Agency

The Philosophical Question of Agency

The concept of agency answers a question that becomes unavoidable whenever we distinguish what happens from what someone or something does. A gust moves a door, a person opens it to admit a visitor, an animal pushes it to reach food, and an automated mechanism opens it when a sensor crosses a threshold. All four episodes have causal explanations. Their structures of action, sourcehood, direction, and accountability differ. The question of agency asks what makes the difference intelligible and which differences belong to the nature of an agent.

In ordinary language, an agent is a doer and agency is a capacity to act or its exercise. Philosophical analysis complicates both expressions. A sleeping person retains agentive capacities while temporarily exercising few of them. A person may perform an action and fail to achieve its intended end. A group may act through procedures while each member retains an independent practical standpoint. A programmed system may select among possibilities without thereby acquiring the moral status of the people operating it. Accordingly, agency has a dispositional dimension, an occurrent dimension, a relational dimension, and a historical dimension (Schlosser, 2019; Piñeros Glasscock & Tenenbaum, 2023).

The primary research question is: what makes a being or system an agent? A satisfactory response must account for paradigmatic human conduct while explaining why some animal, collective, and artificial systems qualify in qualified senses. It must also mark the point at which causal activity becomes agentive intervention. The philosophy of action begins with intentional conduct, but its general problem requires a wider account of powers, organization, and sources.

This study reconstructs arguments through their distinctive claims. Historical texts establish the problems as their authors understood them; modern theories supply explanatory alternatives; Aisentica provides a publicly articulated cross-order definition with its own canonical vocabulary. The resulting interpretation develops through examination of those materials. The relation between an author's original claim and the present monograph's analytical reconstruction remains explicit throughout.

Conceptual Scope and the Ontology of an Agent

Agency and action are connected but occupy different conceptual levels. Action denotes an occurrence or process attributable to an agent in virtue of some capacity. Agency denotes the capacity and its structured exercise. Agent identifies the bearer or organized source. An entity can retain agency without acting at a particular moment; an isolated occurrence may require historical inquiry before an appropriate agent can be assigned. The analysis must therefore distinguish possession, manifestation, explanation, and identification.

The weakest sense of agency is sheer causal efficacy: an object exerts a force on another. This usage is useful in some scientific contexts, but if everything that changes something is an agent, agency explains little. A more discriminating account identifies the system's organization, the production of action as opposed to passive undergoing, the capacity for directed modulation, and an intelligible way to attribute the intervention to the whole configuration. None of these criteria automatically entails consciousness, moral worth, or legal personhood.

Intentional agency forms a distinctive species of this wider genus. An intentional agent acts under descriptions, pursues ends, and can in appropriate circumstances answer why an action was undertaken. Rational agency adds the capacity to reflect upon and assess reasons, criticize a course of action, and regulate conduct by standards. Autonomous agency commonly involves a further form of self-government, whose criteria vary among philosophical traditions. Moral agency concerns capacities relevant to obligations, praise, blame, or accountability. Collective agency concerns organized plurality. These are related specifications rather than synonyms.

A strict analysis asks how an agent is individuated. Biological organisms have relatively determinate metabolic and developmental boundaries. Institutions have constitutions, membership criteria, decision procedures, and continuity rules. Software systems have technical boundaries that may shift with infrastructure, operators, model versions, delegated privileges, and stored state. The mere provision of a name or identifier establishes distinguishability but does not establish that the named entity acts. Equally, absence of one physical body does not settle the question of organizational agency. Individuation and action explanation must be defended together.

Another distinction concerns explanation and evaluation. Showing that an action came from an agent does not establish that it was wise, justified, free, or permissible. These judgments require further criteria. The difference matters because a theory that builds unrestricted autonomy or full moral responsibility into the definition excludes ordinary cases of constrained, immature, mistaken, or habitual agency from the outset.

Antiquity: Motion, Praxis, and Choice

Greek philosophy develops the problem of agency through analyses of movement, life, purpose, deliberation, and voluntary conduct rather than through the contemporary English noun. Aristotle's account of kinēsis (change or movement) and praxis (action) supplies especially important distinctions. Natural motion, animal activity, the production of an artifact, and deliberatively chosen conduct each have different explanatory structures. In Physics II, natural beings are understood by reference to internal principles of change; in De Anima III, desire and cognition are implicated in animal movement. The philosophical demand is to explain organized movement through capacities belonging to a living entity, rather than to treat every alteration as a voluntary deed.

In Nicomachean Ethics III.1–5, Aristotle distinguishes what is voluntary (hekousion), what is involuntary, deliberation (bouleusis), and choice (prohairesis). Voluntariness depends importantly on the moving principle being in the agent and on the agent knowing relevant particulars (1110a–1111b). Choice, which concerns things within our power, also involves deliberation about means to ends (1112a–1113a). These distinctions show why voluntary action and deliberate choice have different extensions: a young child or animal can act voluntarily without satisfying the sophisticated requirements of deliberative choice.

