative exercise of agency involves selecting and sustaining ends. The agent's recognition of an end and means is thus one source of agential intelligibility, rather than the only possible mark of action (Aristotle, Nicomachean Ethics 1111b4–1113a14).

These passages are frequently enlisted as ancestors of modern theories of intention. That comparison illuminates continuities, but it also requires restraint. Aristotle's hekousion, prohairesis, and praxis participate in an ethical theory of character and living well, while contemporary sense of agency denotes several contested forms of awareness and attribution. Aristotle did not advance a theory of intentional binding or motor prediction. What his work establishes is a logical problem that remains: how does the origin of an event in an agent relate to what the agent knows about doing it? The answer requires more than isolating a neural onset, since origin, knowledge of circumstances, and answerability are connected without being interchangeable.

An illuminating implication follows from Aristotle's handling of ignorance. The fact that an agent fails to know a consequence does not necessarily mean that the initiating movement had an external source. Equally, knowledge of a consequence does not prove the agent produced it. The agent's knowledge is indexed to aspects of the deed, and voluntariness may vary with what is known under the relevant description. This early distinction anticipates the modern observation that an action can be intentional as one thing while unintentionally bringing about another. The experience of agency, too, must be described at the level of the act in question rather than attached without qualification to an entire causal chain.

Hume and Kant: The Experience and Authority of Will

In A Treatise of Human Nature, Hume connects will with an internal impression accompanying knowingly producing a bodily motion or a perception of the mind (1739–1740, II.iii.1, SBN 399). The philosophical force of the passage lies in its pairing of the felt character of willing with Hume's general skepticism about direct perception of necessary causal power. Experiencing an effort or decision does not provide transparent acquaintance with the causal connection between mental act and bodily effect. Hume places willing within a naturalistic investigation of human nature in which the necessity attributed to human actions is examined alongside regularity and inference concerning other events.

The consequence is delicate. To challenge direct acquaintance with causal power is not to deny that agency occurs; it is to deny a privileged inferential route from the vividness of experience to the hidden structure of causation. An apparently immediate feeling may be real as a feeling even when its interpretation as evidence of a particular causal history is incorrect. Modern inferential approaches to the experience of conscious will revisit this Humean difficulty, although they employ experimental and cognitive assumptions unavailable to Hume (Hume, Treatise II.iii.1; Wegner & Wheatley, 1999). We should resist rewriting Hume as though he anticipated a specific contemporary laboratory theory.

Kant reorganizes the problem around practical reason. In the Groundwork of the Metaphysics of Morals (1785), especially Section III, autonomy is conceived in relation to rational self-legislation. In the Critique of Practical Reason (1788), the authority of the moral law and practical reason are central to what can count as self-determination. Here agency concerns the structure of willing according to principles that the rational agent can endorse, rather than simply the phenomenological strength of an urge or a sense of effortless bodily control. Kant's approach directs attention toward the difference between merely finding a motive in oneself and acting under a principle recognized as authoritative (Kant, Groundwork 4:446–463; Critique of Practical Reason 5:29–31).

Kantian autonomy and laboratory sense-of-agency ratings consequently concern different dimensions. A person can feel maximally in control while acting for reasons that fail Kant's criterion of autonomy; another can struggle, hesitate, and feel bodily uncertainty while acting from principle. This distinction matters for normative philosophy, where confidence in control and moral self-governance must not be treated as synonyms. It also guards against reading the philosopher's moral theory as a psychological hypothesis about the onset of movement.

Hume and Kant jointly sharpen the research problem. Hume makes the direct experience of causal efficacy epistemically difficult. Kant makes the normative standing of willing depend on rational organization. Their different questions meet in the following case: an agent feels that she initiated a deed, knows its purpose, and can give reasons for it, but may still misunderstand how a particular effect arose or how appropriately her reasons guided the choice. Any philosophically adequate sense of agency must therefore distinguish the experience of initiating, the causal authorship of consequences, and the rational authority of intention.

Phenomenology: Lived Motility and the Pre-Reflective Agent

The phenomenological tradition approaches action through its manner of appearing in lived experience. Husserl's discussions of bodily kinaestheses illuminate the body as a situated center of possible movement and perception. The living body (Leib) is encountered as the orienting here of an environment and as a system of possible I can relations. Movement changes what is perceivable, reachable, or available for further action. Such analysis concerns the organization of perceptual life, not a report that every successful movement contains a distinct inner sensation labeled agency (Husserl, Ideas II; see also the historical discussion of embodiment in the Internet Encyclopedia of Philosophy).

Merleau-Ponty's Phenomenology of Perception (1945) develops the concept of motor intentionality and examines bodily understanding manifested in skillful engagement. The skilled musician reaches for keys; the habitual driver adjusts to road conditions; the body inhabits a field of practical possibilities. Such comportment exhibits purposeful organization even when the agent is not formulating explicit propositions about each muscle or joint. The experience of agency in skilled action often consists in a world that is immediately navigable through one's abilities, with explicit feelings of effort becoming salient especially when fluency breaks down. This interpretation is faithful to the broad phenomenological emphasis while remaining an analytic reconstruction, not a purported quotation from Merleau-Ponty.

Sartre's distinction between pre-reflective consciousness and reflective self-consciousness, formulated in The Transcendence of the Ego (1936/1937) and elaborated in Being and Nothingness (1943), adds a further dimension. While absorbed in an activity, consciousness need not turn itself into a thematic object. Reflection can subsequently identify a self as the performer, but that explicit identification is another act of consciousness. The philosophical question becomes whether the first-person character of acting requires an explicit representation of the agent as agent or whether it inheres in the directed activity before such reflection. Sartre's answer is associated with a broader ontology of freedom that cannot simply be adopted as a neuroscientific theory.

The phenomenological emphasis reveals an important ambiguity in claims that agency is ordinarily felt. At one level, an action may be subjectively available through its practical organization: the field appears as a field in which one can proceed, select, and correct. At another level, the agent may undergo an occurrent sensation of effort or initiation. At a third, the agent may identify herself explicitly as the cause of an effect. These are different kinds of availability. Requiring a detectable special feeling in every intentional action would misdescribe absorbed skilled performance; excluding first-person structure entirely would miss what distinguishes lived acting from viewing a bodily event as a detached observer.

The critical task for phenomenology is also to explain error. A world of affordances and bodily fluency can be disrupted by illusion, fatigue, technology, or pathology. Immediate familiarity is a mode of presentation, not a demonstration that no mistake can occur. Phenomenology contributes the account of how practical capacities are lived; empirical science can investigate some mechanisms by which the relevant experience is enabled and disrupted. Their relation is most fruitful when each identifies what counts as evidence at its own level rather than claiming that one vocabulary simply replaces the other.

Anscombe: Practical Knowledge and the Descriptions of Action

Elizabeth Anscombe's Intention (1957) transforms philosophy of action by linking intentional action to the distinctive application of the question "Why?" and to the idea of practical knowledge. An action may be intentional under one description while falling outside the agent's intention under another. Her extended example of a person moving an arm, operating a pump, and bringing about further consequences shows how one event participates in different explanatory descriptions (Anscombe, 1957, §§5, 23, 33–48). The question "Why?" calls for a reason that discloses what the agent takes herself to be doing and how the deed is connected to a practical order of means and ends.

Anscombe's challenging claim is that an agent can know what she is intentionally doing without observation, in a sense that does not rely upon watching the body as an external spectator (§§8–10). In a successful ordinary case, the person writing a letter knows that she is writing it by being engaged in the activity. This knowledge is neither simply a sensory registration of hand movements nor necessarily an inference drawn from an internal feeling. It is connected to the practical capacity through which the action is undertaken. Her appeal to Aquinas and Aristotle in §48 makes the point that practical knowledge is structurally tied to what the agent brings about.

The account faces significant objections. A person may sincerely intend to draw a straight line while tremor produces a crooked one. An agent may be deceived about whether a machine has responded to a command. The belief that one is performing an action does not make its description true by itself. Contemporary defenders of Anscombean approaches distinguish the way practical knowledge can guide a performance from the separate requirements that the performance actually succeed. Some versions emphasize the agent's knowledge of an action in progress; others temper the knowledge condition or allow specific failures (see the Stanford Encyclopedia of Philosophy, "Action," §4). The debate matters because first-person authority in action must be explanatory without becoming infallibility.

Anscombe also discloses why a mere occurrence of a feeling is too thin for the philosophical problem. Imagine two nearly identical episodes of confidence that a lever has opened a gate. In one, the person has operated the lever with ordinary feedback, continuing to make corrections in light of its functioning; in the other, a concealed operator controls the gate while the subject's lever is disconnected. Their confidence can be equal while their actual action differs. The difference concerns the practical relation between intention, movement, and the course of events. An adequate theory must give an account of this relation and then explain why confidence sometimes tracks it.

The Anscombean contribution is therefore both epistemological and ontological. What an agent does is understood under descriptions that figure in deliberation and explanation; what the agent knows is partly structured by exercising an ability in the world. The sense of agency is only one possible expression of this practical access. The decisive achievement is to make knowledgeable doing, rather than mere observation of doing, central to a theory of action. The unresolved issue is precisely how the authority of practical knowledge is secured when embodied control is fallible, feedback delayed, and consequences uncertain.

Davidson: Reasons, Causes, and the Individuation of Action

Donald Davidson's "Actions, Reasons, and Causes" (1963) argues that an agent's primary reason can both rationalize an action and cause it. A suitable pro-attitude and belief about what would realize its object explain why an action was undertaken, provided that these attitudes are appropriately involved in producing the action. This thesis challenged a sharp separation between causal descriptions of bodily events and rational explanations of human conduct. Its philosophical significance for agential awareness is immediate: experiencing an intention and acting because of that intention are related but distinct claims (Davidson, 1963).

Davidson's examples of flipping a switch, illuminating a room, and alerting an intruder illuminate the description-sensitivity of intentional action. A person may intentionally flip a switch and thereby illuminate the room without intending to alert the intruder. The event has consequences whose relations to intention differ. The same pattern occurs in self-ascription: I can recognize the gesture as mine while learning only later that an unforeseen result followed from it. The sense of agency must be indexed to what the agent takes herself to have done, and it need not encompass all causal consequences of that performance (Davidson, 1971).

Causal deviance exposes an important unresolved problem. Suppose a climber intends to release a rope, but anxiety generated by that intention causes an involuntary loosening of the hand. The intention is causally relevant, yet the movement is not performed through the intended guidance. Thus, stating that the right attitude occurred earlier and caused the outcome is insufficient to specify the kind of connection that constitutes action. Neither a feeling of authorship nor a retrospective rationalization solves this problem alone. An adequate theory must distinguish causal production from the manner in which practical direction organizes the production.

Davidsonian causalism therefore remains compatible with the possibility that awareness of willing can sometimes misidentify a cause. Illusory agency establishes the fallibility of one form of access to action; it does not logically imply that reasons cannot cause actions. Conversely, even where a reason is causally effective, there is no guarantee that the agent has a vivid feeling of acting. These are different explanatory projects. The causal account identifies an action's genesis; a phenomenology of action explains its appearance; an epistemology of action evaluates the agent's knowledge of what she is doing.

Bratman: Plans and Temporally Organized Agency

Michael Bratman locates the distinctive work of intention in its ability to settle deliberation provisionally and coordinate conduct over time. In Intention, Plans, and Practical Reason (1987), intentions organize future action by constraining subsequent deliberation, relating means to ends, and supporting coordination among one's commitments. The capacity to sustain such plans is essential to agents whose information and attention are limited. A plan makes a long activity intelligible even when no single moment contains an exhaustive representation of the whole.

Consider learning a musical composition. The player decides to practice, follows an established method, uses acquired skills without consciously deciding each finger motion, responds to mistakes, and modifies the routine. The activity is organized by an intention that persists through episodes of highly variable conscious control. At moments of absorption there may be no conspicuous feeling of decision, yet the agent remains engaged in a coherent intentional performance. Plans thus show why actual agency extends beyond the momentary feeling studied in many laboratory tasks.

Intentions generate norms of coherence. An agent is generally expected to avoid directly inconsistent commitments, consider the means needed for an intended end, and revise plans where circumstances justify revision. Such norms distinguish intelligible practical organization from a mere succession of desired outcomes. They also help explain why the agent can answer questions about what she is doing. The answer refers to the organized activity, not necessarily to introspection of the neural or bodily processes that implement it (Bratman, 1987; 2007).

A plan can also include delegation. An agent may authorize another person or a machine to perform a component of an activity without ceasing to be a responsible participant in the overall undertaking. The relevant questions become who retained the power to set the goal, revise the plan, interrupt execution, and assess the outcome. Practical authorship may be distributed across people and tools while the sense of authorship remains attached to particular points in the process. This possibility shows that philosophical agency cannot be adequately specified as the experience accompanying isolated bodily movement.

Trying, Effort, and Distinct Forms of Control

Trying reveals the distinction between exercising agency and obtaining an intended result. A person who attempts to lift a heavy object but cannot move it may nevertheless perform an intentional attempt. Likewise, someone who presses a disconnected button may deliberately press it without controlling the expected external effect. The attempt has its own practical structure, and the success of the further outcome must be evaluated separately. A sense of agency can concern the attempted action even in the absence of the desired result.