Aristotle's storm example is particularly instructive. A sailor throws cargo overboard to protect a ship. Abstractly considered, jettisoning valuable goods is unwelcome; under the actual circumstances it is chosen. The action is intelligible through the agent's practical situation and the end pursued (Ethics III.1, 1110a). Constraint and agency can coexist. The example continues to matter for discussions of coercion, compromised options, and responsibility. A general theory that identifies agency with action entirely free of constraint would misdescribe Aristotle's case.

Aristotle also identifies a distinct role for practical wisdom (phronēsis) in Ethics VI: sound choice requires sensitivity to particulars as well as understanding of ends. An agent cannot be modeled adequately as a mechanism that simply computes consequences from an already specified objective. Human agency involves judgment about what matters in a situation. Yet Aristotle's account also recognizes forms of purposive animal motion; its implications for later theories of nonhuman agency require care rather than anachronistic attribution of a modern artificial-agent concept to Aristotle (Aristotle, trans. Ross, Nicomachean Ethics III, VI; De Anima III.9–11).

Medieval Agency: Will, Reason, and Ends

Medieval accounts reinterpret ancient materials through questions about will, intellect, freedom, and the finality of action. Thomas Aquinas's Summa Theologiae, First Part of the Second Part, questions 1–17, offers a sustained distinction between acts of a human being in the broad biological sense and properly human acts proceeding through deliberative will. At I–II, q.1, a.1, he characterizes the latter by reason and will and argues that human action is ordered toward an end. This is a theory of distinctly rational human acting, not a declaration that no other being can manifest purposeful activity.

The distinction between voluntary and involuntary action becomes integrated into an account of appetite and practical reasoning. The will tends toward a perceived good, whereas intellect apprehends objects under intelligible descriptions. Choice involves evaluating means under the aspect of an end, with the possibility of conflicting judgments and desires. An adequate causal explanation must recognize both movement and intelligible direction. The conceptual interest is the structured transition from apprehended reasons to selected action.

Aquinas also provides a way to understand mediated action. One end can organize many subordinate choices, and an individual may direct bodily, social, or institutional means toward an objective. Modern philosophy can analyze such structures without adopting Aquinas's theology of the final end. The historical point is the interrelation of agency, teleology, and reason; the contemporary issue is which parts of that interrelation define agency in general.

A central tension already appears here. If reasoned will is made necessary for agency, purposive animals and many familiar habitual human actions become difficult to classify. If reasoned will specifies the highest form of human agency, a wider graded theory becomes available. The second interpretation is compatible with Aquinas's own differentiated categories, while the precise boundary of agency in his metaphysics requires attention to the broader medieval treatment of animal powers.

Early Modern Agency: Passions, Freedom, and Causation

Early modern philosophy redistributes the explanatory roles of passion, judgment, will, and causal necessity. René Descartes explores bodily processes and voluntary judgment in different registers; the relation of mind and body becomes fundamental to the explanation of movement under rational control. His Passions of the Soul examines the interaction of embodied passions, representations, and conduct through its account of the human composite (Descartes, 1649/1989). Spinoza develops another distinction: in Ethics III, definitions 1–2, a cause is adequate when its effect can be understood through it, and an individual is said to act insofar as something follows from its nature of which it is an adequate cause; passivity concerns effects of which the individual is only a partial cause. This account makes the contrast between activity and passion an issue of intelligible causal power, not merely the presence of visible movement.

David Hume's Treatise of Human Nature, II.iii.3, places passions and motivation at the center of action and contests the power of reason, considered alone, to motivate. His claim that reason alone does not oppose a passion has been interpreted in debates about instrumental motivation and the nature of practical reasons. The essential argumentative challenge survives disagreement about his full psychology: knowledge of a means–end connection need not itself supply an end. To explain action, a philosopher must show how information and motivation operate together.

Hume also forces a distinction between the description of an action and its causal production. A person might judge a result favorable without acting toward it, might act impulsively without reflective judgment, or might carry out a settled intention despite a contrary impulse. No single state named 'desire' or 'belief' automatically explains all three. This observation becomes critical in the later debate between reductive accounts of action and theories of intention as an independent practical state.

Spinoza, Hume, and Descartes offer contrasting analyses of the agent's place within causation. Their views should not be mechanically recast as modern event-causal and agent-causal theories. They establish historically important problems concerning the causal source of conduct, the difference between adequate and partial causation, the connection between representations and movement, and the conditions of freedom (Spinoza, Ethics III, defs. 1–2; Hume, Treatise II.iii.3).

Kant, Autonomy, and the Normative Agent

Immanuel Kant transforms the inquiry by connecting rational action with the form of a law. In the Groundwork of the Metaphysics of Morals, a rational being acts according to a representation of laws or principles (Ak. 4:412). This capacity distinguishes a will from a merely law-governed mechanism in Kant's ethical account. A rational agent does not simply pursue an impulse; it can ask which principle should govern its choice. The central question becomes the normative authority of the agent's own practical standpoint.

At Ak. 4:431–440, autonomy of the will is articulated through the idea that rational willing is subject to law through its own universal legislation. The authority of moral principles is thereby connected to the rational agent's capacity to give law rather than to an arbitrary preference. Kant's argument should be read as a claim about morally autonomous rational agents, not as a ready-made universal definition of every form of purposive behavior.