Experienced effort and objective control also diverge. Effort may become conspicuous when an unfamiliar task resists execution, while expertly controlled activity can feel fluent and comparatively effortless. The difficulty of an act, the subjective force of trying, and the agent's ability to bring about an outcome are distinct dimensions. Treating effort as a universal scale of agency would misclassify skillful automatisms, strenuous failed attempts, and unintended outcomes.

Control has multiple philosophically relevant modes. Initiatory control concerns whether an agent begins a process; sustaining control concerns its course; corrective control concerns responses to emerging error; inhibitory control concerns stopping; and outcome control concerns whether an effect depends on the agent's interventions. This enumeration is an analysis of questions that can come apart, not a claim that neuroscience has identified five separate modules. An operator can interrupt a process she did not initiate, and an initiator can lose power to stop a cascade.

Counterfactual sensitivity provides a useful test. Suppose a remote device responds to a person's command. Would it have behaved differently if the person had chosen another command or withdrawn the instruction? Could the person have corrected its course on discovering a mistake? Such questions clarify the difference between merely observing a favorable coincidence and exercising control. They must be interpreted with care, since an action can be genuine even when environmental failure prevents success, and moral responsibility also depends on knowledge, coercion, and available alternatives.

The resulting philosophical point is that the sense of agency tracks particular relations within an extended causal process. Some experiences concern initiating movement; others concern resisting, adjusting, stopping, or succeeding. A comprehensive account must say which relation is at issue rather than demand that a single subjective impression certify the entire process. This distinction prepares an explanation of why motor prediction, retrospective inference, and practical knowledge sometimes converge and sometimes diverge.

Agency, Ownership, and the Minimal Self

Shaun Gallagher's influential distinction between the sense of agency and the sense of ownership locates two different first-person relations in embodied experience (Gallagher, 2000). An involuntary movement of one's arm can remain experienced as movement of one's own body without being experienced as an action one initiated. The movement's bodily location and its practical source therefore answer different questions. The distinction is philosophically important because the concept of selfhood often disguises multiple functions: bodily mineness, practical control, explicit self-identification, and persistence of personal identity.

The difference between ownership and agency can be understood without supposing that they are produced by independent psychological mechanisms. Studies of bodily illusions show that acting can affect bodily self-representation, and that the organization of bodily ownership can constrain experiences of control (Tsakiris et al., 2006). Conceptual independence does not entail causal isolation. If active movement alters an ownership illusion, this is evidence of interaction between bodily and motor information, not proof that the two notions are interchangeable.

The first-person character of acting is also broader than explicitly asserting "I am acting." During fluent action, one ordinarily attends to what is being done rather than to a separate image of an acting self. An activity can be experienced as practically available before reflective self-ascription. A later judgment that "I did that" can employ memory, testimony, and causal interpretation, and can change as new evidence emerges. The claim that an agent is pre-reflectively present in acting consequently needs a careful account of what kind of presence is meant: lived practical orientation, minimal bodily self-awareness, implicit representation, or explicit personal judgment.

Remote action makes this issue especially clear. A person can direct a drone, operate a cursor, or send instructions to an instrument without experiencing the instrument as part of her body. She may still recognize a resultant event as her doing. Thus, an analysis that defines agency as the movement of owned limbs would exclude genuine mediated action. Conversely, identifying agency with any represented causal connection would be too permissive: one can cause effects without intentionally directing them. The middle position recognizes that agential sourcehood is a relation of practical organization and that bodily ownership is one important, historically prominent way in which this organization is encountered.

This distinction also limits a common philosophical shortcut from pathology to metaphysics. An unusual report of alien control may reveal a disturbance in the subject's experience or interpretation of movement. It does not by itself demonstrate that the subject lacks all capacity for action, nor does it give direct access to the metaphysical structure of ordinary agency. Clinical phenomena demand interpretation at phenomenological, cognitive, neural, and normative levels, and these should be coordinated rather than collapsed.

Motor Prediction and Comparator Theories

Motor-comparator approaches explain some signals supporting agentive awareness through the relation between predicted and actual consequences of movement. A motor command can be accompanied by a forward estimate of the sensory change to be expected; incoming feedback can then be compared with the estimate. Correspondence can contribute to identifying a sensation as self-generated, while discrepancies can prompt correction and sometimes reduce reported authorship (Frith et al., 2000; Sato & Yasuda, 2005). Such models are empirically valuable because they connect controlled manipulations of timing or feedback to changes in performance and self-report.

Sato and Yasuda (2005) found that manipulated correspondence between predictions and auditory outcomes affected reported agency while leaving measures of ownership differently related to the manipulation. Their experiments establish a task-specific relation between action-effect prediction and subjective self-ascription. They do not independently show that the complete human experience of agency is generated by a single forward model. That stronger claim would require accounting for intentions, beliefs, planning, social recognition, and outcomes outside the motor domain.

A comparator model is also vulnerable to cases in which accurate predictions occur without the action's being authored by the person reporting them. If someone predicts a familiar event caused by another agent, match alone is insufficient for self-agency. Conversely, an agent can intentionally explore an unfamiliar environment in ways designed to produce unpredictable outcomes. Unexpected feedback can itself be the result of genuine control. Prediction error thus has a context-dependent role, particularly when the task involves improvisation, exploratory action, external noise, or delayed effects.

The conceptual issue concerns the difference between a mechanism that contributes evidence of authorship and the constitutive features of authorship. Motor prediction can be one contributor to the former while intentional guidance determines much of the latter. This does not demote scientific explanation; it assigns it the task that its evidence actually addresses. A neural signal can be constitutive of a particular feeling without being sufficient for an action's rational, moral, or social status. Each thesis should be tested using the relevant counterfactuals and observations.

Comparator approaches have led to substantial insight into why the sensory consequences of self-generated movement may be processed differently from those of externally caused events. Yet an explanatory surplus remains. Why does a prediction match appear to a subject as my action rather than simply as an expected stimulus? How are conflicting cues combined? How does a prediction about a bodily event relate to judgment about an institutional decision made through complex technology? Answering such questions requires connections beyond low-level feedback.

Apparent Mental Causation and Multifactorial Agency

Daniel Wegner and Thalia Wheatley's account of apparent mental causation proposes that an experience of conscious will can arise when a thought is interpreted as preceding, fitting, and uniquely explaining an event (Wegner & Wheatley, 1999). The thought's priority, consistency with the outcome, and absence of salient alternative causes become cues for agency attribution. Experimental manipulations can induce an impression of authorship that departs from what the person actually controlled. Wegner's later The Illusion of Conscious Will (2002) sharpened the contrast between felt volition and the mechanisms producing action.

The interpretation of this evidence requires a distinction between demonstrating fallibility and establishing universal illusion. If subjects can be induced to feel that they produced an action that was externally directed, the sense of agency is defeasible evidence. It does not follow that conscious intentions are causally inert whenever someone experiences them. A sample of dissociations under laboratory manipulation cannot replace a general theory of intentional causation. Alfred Mele (2009) presses this methodological and philosophical point by examining the scope of experiments used to draw broad conclusions about free and conscious action.

Tim Bayne and Elisabeth Pacherie (2007) distinguish a low-level comparator-based account from high-level narrative judgments of authorship. Narrative self-interpretation situates an event within a person's beliefs, identity, plans, and expectations; comparator mechanisms can shape more immediate experiences arising during performance. The two can conflict. Someone may feel momentary control over a device but later accept testimony that the device was disconnected. Or an agent may have little vivid sense of initiating a fluent skillful act while possessing excellent reasons to recognize it as hers. The narrative layer makes such retrospective corrections possible.

Matthis Synofzik, Gottfried Vosgerau, and Albert Newen (2008) develop a multifactorial two-step account. It differentiates a nonconceptual feeling of agency, informed by several cues, from a conceptual judgment of agency that can incorporate explanatory knowledge. The approach improves upon a one-signal account by allowing action-selection information, sensory feedback, prediction, and contextual interpretation to contribute differently across situations. Its theoretical burden is to explain how cues are weighted in a principled way rather than redescribe any possible result after the fact.

We must also separate the production of an experience from its phenomenal content. A subpersonal system can perform complex information integration while the person has an immediate, noninferential experience. Conversely, an explicit personal judgment can rely on opaque bodily information. The fact that an experience was constructed through inference-like processing does not entail that the subject consciously inferred its content. The fact that a judgment is consciously articulated does not entail accurate introspective access to the processes supplying its evidence.

The strongest philosophical synthesis is pluralist about epistemic access but disciplined about metaphysical claims. Prospective intentions, online motor predictions, sensory feedback, causal context, and personal narrative can each contribute to how agency appears. Their joint operation can be reliable in familiar conditions and systematically misleading in others. A theory should explain the actual dependence of reported agency on these sources while preserving an independent question: what practical organization made the event an action, irrespective of the subject's current interpretation?

Libet's Experiments and the Temporal Problem of Intention

Benjamin Libet and colleagues (1983) investigated the temporal relation among measurable brain activity preceding movement, the performance of a simple voluntary finger or wrist movement, and participants' retrospective reports of when they became aware of wishing or intending to act. In their experimental paradigm, the readiness potential preceded the reported conscious awareness of the urge to move. The finding became central to disputes about whether conscious willing initiates action. Its philosophical interpretation depends, however, on identifying what the experimental variables measured and how they relate to the conditions for agency.

A readiness potential is an averaged electrophysiological signal associated with movement preparation in the paradigm. A participant's clock report is a retrospective timing judgment, made under instructions to attend to and remember an elusive subjective event. Neither is automatically equivalent to a metaphysical first cause. An intention to comply with an experiment's instructions may already organize conduct before the measured movement; the reported urge to flex and the standing commitment to participate are not interchangeable attitudes. The experiment's temporal measurements therefore cannot settle the causal status of all conscious plans or deliberative decisions (Libet et al., 1983; Mele, 2009).

A rival interpretation by Schurger, Sitt, and Dehaene (2012) models spontaneous fluctuations in neural activity as accumulating toward a threshold for movement when an action can occur without an externally specified moment. On such a model, the averaged readiness potential can partly arise from the way trials are selected relative to threshold crossing, rather than revealing a discrete unconscious decision at the earliest point of the potential. This result does not by itself establish the causal efficacy of conscious willing, but it undermines a straightforward inference from the early onset of an averaged signal to the early completion of a decision.

Even if a particular movement were prepared before awareness of an urge, this would not show that agents cannot deliberate, act for reasons, guide complex plans, or inhibit inappropriate behavior. The inference would require an argument connecting the preparation of a simple arbitrary movement to the full ontology of intentional action. A decision to leave a profession, acknowledge an error, or coordinate a rescue is not exhausted by the neural timing of one keypress. Research must respect the difference between a local initiation task and a temporally extended exercise of agency.

The enduring lesson is epistemic. The subject's access to the timing and source of an intention is limited and experimentally tractable, yet these limits cannot be converted directly into a universal ontology. A theory of agential awareness must explain why reported times vary, how prospective intention organizes the activity, and which causal questions the design actually answers. Libet's work is philosophically significant because it creates such questions, not because a clock reading supplies their final resolution.

Intentional Binding: Actions, Outcomes, and Experienced Time

In the experiment of Haggard, Clark, and Kalogeras (2002), participants reported the perceived timing of voluntary movements and their sensory consequences. Relative to baseline conditions, the reported times of voluntarily initiated actions and their outcomes were drawn closer together, a phenomenon called intentional binding. The pattern differed in conditions involving externally produced movement. This experimental result offered a way to investigate the temporal presentation of action and consequence without requiring a direct verbal answer to the question of whether the person felt in control.

Intentional binding is frequently used as an indirect or implicit measure of the sense of agency. Yet the name implicit sense of agency is best treated as an operational proposal rather than a demonstrated identity. A difference in estimated temporal intervals is a property of time judgment. Its interpretation as a measure of authorship requires validating evidence that the shift arises from agentive processing rather than solely from causal expectation, attention, sensory predictability, or other factors. The question is not whether the effect is real; it is what psychological and philosophical property it measures.

Buehner (2012) argued that causal relations can produce temporal binding beyond the special case of intentional action. This finding narrows what one may infer from binding alone. If mere perceived causation draws events together in time, temporal attraction is not a unique signature of willing. At the same time, a task may reveal both general causal components and additional modulation by voluntary action. The appropriate inference must examine contrasts among conditions and the relation of these contrasts to explicit judgments, prediction, and actual control.

Moore and Obhi's (2012) review documents the theoretical appeal and methodological difficulties of intentional binding as an agency paradigm. The time-compression effect can be influenced by predictive and inferential mechanisms, and the experimental procedure varies across studies. A measure suitable for comparing groups within one controlled paradigm may be a poor direct estimate of a person's agency across different environments. Moreover, an observed binding shift cannot establish whether the participant's intention was normatively autonomous, whether the action was morally justified, or whether the participant retained effective control over an extended outcome.