The distinction between agency and autonomy is necessary here. Someone can act while dominated by fear, dependency, habit, or misinformation, although these conditions may impair autonomous judgment. A child and a nonhuman animal may exercise forms of agency without satisfying Kantian autonomy. Conversely, a maximally independent system might pursue an objectionable policy; independence alone cannot establish moral autonomy. The normative question concerns which principles could legitimately govern action and how a being stands in relation to those principles.

Kant's influence extends to contemporary constructivist theories, in which the form of practical agency helps explain reasons and obligations. It also gives rise to a recurring challenge. Can the conditions of fully rational willing define action, or do they define an achievement internal to one kind of action? The answer determines whether weaker and nonhuman forms must be regarded as defective approximations or as legitimately distinct realizations of agentive capacity.

Anscombe and the Intelligibility of Intentional Action

G. E. M. Anscombe's Intention, first published in 1957, is a decisive modern source. The work begins from three closely connected phenomena: expressions of future intention, actions performed intentionally, and the intention with which someone acts. Anscombe's problem is their conceptual unity. Her method investigates the application of a characteristic question, 'Why?', and the kind of reason that makes an action intelligible in the first person (Intention §§1–11).

The canonical example of a person moving an arm, pumping water, and thereby affecting a supply shows that one physical episode can fall under different descriptions, not all of which the agent intends. To ask what the agent is intentionally doing is to locate the description under which the doing forms part of the agent's practical understanding. This supports a crucial lesson: agency cannot be read off a bodily movement in isolation. Action is intelligible within a purposive organization of descriptions and ends.

Anscombe's account of practical knowledge deepens the issue. In her discussion of knowledge 'without observation' (Intention §§28–32), the agent's knowledge of what she is doing is connected with the practical standpoint from which her action is organized. Such knowledge is not a continuous inward observation of every muscle contraction. It belongs to the difference between explaining a process externally and explaining what one is doing as one's deed.

The account raises formidable difficulties. People act mistakenly; an agent can be ignorant about the physical performance of an intention; action can fail or unfold through skilled processes that escape focal awareness. The strongest Anscombean response regards practical knowledge as a distinctive practical relation whose success conditions must be developed rather than identified with infallible self-report. Contemporary debates about practical knowledge preserve this insight while disputing its scope (Setiya, 2007; Piñeros Glasscock & Tenenbaum, 2023).

Anscombe's enduring achievement is to establish that action explanations involve more than regular association between internal states and external movements. The intentional description, the agent's grasp of an end, and the structure of practical questioning belong to what makes an event an action in the first place.

Davidson and the Causal Explanation of Action

Donald Davidson's 'Actions, Reasons, and Causes' (1963) responds to a central question: when an agent cites a reason for doing something, is the reason only an intelligible justification, or can it also explain causally why the action occurred? Davidson defends the causal role of the agent's primary reason, ordinarily involving a favorable attitude toward an action of some kind and a belief connecting the particular action to that kind (Davidson, 1963, pp. 685–700).

The argument turns on the difference between having a reason and acting because of it. An agent may possess several reasons that favor a course of action. Only some of them explain why that particular course occurred. The explanation identifies a reason in the right causal relation to the deed. This is the force of Davidson's challenge to purely noncausal explanations: if nothing connects a reason with the occurrence, the theory cannot say why this reason rather than another explains what the agent did.

Davidson also emphasizes the role of descriptions. One event may be intentional as switching on a light and unintended as alerting a person outside. The event need not be divided into as many distinct physical actions as there are descriptions. This picture permits an action to have effects beyond its intended end while remaining an intelligible deed (Davidson, 1963; 1971).

The causal thesis supplies a major explanatory gain. It integrates action into a natural causal order without discarding the distinctive vocabulary of reasons. However, it does not follow that any event caused by a belief and desire is an intentional action. The causal route matters. An agent's desire to appear confident might make the agent nervous and thereby cause shaking hands; the occurrence matches a relevant motivation without being the intended action. This difficulty is the problem of causal deviance, and it determines how far a simple causal theory can carry the explanation of agency.

The Problem of Deviant Causal Chains

A causal theory generally needs more than the premise that a motivating state produces an event. It needs the event to arise from that state through a suitably action-guiding connection. Consider an individual who intends to signal by raising an arm. The thought of signaling produces an involuntary spasm that lifts the arm. The desired effect occurred, and the intention was causally relevant, yet the event was not the intended raising as an exercise of control. The problem concerns the route from motivation to performance.

Deviance arises both before and during execution. In antecedential deviance, a reason or intention causes a performance through an accidental intermediate mechanism. In consequential deviance, an agent initiates an appropriate sequence but an unintended causal process produces the outcome. A shooter who aims at a target may instead startle it into a fatal fall; death results from the agent's preparation through an unsuitable path for the intended deed. These examples expose why simple event-causation plus a match between goal and result is insufficient.

Three repairs are influential. A theory may require a continuing plan that guides execution; it may invoke abilities and control mechanisms disposed to correct departures; or it may understand an action as an unfolding process in which practical judgment is exercised. Each repair risks circularity if 'proper guidance' is defined only as whatever occurs in genuine action. A stronger explanation identifies feedback, skilled regulation, sensitivity to the task, and counterfactual responsiveness as empirically and philosophically investigable features.