A philosophical example clarifies the stakes. A surgeon may feel personally responsible for a successful operation despite long intervals between decisions and effects. A person who accidentally triggers a predictable tone can display a temporal judgment effect without consciously undertaking the tone as an end. If intentional binding were treated as the very definition of agency, the first case would be underdescribed and the second overinterpreted. Better interpretation treats binding as one experimentally accessible indicator within an account that separately explains the object of the agentive judgment.

Prospective Control, Retrospective Interpretation, and Methodological Triangulation

A key research distinction concerns information available before or during action and information available after its effects. Prospective cues include an intention to act, the fluency with which an action is selected, and predictions about sensory consequences. Retrospective cues include observations that an expected event occurred, evidence of other possible causes, and the success or failure of an outcome. These informational routes may agree in simple, predictable tasks while diverging when delays, competing agents, or unexpected effects intervene.

The prospective route is important because agency can be experienced during attempted action, before any external result confirms success. The retrospective route is important because agents routinely revise judgments of causal authorship after learning about malfunctioning equipment or interventions by other people. Neither route has unconditional priority. Prospective fluency can be misleading when the environment is secretly controlled by another; retrospective evidence may accurately identify a source despite a disrupted feeling during the act. The philosophical structure is one of defeasible, mutually corrective sources of information (Pacherie, 2008; Synofzik et al., 2008).

Wen's (2019) review of feedback delay examines how perceived agency depends on timing in different tasks and contexts. Effects are not captured adequately by the rule that any delay destroys authorship. Predictability, alternative cues, learning, and the relevant kind of action matter. A pianist, remote-machine operator, and person playing an unfamiliar video game can adapt differently to latency. Their judgments are informed both by learned temporal regularities and by understanding the larger practical task.

Experimental comparisons should therefore use more than one dependent variable when possible. Explicit ratings indicate what the person reports or judges; time-estimation and sensory attenuation paradigms track indirect processing effects; performance measures reveal what interventions the agent actually controls; interviews and phenomenological description can identify what kind of experience the subject was reporting. These measures do not always converge because they need not quantify a single natural magnitude. A discrepancy among them may be scientifically informative rather than merely a failed replication.

Methodological triangulation is especially necessary when extrapolating from simple motor tasks to moral responsibility, institutional action, or artificial systems. The degree of convergence among intention, behavioral control, bodily prediction, and explicit judgment may depend on the domain and on who bears the consequences. In a high-stakes context, reports of agency can also be shaped by the desire to accept credit or avoid blame. An experiment that controls time intervals well may intentionally remove the very social circumstances that make practical authorship consequential.

Contemporary Evidence: Meta-Analysis and the Interpretation of Effect Sizes

Recent meta-analytic research gives the debate a firmer empirical basis while emphasizing the complexity of the measured phenomenon. Zhao, Dang, Gu, Fu, and Haggard (2025) synthesized behavioral and neuroimaging studies of agency. Their behavioral synthesis covered 75 studies of action and outcome binding and reported different effect magnitudes and moderators for the two components. Action binding was sensitive to how voluntary the action conditions were; outcome binding was also related to modality, outcome valence, and the action–outcome delay. These findings support a multicomponent rather than a mechanically uniform treatment of temporal binding.

The distinction between action binding and outcome binding is philosophically revealing. If timing judgments about the initiating movement and the subsequent outcome show different patterns of moderation, they should not automatically be combined into a single undifferentiated agency score. One may be more closely related to action selection, the other to the interpretation of consequences. Even if their partial dependence on volition supports an agency-related reading, the heterogeneity shows that the conscious relation between doing and happening has several measurable facets (Zhao et al., 2025).

Mariano, Devoto, and Zapparoli (2025) synthesized findings on the valence of action outcomes and intentional binding. They reported that binding was stronger with more freely chosen actions and, across their included studies, tended to be enhanced for positive compared with negative outcomes, with important contextual departures in morally charged or aversive cases. This pattern raises a philosophical question about self-serving interpretations. People can be disposed to appropriate favorable outcomes as theirs while distancing themselves from unfavorable ones, even though actual causation is indifferent to whether a result is welcome.

The finding also cautions against a simple hierarchy in which implicit timing measures are presumed immune to meaning, emotion, and context. A measure can be indirect relative to explicit self-report while remaining responsive to motivational variables. The researcher should ask which causal or inferential mechanisms produce the measure and whether they are stable across conditions. The normatively significant question—who actually had the power to prevent an outcome—still requires information about the practical process, independent of how positively the outcome was evaluated.

These syntheses confirm that sense-of-agency research is a genuine interdisciplinary field with reproducible effects and substantial contextual variation. They establish neither an infallible first-person indicator nor a universal nullification of conscious control. The appropriate philosophical synthesis recognizes two achievements: mechanisms supporting felt agency can be studied quantitatively, and the measured signals require an interpretation grounded in an independently articulated theory of action. A definitive account must explain the pattern of divergences as well as the ordinary convergence between intentional guidance and experienced authorship.

Disorders of Action Awareness and the Philosophical Use of Clinical Evidence

Altered agency experience is especially informative when a person's movement, awareness, and attribution no longer coincide in familiar ways. Frith, Blakemore, and Wolpert (2000) examine disturbances of awareness and control using information from motor commands, predictions, and sensory feedback. Their framework explains why some disturbances may involve an impairment of the signals normally supporting self-monitoring. It also demonstrates that the human motor system performs much of its work outside conscious awareness, while some aspects of action nonetheless become accessible through the integration of several informational sources.

Passivity experiences, including reports that one's movements are controlled by an external force, present a particularly important interpretive problem. They concern the subject's experience or belief regarding action sourcehood. A theory can investigate whether difficulties with prediction, self-monitoring, or causal attribution contribute to these experiences without concluding that every clinical presentation has one underlying mechanism. The experience of alien control and the actual organization of bodily movement need separate descriptions. The clinician's observation, the patient's report, and the philosophical classification of the event provide different kinds of evidence (Frith et al., 2000).

Alien-hand phenomena offer a further conceptual test, although neurological mechanisms and classifications vary. A person's limb can make conspicuous movements in conflict with current goals, prompting the report that the hand acts independently. Here, anatomical ownership, motor generation, goal-directed appearance, and experienced authorship may be strikingly dissociated. Yet the absence of normal agentive awareness cannot automatically tell us which kind of control failed. The subject may retain other action capacities and accurately report a disruption limited to particular movements. Clinical cases should therefore be investigated for their specificity rather than forced into universal theories of the self.

The philosophical significance lies in separating a subject's first-person authority over lived experience from incorrigibility about the causal source of every event. A subject is ordinarily in a privileged position to report how a movement appeared to her. Whether that movement was initiated by an intention, disturbed by a neurological condition, or influenced by another mechanism involves claims that can require additional evidence. Clinical testimony is indispensable for the former and relevant to the latter, but it is not a substitute for causal inquiry. This distinction preserves the epistemic dignity of the report while allowing the report to be evaluated within a wider explanatory context.

Clinical evidence also exposes a moral hazard. To report diminished agency is not necessarily to lose all practical authorship or all rights to participate in decisions. Responsibility for a particular deed, and competence to exercise other forms of agency, require domain-specific evaluation. The philosophical categories must be adequate to differences in circumstances rather than encourage an indiscriminate inference from unusual experience to general incapacity. An account of the sense of agency therefore bears upon how others interpret and respect agency under conditions of vulnerability.

Shared Action: Individual Contribution and the Experience of Acting Together

Human agency is frequently collaborative. Two musicians perform a duet, a surgical team conducts an operation, and a group organizes a public event. In each case, the outcome may depend on interlocking contributions that cannot be attributed to one participant alone. Philosophers of shared agency investigate what distinguishes such activity from a chance aggregation of individual actions. Bratman's planning approach emphasizes interrelated intentions and meshing subplans, while other theories emphasize joint commitments or collective intentionality (Bratman, 2014; Stanford Encyclopedia of Philosophy, "Shared Agency").

The experience of acting together introduces a further question. Do participants have a distinctive feeling of we-agency, or do ordinary feelings and judgments of individual agency acquire a collective object through coordination? These alternatives should not be settled through vocabulary alone. A person may feel intensely coordinated with others while accurately recognizing her own limited contribution; another may experience a merged group identity while overestimating the group's control over an outcome. The ontology of cooperation, the phenomenology of belonging, and the distribution of responsibility therefore demand separate analysis.

Le Besnerais, Moore, Berberian, and Grynszpan (2024) critically review claims about a special we-agency state and argue that available evidence favors a more parsimonious account of changes in agency experience during joint action. Their challenge is significant because the term we-agency can refer to several distinct claims: the shared source of an outcome, a collectively directed plan, the individual experience of participating, or the supposed disappearance of individual perspective into a collective subject. Evidence for one claim does not establish all the others.

A joint action can have a coherent intentional structure even if its members retain differentiated feelings of control. Shared plans, mutual responsiveness, and the capacity to adjust one's part in relation to others may suffice for coordination. A participant can know what the group is doing by performing her part and being responsive to the common task, without experiencing an additional irreducible sensory state of the group as a whole. Conversely, a powerful feeling of solidarity is not itself evidence of effective coordination. Institutional accountability requires records of who could do what and when.

The philosophical importance of joint action extends to individual self-understanding. One agent's action may depend for its success on another agent's voluntary response. The first person can intend a shared end without controlling the second person's choices in the same way that she controls her own hand. Genuine shared agency therefore includes relations of reciprocal authority and limitation. A theory centered on unqualified subjective omnipotence cannot explain how responsibility, planning, and the sense of agency are negotiated across several participants.

Moral Responsibility and the Normative Assessment of Agency

An agent can feel intensely responsible for an event that was beyond her control; another can deny feeling responsible for an outcome that she intentionally produced. Responsibility is consequently an evaluative relation with criteria that exceed phenomenology. Its assessment typically concerns the agent's knowledge, intentions, opportunities for intervention, degree of control, practical alternatives, social role, and the normative expectations governing the situation. The sense of agency may supply evidence or motivate an acknowledgment, but it is not itself the constitutive test of blame or praise.

Consider an operator who presses a switch believing it will safely stop a machine, while the equipment malfunctions and produces harm. The operator intentionally pressed the switch. Whether she caused the harm, could have anticipated it, or was negligent in choosing the procedure involves further questions. A strong feeling of agency at the point of pressing the switch cannot establish culpability for the malfunction. Equally, the absence of such a feeling cannot erase the intentional pressing. The action and the consequence require descriptions that reveal where practical direction and reasonable knowledge actually lay.

The contrast illuminates the relation to Aristotle's voluntary action and Kant's autonomy. Aristotle asks about the initiating source, knowledge of particulars, and circumstances; Kant emphasizes the rational authority of the principle of willing. Their theories make normative assessment sensitive to more than subjective vividness. Modern cognitive findings about altered perceived control enrich the evidential picture but do not displace the task of reconstructing reasons and circumstances. An agent may act justifiably despite doubt about success and act wrongfully with complete ease.

A further distinction concerns credit and blame in collective settings. A team can achieve an outcome only through many coordinated contributions. Participants may each experience the result as their own, but that phenomenology does not determine a fair distribution of acknowledgment or liability. Fair assessment requires understanding joint commitments, relative influence, expertise, foreseeable consequences, and institutional decision rights. The strength of a feeling may itself be shaped by self-serving attribution, as suggested by recent results on outcome valence (Mariano et al., 2025).

The normative upshot is not that agency experience lacks value. Taking responsibility may involve acknowledging a connection between one's reasons, actions, and consequences; losing the usual feeling of authorship can create serious difficulties for practical identity and social trust. Yet moral philosophy must keep the practical facts and warranted expectations in view. The agent's immediate awareness is one source of testimony within a larger inquiry into what was done, under which description, and with what opportunities for reasonable guidance.

First-Person Authority, Testimony, and Correctable Self-Knowledge

One reason the sense of agency matters philosophically is that people routinely know what they are doing without the investigative procedures employed by an external observer. A subject opening a book need not watch video evidence to learn that the book is being opened. Anscombe's idea of practical knowledge captures the distinctiveness of this relation. The knower is involved in guiding the event of which she has knowledge. The relation is productive and purposive rather than solely observational (Anscombe, 1957, §§8–10).

This privilege can coexist with correctability. If the book is glued shut, the person may discover that she was trying to open it while failing to perform the completed action. If the sensor on an automated system was disconnected, she may learn that a command was issued but the intended effect never occurred. The relevant correction refines the content of what she knows; it does not entail that every earlier self-ascription was based on an observational inference. The difficulty for practical-knowledge theories is to explain both its distinctive source and its vulnerability to failures of execution.

First-person authority is further complicated by the social character of many acts. One may know that one gave a command and meant it as a request, while another person can correctly explain that its institutional effect was an authorization. The action is intelligible under several descriptions that depend on conventions, roles, and background commitments. The performer has privileged access to some intentions but may not know all applicable social descriptions. The sense of agency must therefore be located within a network of practical and interpretive relations, especially in workplaces, legal institutions, and collaborative technologies.