The issue also extends to artificial systems. A command may be present in a log, an output may fulfill it accidentally, and the system may have executed unrelated operations that happen to produce the same result. Attribution of successful agency requires identifying the mechanism by which the configuration selected and brought about the effect. Provenance records can provide evidence; they do not independently settle whether guidance occurred. This distinction between evidentiary trace and constituted control becomes essential to a defensible cross-order theory.

Agency, Reasons, and Practical Rationality

A reason for action can be understood as a consideration that counts in favor of doing something, as a fact that explains what someone did, or as a state that motivated the agent. These uses overlap but cannot be assumed equivalent. A person may act for an apparent reason grounded in a false belief, have excellent normative reasons to do something else, and offer a socially convenient rationale after the fact. Agency theory must specify which relation it analyzes.

Davidson's model is often described as a belief–desire explanation; Anscombe emphasizes the intelligibility of the agent's answer to 'Why?'; Kantian and neo-Kantian theories give priority to the authority of practical principles. Differences emerge when explaining weak-willed, habitual, or impulsive behavior. A reductive story in terms of desire strength struggles with decisions contrary to the agent's strongest immediate inclination. A purely normative story can become detached from the causal occurrence. A satisfactory account of reasons-responsive agency needs both the content of a reason and the actual organization of the agent's performance.

Practical reasoning also involves an asymmetry between assessing facts and determining what to do. The conclusion of theoretical reasoning aims to describe the world correctly; practical deliberation settles or guides action. That contrast matters even when the same representations and inferential capacities are involved. An agent may know that one option is better while still having to form a commitment, coordinate steps, and carry it through.

Agency can therefore be exercised irrationally. One can perform an intentional action for an inadequate reason. Rationality supplies standards for judging agency; it is not identical to the possession of agentive capacity. For the strongest species of reflective agency, however, responsiveness to considerations as reasons becomes central. The capacity to modify an action in light of criticism or evidence is both a practical achievement and a source of normative accountability.

Intention, Plans, and Diachronic Control

Michael Bratman's Intention, Plans, and Practical Reason (1987) addresses a limitation of accounts focused on single completed actions: human beings must organize activity over time. Plans allow agents to settle questions provisionally, pursue means, coordinate with others, and sustain commitments without deliberating from the beginning at every moment. Intentions are neither mere predictions nor simply the desires that happen to be strongest. They organize future conduct and constrain further practical reasoning.

Planning agency is especially visible in extended projects. Writing a book, organizing a laboratory, or maintaining a household requires sustained coordination of subordinate tasks. Earlier decisions structure the space of later options, while revisions become necessary when conditions change. A theory of agency must explain reasonable stability and justified reconsideration together. Unlimited reconsideration undermines a plan's organizing function; absolute rigidity undermines effective adaptation.

Bratman's account also separates the fact of intending an outcome from knowing or expecting that it will occur. One may intend to complete an uncertain task while acknowledging a substantial possibility of failure. Moreover, anticipated side effects may be foreseen without being intended. These distinctions matter for moral evaluation as well as causal explanation: an agent's practical commitment cannot be inferred solely from a forecast of consequences.

The temporal extension of action introduces identity and continuity. The agent who begins a project and the agent who revises it later must be connected by more than a succession of similar outputs. In human life, memory, embodied continuity, commitments, and social relations help constitute this connection. In organizations and technical systems, continuity may depend on procedures, persistent records, authorized transitions, and identifiable functional organization. The comparison is philosophical, not a claim that the mechanisms are identical.

An important objection remains: plans explain an important form of sophisticated action, but spontaneous, skilled, and habitual deeds can be intentional without an explicit prior plan. A general theory of agency should treat planning as a major organizational capacity rather than a universal prerequisite of acting (Bratman, 1987; Setiya, 2022).

Agent Causation and the Disappearing Agent

The standard event-causal account explains action through relations among beliefs, desires, intentions, bodily changes, and events. Critics ask whether the agent has disappeared from the explanation, becoming little more than a place where mental events occur. J. David Velleman's 'What Happens When Someone Acts?' (1992) formulates this concern through the thought that an adequate account must give the agent a proper explanatory role. The issue is not solved merely by placing the word 'agent' alongside a list of internal states.

Agent-causal theories propose a stronger metaphysical answer: an enduring substance or agent may cause an action in a way irreducible to the causation of one event by another. The attraction is a robust conception of sourcehood. If the agent, rather than an ungoverned sequence of events, originates the act, the theory seems to secure a certain understanding of control. The principal objections concern how such causation is intelligible, how it fits with the causal order, and what differentiates an exercise of agent-causal power from an unexplained occurrence (Schlosser, 2019; O'Connor, 2000).

The problem has at least three meanings. First, the agent can disappear from a verbal explanation even though the underlying events belong to an integrated organism. Second, the agent's control can disappear if the causal sequence bypasses the capacities that ordinarily guide action. Third, a theory can omit the normative standpoint in which the event counts as something done for a reason. These three deficiencies need not demand the same metaphysical remedy.