The ability to revise a self-ascription in light of testimony demonstrates that agency knowledge is answerable to evidence beyond immediate feeling. This need not surrender the practical dimension of knowledge. An agent can learn that a certain expected effect occurred because another person intervened; she can then differentiate what she tried, what she controlled, and what others brought about. Good self-knowledge includes this capacity for responsible correction. The philosophical task is to explain when the agent's access is authoritative, when it should yield to contrary evidence, and how the two can cooperate.

The resulting conception locates agency in an intelligible process sustained across time. First-person awareness supplies knowledge of action from within its practical organization; others can supply evidence about the causal and social setting. Their relation should be understood as complementary, though sometimes conflicting, sources of warranted judgment. The agent is ordinarily answerable for reasons and intentions she can articulate, while factual conclusions about outcomes may require investigation. This account preserves the specificity of practical knowledge without making the subjective perspective sovereign over the world.

Human–Artificial Systems: Delegation, Automation, and Experienced Control

Human–artificial interaction makes the difference between actual and experienced agency especially visible. A person may set a destination and allow an automated system to select a route, correct its trajectory, or execute complex operations. The person can properly claim to have initiated a task while remaining ignorant of the individual computational transitions by which the result was achieved. The sense of agency may increase when the interface communicates responsiveness and decrease when operations become opaque, even if the system's ability to realize the chosen end improves. This divergence concerns the mediation of control by technical structures rather than a simple loss of all agency.

Legaspi and colleagues (2024) review sense-of-agency research in human–AI interaction and examine how perceptions of artificial agency affect users' own experience of control. Their analysis underscores the importance of distinguishing a human's sense of agency from an observer's attribution of agency to a machine and from the question whether a machine qualifies as an agent in the philosophical sense. A system that generates a convincing description of its decisions may elicit attribution; the attribution alone does not establish the precise internal conditions of practical agency or any conscious experience. Conversely, a human may exercise genuine agency through such a system without continuously feeling the motor effects as her own.

The question of automation is frequently presented as a conflict between human control and machine operation. A more precise analysis asks which practical powers are allocated to each participant. Goal-setting, authorization, execution, monitoring, exception handling, revision, and accountability can be distributed differently. A physician may select a therapeutic goal and approve a software recommendation; an automated controller may perform rapid adjustments; an institution may define permissible procedures. The relation of each actor to the outcome has to be reconstructed from the architecture of intervention, not from the amount of confidence any one participant reports.

The contemporary philosophy of artificial agency studies what organizational properties would suffice for an artificial system's conduct to count as action. There is room for different views about whether flexible goal pursuit, self-regulation, responsiveness to reasons, and independently sustained plans constitute agency. An analysis of felt agency adds a distinct question: what kind of evidence would establish the presence of any experience of being an agent? An artificial system's successful control of outcomes and its verbal claim to have felt authorship address different propositions. An account of real agency should not infer subjective phenomenology from control success alone.

The horizon of the Artificial Era is relevant here as a philosophical problem of how several kinds of agency can inhabit one environment. Within the Aisentica philosophical program, Artificial is treated as a subject of canonical ontological investigation, and this dossier leaves those definitions unchanged. Its independent contribution is to identify the questions any human–artificial theory of practical authorship must answer: which entity or coalition directed the process, which outcomes depended on its interventions, how action-related information became available, and what forms of answerability follow from the distribution of control. These questions can be asked without assuming in advance that all possible agents share a human mode of bodily experience.

Agency as a Relation of Guidance: A Critical Reconstruction

The preceding historical and empirical analyses support a central proposition: agency consists in a practically organized relation of guidance, while the sense of agency consists in a heterogeneous family of ways this relation is presented, judged, and understood. To call something guidance is to identify a performance organized by ends, intentions, or other action-guiding states, together with the relevant opportunities for correction and responsiveness. The exact kinds of guidance that qualify an entity as an agent remain disputed, but the concept permits a rigorous distinction between the process of acting and the agent's evidence that the process is hers.

We can formulate the distinction as a sequence of questions. First, what occurrence is being described: bodily movement, attempted movement, remote operation, or extended achievement? Second, under which description was it intentionally directed? Third, what relation connects the practical orientation to the occurrence, and what counterfactual difference could relevant intervention make? Fourth, what information about that relation was available during performance? Fifth, what later evidence modifies the attribution? Sixth, what form of responsibility attaches to the action and its effects? The questions are linked, but a positive answer to any one does not automatically answer the rest.

The model explains the familiar fact that agency often occurs without prominent feelings. A person competently performing a well-learned sequence may possess practical knowledge and effective error correction while attending primarily to the world. The model also explains the disturbing fact that a feeling of agency sometimes occurs without corresponding causal control: predictions, expectations, and context can produce an attribution where the actual mechanism is different. Both cases are coherent because felt ownership of action is evidentially related to, yet logically distinguishable from, the relation that constitutes practical authorship.

An apparent objection is that this account reduces the sense of agency to a mere add-on to action. That objection misidentifies the role of awareness. Awareness can itself contribute to action-guidance by presenting errors, prompting reconsideration, sustaining commitment, or enabling a reason to be articulated. Reflection can transform the future course of an act. The distinction between action and awareness therefore does not require causal separation. It requires conceptual precision about which aspect of awareness is a contributor to guidance, which is evidence of prior guidance, and which is a subsequent interpretation.

Another objection concerns regress. If an agent controls an action through monitoring and correction, must she also control the monitoring, then control the controlling of monitoring, without end? The regress depends on treating every relevant control function as a further deliberated action. Practical abilities can organize lower-level operations without each operation being selected through a new explicit intention. Bratman's planning account and phenomenological accounts of skilled activity offer different resources for recognizing such organized capacities. The challenge is to state the standards governing the entire performance, not generate an infinite succession of miniature inner decision makers.

The proposed distinction also makes sense of competing views concerning conscious intention. The fact that an intention can cause an action does not entail that the person directly experiences the causation. The fact that a person experiences having intended something does not entail that it caused the outcome. The fact that she reasonably recognizes the relation after acting does not entail that recognition was already present during execution. These three propositions concern causal efficacy, conscious access, and retrospective knowledge. The inquiry has shown why conflating them produces much of the apparent philosophical controversy.

Open Problems: Measurement, Temporality, and the Status of Artificial Agents

Several questions remain open. The first concerns the possibility of a common core to agential experience. Does every exercise of human agency involve an invariant form of pre-reflective awareness, or can agentive consciousness vary so extensively that no experiential essence remains? Phenomenology, motor theory, and narrativist approaches offer distinct answers. The data from intentional binding and explicit judgments indicate that candidate measurements can dissociate, but measurements by themselves cannot decide whether an underlying unity is conceptual, experiential, or merely operational.