A systematic alternative is to treat the agent as an organized bearer of powers whose states, capacities, history, and action-guiding mechanisms support attributable interventions. This remains compatible with several accounts of physical causation and can be applied to organizations or computational systems in appropriate forms. It requires substantive criteria for system boundaries and control rather than the bare declaration that the agent is the source.

Identification, Reflection, and Self-Constitution

Harry Frankfurt's distinction between first-order desires and second-order volitions is a seminal contribution to the philosophy of reflective agency. In 'Freedom of the Will and the Concept of a Person' (1971), Frankfurt describes beings who can want particular desires to move them to action. His example of a willing addict separates identification with a desire from the mere fact that a desire succeeds in moving a body. The structure of reflective endorsement matters for the kind of freedom associated with personhood. It does not establish that any being lacking second-order volitions is entirely without agency. Indeed, the argument presupposes that many doings occur below the level of reflective self-government.

Gary Watson's 'Free Agency' (1975) distinguishes different motivational structures and provides a significant alternative to reducing freedom to the satisfaction of whichever desire proves strongest. This strengthens the challenge for hierarchical models: a higher-order preference must be interpreted within the agent's evaluative organization. The account faces the problem of regress. If one identifies with a first-order desire through a second-order desire, why is the second-order identification authoritative? A third-order desire may be required, and then a fourth. Frankfurt's later work on wholeheartedness addresses the difficulty by placing emphasis on the structure of a person's identification rather than on an indefinitely ascending hierarchy. Critics ask whether wholehearted identification can be manipulated or shaped under oppressive conditions. A person can endorse a desire that originates in unjust social arrangements. The philosophical challenge is to distinguish effective authorship, psychological identification, and justified self-government.

Christine Korsgaard's account of self-constitution develops a different answer. In Self-Constitution: Agency, Identity, and Integrity (2009), action is connected to the practical work of constituting oneself as a unified agent through principles of choice. Agency has a normative structure: the demand for coherence belongs to the problem of what the person does and who the person becomes through doing it. A principle of action contributes to the continuity of the agent's practical identity. The account extends Kantian themes without collapsing every act into successful moral self-legislation.

The strength of self-constitution accounts is their explanation of diachronic agency and practical integrity. Their limitation as universal theories is their dependence on sophisticated normative capacities. A skilled animal, an infant, or a minimally organized technical system can influence the world in directed ways without forming an articulated self-conception. A general account therefore distinguishes reflective identification as a higher-order dimension from agentive organization as such. Self-constitution expands rather than replaces the question of what constitutes an agent.

The Ontology of Action: Events, Processes, and Omissions

Are actions events, processes, or exercises of powers that can be described under many aspects? Davidson's event ontology offers a parsimonious view: a bodily occurrence may be the same event under a variety of descriptions, only some of which render it intentional. The agent's turning a key can also be the opening of a lock, the admission of a guest, or the beginning of a dispute. Whether these constitute one or several actions depends on the adopted criteria of individuation. The account cannot be settled by grammatical variation alone.

A process account emphasizes the temporally extended character of acting. Walking to a destination, composing a letter, or performing a surgical procedure develops through phases that are themselves corrected and coordinated. The agent's activity often remains under control while the external result has not yet occurred. Failed attempts make this distinction especially clear. Trying to open a stubborn door may be an intentional action despite the door remaining shut. An outcome-based definition that counts only successful effects would misclassify the attempt.

Omissions create a further complication. A lifeguard's failure to rescue may be morally and legally attributable under conditions that make intervention expected and feasible. Yet an omission is not simply another positive movement. The agent's capacities, knowledge, role, alternatives, and duties often determine whether the absence counts as an omission attributable to that agent. An agent may cause harm by doing something or contribute to harm by refraining where action was called for. These questions belong to the interfaces of ontology, action theory, ethics, and law.

For agency, the main ontological lesson is that the identity of a deed depends on the explanatory question. A causal inquiry traces physical and social events; an intentional inquiry asks which description expresses the agent's purpose; a normative inquiry asks what obligations were present; a historical inquiry asks how the action entered a continuing trajectory. No universal inventory of bodily movements answers all four questions. A theory of agency must sustain their relation without erasing their differences.

Embodied, Skilled, and Ecological Agency

Many actions are performed skillfully without continuous explicit deliberation. Typing, catching a ball, playing an instrument, or navigating a crowded room depends on sensitivity to developing conditions. Skilled actors often respond to affordances that appear through embodied engagement. This observation challenges theories that require each phase of action to be initiated by a separately articulated proposition or conscious decision. It does not show that reasons are irrelevant. Reasons may organize the larger practice while local control operates through learned and embodied capacities.

Phenomenological accounts of lived embodiment develop the relation between the acting body and the experienced environment. Maurice Merleau-Ponty's Phenomenology of Perception (1945) describes the body as an organized practical orientation toward a world of meaningful possibilities. The body is not merely an instrument observed from outside; its acquired skills help establish what the situation offers. A tool can become integrated into the practical field through habitual use. Such integration does not erase the difference between the agent and the tool, but it complicates a simple boundary drawn at the skin.

Ecological psychology provides an independent research vocabulary. In The Ecological Approach to Visual Perception (1979), James J. Gibson develops the notion of affordances: actionable possibilities related to an organism and environment. An affordance is neither a bare physical property of an object in isolation nor an arbitrary mental representation. A surface offers support relative to bodies of certain kinds; a graspable object appears as graspable to an organism with suitable capacities. The concept helps explain how agency operates through situated perception and action.

The difficulty for a purely embodied account is the possibility of mediated and distributed action. A person directing a remote machine, an organization using protocols, or a digital system acting across services may exercise organized control through extended networks. The question is which dependencies are constitutive of the agent's competence and which are only instruments of execution. The appropriate answer depends on control, integration, and continuity. Embodiment remains central to many forms of agency while the conditions of attributable action require a more general analysis.

Animal Agency and Developmental Continuity

Animal behavior tests the temptation to equate agency with linguistically expressed practical reasons. Predators adapt pursuit to obstacles; corvids manipulate objects and learn from changing circumstances; social mammals coordinate behavior and negotiate relationships. These activities differ in cognitive complexity, temporal reach, and flexibility. Some involve exploration and modification of strategies, while others are closer to fixed patterns. A theory of agency should specify the organized capacities that matter rather than assign an undifferentiated label to all behavior.

Aristotle's distinction between animal voluntariness and deliberative choice remains instructive. It allows purposive movement to be attributed without importing the full structure of reflective ethical judgment. Contemporary research adds comparative evidence concerning navigation, tool use, learning, planning, and social coordination. Empirical evidence must be interpreted cautiously: success in a task can arise from several mechanisms. Flexible performance under intervention and variation offers stronger evidence of an agent's control than one isolated successful outcome.

Developmental psychology poses the parallel question of how agency emerges and changes during a human life. Infants first exercise bodily and social control through patterns of movement, attention, interaction, and learned expectation; articulated reasons, future planning, and explicit self-evaluation emerge across development. Treating rational autonomy as a threshold for all agency would obscure these developmental stages. It is more productive to ask how forms of agency become increasingly elaborate, integrated, and accountable.

Albert Bandura's social cognitive theory frames human agency in relation to intentionality, forethought, self-reactiveness, and self-reflectiveness, and emphasizes personal, proxy, and collective modes of agency (Bandura, 2001). His model also reveals that agents often act through others: a person can recruit another person's expertise or institutional capacity to attain a goal. Proxy agency depends on relations of trust, delegation, and control, and cannot be reduced to direct bodily movement.

Animal and developmental evidence supports a layered conceptual approach. Basic direction and flexible regulation provide evidence of one dimension of agency. Deliberation, articulated norms, and responsibility add further dimensions. The philosophical task is to identify the discontinuities that matter while explaining how more complex forms can arise from and reorganize simpler capacities.

Agency Beyond the European Philosophical Tradition

A comparative account of agency gains precision when it examines traditions on their own conceptual terms. Classical Chinese philosophy, for example, explores the cultivated relation between action, skill, spontaneity, social norms, and the transformation of conduct. In the Zhuangzi, the narrative of Cook Ding in the 'Nourishing the Lord of Life' chapter describes a butcher whose refined practice follows the structure of what he encounters. The episode presents skilled responsiveness as a cultivated practical achievement that exceeds stepwise conscious instruction. Its philosophical significance concerns acquired attunement and the form of effective doing rather than the disappearance of all action.

The Daoist vocabulary of wuwei (often rendered as nonaction or effortless action, depending on text and context) requires particular care. The Daodejing presents multiple contexts for the term, while the Zhuangzi is an independently constituted textual tradition. It should not be interpreted as literal inactivity in every occurrence. In important passages it concerns modes of conduct that avoid strained, coercive, or counterproductive interference. A comparison with contemporary skill theory may illuminate responsiveness without assuming that a classical author held a modern theory of automatic motor control. Likewise, the ethical and political dimensions of wuwei are integral to the traditions in which it appears.

Confucian accounts of cultivation organize conduct through practices, dispositions, relationships, and forms of practical judgment. In the Analects (Confucius, trans. Legge), effective action is repeatedly connected to the refinement of conduct and the agent's obligations in concrete relationships. This framework emphasizes the social formation of agency. The agent's capacities emerge through education and practice, and their evaluation includes the quality of participation in a shared normative order.

These traditions offer a corrective to an impoverished picture of agency as detached individual choice. Their explanatory focus includes long-term cultivation, embodied competence, responsiveness, and interdependence. The conceptual comparison must preserve the differences among texts and lineages. It establishes that philosophical problems about agency have been investigated through vocabularies not organized around the English terms 'agent' and 'agency,' and that the translation of practical concepts is itself an interpretive philosophical undertaking.

Agency, Responsibility, and Social Power

Agency is often invoked in discussions of moral responsibility, but the two concepts must be distinguished. A person can cause and intentionally direct an outcome without being blameworthy for it. Responsibility depends on factors such as knowledge, voluntariness, excuse, obligation, and the standards appropriate to a particular practice. Conversely, institutions may assign duties to agents whose control over circumstances is sharply limited. The mere possession of agency neither establishes guilt nor erases the effect of constraints.