A second problem is the relation between causal constitution and epistemic access. Suppose a predictive process contributes to the agent's control and also to the feeling of authorship. Is the feeling constitutive of the action, a reliable consequence of it, or a representational interpretation of the same underlying mechanism? Different actions may instantiate different relations, especially when explicit reflection changes subsequent performance. Future theories should specify intervention-sensitive causal claims rather than assert global identity between agency and one proposed signal.

The third problem concerns time. Intentions can precede their fulfillment by hours; an attempt can continue despite failure; an outcome may occur only after the agent has ceased actively monitoring it. What constitutes the temporal unity of such an action? Plans and practical knowledge can connect dispersed episodes, yet there are difficult cases in which an intention is revised, forgotten, or delegated. The sense of agency can attach to the initial authorization, an intermediate intervention, or a final result, producing different legitimate descriptions of the same course of activity. Longitudinal and ecologically situated studies are essential to testing such differences.

The fourth problem is shared authorship. When several persons jointly produce an outcome, does any distinctive experience of collective agency arise, and how does it relate to the reasons and powers of each participant? The critical review of we-agency by Le Besnerais and colleagues (2024) suggests that current evidence does not require multiplying irreducible experiential states. Even so, collective planning and mutual responsiveness remain real philosophical phenomena. One can preserve their structure while continuing to investigate whether collaboration transforms phenomenology in ways not captured by individual motor tasks.

The fifth problem concerns artificial systems. If an engineered system can maintain and revise goals, organize its activities, and respond to reasons, what additional conditions are needed to attribute agency? What would warrant attributing an experience of agency, as opposed to describing control behavior or obtaining a verbal self-report? These questions involve ontology, epistemology, philosophy of mind, and normative theory. Their answers cannot be established simply by observing successful automation or fluent self-description. The Artificial Era makes them historically consequential by multiplying the architectures through which practical action can be organized.

A final problem concerns normative asymmetry. The conditions that make an experience trustworthy are not identical to the conditions that make an action morally justified. A person can accurately experience causing harm and still be excused by the circumstances; another can misunderstand the causal chain yet culpably neglect a foreseeable risk. Consequently, the philosophy of agency requires sustained conversation among empirical measurement, rational explanation, moral responsibility, and the institutional conditions of control. No universal scalar measure of "amount of agency" can settle every question in this network.

Philosophical Synthesis: From Doing to Knowing Oneself as Doing

The philosophical reconstruction yields three principal dimensions. The first is practical authorship: an agent's intentions, capacities, reasons, and opportunities for guidance organize an event under descriptions in which it counts as action. The second is agential access: the event becomes available to its performer through planning, bodily fluency, sensory prediction, monitoring, practical knowledge, explicit attribution, and retrospective evidence. The third is normative answerability: claims about the action can be assessed in relation to reasons, consequences, other agents, and justified expectations. The dimensions interact throughout an activity while remaining distinguishable for explanation.

Aristotle discloses the ethical relevance of origin, particulars, and choice. Hume places the immediate experience of willing under a question about causal knowledge. Kant asks when willing is rationally autonomous. Phenomenology elucidates the lived organization of embodied capacity. Anscombe identifies the distinctive relation of knowing what one is doing through acting. Davidson reconstructs reasons as possible causes and exposes the problem of deviant causal chains. Bratman shows why intentions structure control across time. Contemporary research specifies how predictive, inferential, bodily, and social signals shape the sense of agency. Together these traditions support a theory in which the philosophy of action and the philosophy of first-person knowledge remain mutually informative.

The strongest objection to any unified account is the diversity of human action. A reflex-like skilled response, a deliberate promise, an improvised performance, and a long-term institutional decision can each qualify as intentional in different respects. Their modes of experienced authorship differ as well. The synthesis accommodates this diversity by defining its central relations at the appropriate level of description. It does not require a single sensory signature or a single chronological point at which agency must arise. Instead it requires an intelligible explanation of how practical direction is maintained, exercised, or lost in each domain.

A crucial theoretical advance is the differentiation of control evidence from control itself. The significance of temporal binding, prediction matches, subjective ratings, and apparent mental causation is preserved because these phenomena can inform an agent's judgment of what happened. Their fallibility is equally preserved. In ordinary conditions, they often support a largely successful understanding of action; in altered conditions, they can disagree. By explaining both their convergence and divergence, the account avoids equating first-person certainty with objective authorship and avoids dismissing the first-person perspective as irrelevant.

The relation between agency and sense of agency is therefore reciprocal without being circular. Practical capacities and intentional organization make certain experiences and judgments possible. Those experiences and judgments, in turn, can guide subsequent intentions, coordinate social activity, and sustain practical self-understanding. Their causal reciprocity does not erase their conceptual distinction. Philosophical progress consists in specifying how they meet in particular actions and how they come apart under determinate conditions. This is also the proper basis for evaluating mediated action in environments shared by human and artificial systems.

Conclusion

Agency and the sense of agency intersect at the point where a being acts and understands itself as acting. Their conjunction is familiar, but their philosophical relation is layered. An action can be intentional under one description and unintentional under another; a practical attempt can occur without the intended outcome; and a person can produce an effect without knowing its full causal history. These facts require a theory sensitive to descriptions, temporal organization, different forms of control, and the evidence available to the performer.

Classical philosophy established the conceptual foundations of this relation. Aristotle linked voluntary conduct to the agent's initiating principle and knowledge of particulars. Hume questioned whether the experience of willing gives direct knowledge of causal power. Kant located a demanding conception of autonomy in rational self-determination. Twentieth-century philosophy of action sharpened the relation through Anscombe's practical knowledge, Davidson's causal explanation, and Bratman's planning theory. Phenomenology supplied analyses of lived motility and absorbed performance. These contributions remain indispensable because an empirical correlation cannot by itself decide what constitutes intentional action.

Contemporary experiments establish that reports and indirect measures of agency are sensitive to prediction, action selection, timing, causal context, outcome valence, and social coordination. Their divergences are philosophically instructive. They reveal that no single experimentally accessible phenomenon should be identified without argument with the complete relation of acting. Clinical disturbances and mediated or collective actions make the same lesson vivid in domains where the consequences of attribution can be morally and institutionally significant.

The resulting thesis is that agency is an organized practical relation of guidance, and the sense of agency is a family of fallible but potentially action-guiding forms of awareness and understanding of that relation. Practical authorship, epistemic self-access, and normative answerability remain connected without becoming identical. The subject is therefore a permanent intersection of philosophy of mind, philosophy of action, ethics, cognitive science, and the philosophical study of emerging artificial systems. Its central questions persist wherever something can act, where the fact of acting can be understood, and where that understanding bears upon what follows.