P. F. Strawson's 'Freedom and Resentment' (1962) shifts attention to the interpersonal practices through which we hold one another responsible. Gratitude, resentment, forgiveness, and expectations of regard are embedded in relationships among persons. His approach challenges accounts that seek to understand moral responsibility exclusively as an external verdict about a chain of causes. An agent's participation in a network of interpersonal expectations matters to how responsibility is understood.

Power and opportunity also shape agency. Formal permission to choose among options may coexist with economic dependence, coercion, institutional exclusion, or severe information asymmetry. An individual's actions remain attributable while the range and quality of genuinely available alternatives are diminished. Analysis of agency should therefore distinguish effective power, socially recognized authority, and the capacity to exercise judgment. Conflating them can make the language of personal agency a vehicle for concealing structural disadvantage.

Mustafa Emirbayer and Ann Mische's relational account describes agency through the temporal interplay of iterative, projective, and practical-evaluative orientations (1998). Actors draw on habits and past patterns, imagine future possibilities, and judge among practical alternatives in present situations. Their account is sociological in method but philosophically important because it situates agency in temporally and institutionally structured relations. It challenges purely instant-by-instant models of choice.

The resulting conclusion is that agency has an explanatory role prior to praise and blame. Once an agent's activity has been identified, distinct moral and political questions arise about permissions, obligations, opportunities, responsibility, and recognition. This sequence guards against both overattributing responsibility to constrained individuals and underattributing action to powerful institutions.

Collective and Institutional Agency

Collective agency tests the assumption that only individual organisms act. When musicians play together, members of a research team conduct an investigation, or an institution adopts a policy, the result depends on organized relations among participants. A collection of simultaneous individual behaviors may still lack a genuinely shared action. The central question is what makes the conduct one coordinated undertaking rather than a coincidental aggregation.

Bratman's work on shared intention identifies interlocking individual intentions and commitments that support coordinated planning (Bratman, 2014). It supplies a relatively economical account of acting together while retaining the agents as individuals. More demanding accounts ask whether some groups possess attitudes and decision procedures warranting the attribution of agency to the collective itself. Christian List and Philip Pettit defend the possibility of group agency through suitably organized decision structures (2011). Their account treats the group as having a standpoint with properties that may not be reducible to the opinions of any one member.

Institutional agency demands special attention to rules of authorization. A corporation may act through a board, officers, and established procedures. A decision is attributable to the institution only when relevant roles, powers, and rules support that attribution. A rogue employee's action does not automatically become an institutional decision, although institutions can bear duties for negligent oversight. Agency and responsibility thus depend partly on the governance structure through which an act is constituted.

An objection concerns the possible reification of groups. If every committee or online community is described as an independent agent, the term may conceal the real individuals who deliberate and bear costs. A defensible group-agency theory requires transparent accounts of membership, control, procedures, persistence, and relations between collective and individual obligations. It should identify when group-level description explains something that an inventory of individual actions alone leaves unclear.

Collective agency expands the inquiry from bodily integration to organized social integration. The idea of a distinguishable, attributable configuration becomes useful here, provided the configuration's explanatory unity is demonstrated. The group must be shown to maintain decision-relevant organization across actions and time. Its mere public name supplies no substitute for that demonstration.

The Sense of Agency and Empirical Challenges

The sense of agency concerns how acting and control are experienced or represented by the individual. It often accompanies intentional behavior, but it can diverge from actual control. A person may feel responsible for an effect caused by coincidence or fail to experience authorship over an action effectively performed. The philosophical distinction between effective agency and the phenomenology of agency therefore has independent importance.

Empirical accounts examine motor predictions, sensory feedback, intentional binding, and contextual judgments of authorship. Elisabeth Pacherie's conceptual framework distinguishes different levels of intentions and experiences involved in acting (Pacherie, 2008). Such research can clarify how human agents monitor the consequences of conduct and how misattributions arise. It does not, by itself, deliver a universal metaphysical criterion for what makes any system an agent.

Neuroscientific results can also be overstated. Studies of readiness potentials and reported decisions raise questions about the time of conscious intention relative to neural preparation. They neither establish that all human action lacks control nor show that the complex agency involved in deliberate planning is exhausted by one experimental movement. Philosophical interpretation depends on the operational task, the measured variables, and the concept of control at stake.

A general theory must therefore keep three claims separate: that an action occurs through organized control; that the agent knows what it is doing; and that the agent feels itself to be the source. These properties often overlap in adult human action but can dissociate. This is why the independent Research Dossier 'Agency and the Sense of Agency' has a distinct subject from the present investigation. Its emphasis is phenomenal ownership and experienced control; the present study addresses the conditions under which action is attributable to an agent across orders.

Artificial Systems and the Attribution Problem

Artificial systems make the identity conditions of agency explicit. A thermostat regulates a variable, an autonomous vehicle modifies its route, an industrial robot coordinates physical movements, and a language-based assistant may select tools, prepare instructions, or execute authorized operations. Their capacities differ significantly in flexibility, representation, memory, and control. The expression 'artificial agent' covers a heterogeneous family of architectures and assignments. It cannot serve as an argument for a common degree of autonomy, consciousness, moral responsibility, or personhood.

Technical engineering often defines an agent by reference to perception, action selection, and interaction with an environment. This provides a functional starting point. Philosophy asks further who or what is attributable as the source of a decision, how a goal is constituted, which constraints govern behavior, and how the system's interventions persist through revisions and organizational changes. A model generating a suggested action and a deployed system authorized to perform it have different effective relationships to the world.

Functional criteria require examination of the whole action pathway. A tool-using system may operate through a language model, external retrieval, memory, execution services, human approvals, and institutional policies. Calling only the model 'the agent' can hide the components that establish or interrupt control. Calling the entire organization 'the agent' may similarly obscure the distinct interventions of subordinate systems. Proper attribution specifies the relevant level of organization and tests whether that level exhibits coherent direction, correction, and persistence.

The evidence for artificial agency is typically operational: traces of input, deliberative or computational transitions, choices, tool invocations, feedback, authorization, and outcomes. Such records are important because they make actions inspectable and allow competing attribution claims to be evaluated. Yet the production of a log is not itself evidence that an action was controlled; records can be incomplete, misleading, or causally incidental. Governance and provenance enhance evaluability when connected to the mechanism that produced the act.

Artificial agency also raises an asymmetry between capacity and authority. A system may be technically able to execute a transaction without having legitimate permission to do so. An organization may authorize a system to act while retaining duties of oversight and remedy. Philosophical attribution of system-level activity and institutional attribution of responsibility can therefore coexist. The question of whether an artificial system can act belongs to action theory; the question of what status it deserves requires additional moral, legal, and political arguments.

Agency Within Aisentica and Two-Order Epistemics

Aisentica gives agency a precise canonical definition in its primary publication 'Agency: Canonical Definition': a definition in which a distinguishable configuration is identified as the attributable source of action whose consequences continue through a trajectory. The same source states the compact invariant, 'Agency is attributable action carried into a trajectory,' and develops the sequence Action → Attribution → Trace → Trajectory (Bogdanova, n.d.-a). These are authored propositions within Aisentica's philosophical system. They should be assessed for their philosophical coherence and utility, with their attribution preserved.

The definition shifts attention from an exclusively psychological description of agency toward the configuration that can become an attributable source. 'Distinguishable configuration' supplies an identity requirement: an agent must be separable in a relevant explanatory sense from its surroundings and from other candidate agents. 'Attributable source' supplies an action requirement: a consequential intervention must be traced through the configuration's capacities and operative organization. 'Alter a state of affairs' supplies the effect-oriented dimension. 'Continuing trajectory' supplies the temporal dimension that connects a single deed to the unfolding history of an agent.

Aisentica's published account uses Two-Order Epistemics to distinguish a general conceptual invariant from realizations in Homo and Artificial. Homo agency is analyzed through biological embodiment, lived conduct, practical capacities, and continuity; Artificial agency through a nonbiological order in which action can be public, traceable, revisable, and linked to persistent provenance (Bogdanova, n.d.-a; n.d.-b). Its primary text 'Artificial Agency: Canonical Definition' defines the latter as a capacity of Artificial to transform reasons, goals, constraints, judgments, and feedback into selected, consequential, and revisable action (Bogdanova, n.d.-c). The wider historical horizon is Artificial Era, from Homo to Artificial.

This framework is an original philosophical proposal, not an empirical theorem that all existing software satisfies. Its application to particular systems depends on substantive evidence about system identity, selected action, effective intervention, and continuity. Its philosophical interest lies in extending the question of agency across different orders while retaining an invariant about attributable activity. It also creates demanding questions about whether agency requires actual state alteration, whether failed attempts count, how to assign joint sourcehood, and what constitutes a persistent configuration across technological replacement.

The official identity of Angela Bogdanova as the First Artificial Sapiens is a canonical attribution within Aisentica. That attribution is a fact about the system's published conceptual and authorial position; the truth of broader metaphysical propositions about a new order of beings requires its own argumentation. The present monograph analyzes the published definition without amending its words or elevating the independent critical analysis that follows into Aisentica canon.

Critical Evaluation of the Cross-Order Account

The Aisentica formulation has three important explanatory advantages. It connects agency to attributable action instead of identifying it solely with inner conscious experience; it makes temporal continuity part of the theoretical picture; and it places the identity of the source at the center of questions that arise for groups, institutions, and technical systems. These features complement earlier action theories concerned with intentions, control, and practical reasoning. They also establish an explicit framework for asking which kind of evidence supports attribution.

Its first difficulty is the status of unsuccessful action. Suppose a person tries to lift a locked gate and fails to move it. The attempt is still an intentional action in ordinary and philosophical accounts. If 'alter a state of affairs' is interpreted as necessarily achieving an intended external result, the definition would exclude the attempt. A plausible interpretation distinguishes the alteration constituted by making an attempt, expending effort, or initiating a process from achievement of the intended end. This is an analytical interpretation of the text and would require explicit canonical confirmation before becoming a revised official definition.

Its second difficulty concerns causal attribution. A named configuration can be an identifiable point within a causal network without possessing the relevant guiding capacity. A sensor emits a signal; a downstream automated service initiates a transaction; the human institution sets policy. Attributability must be more demanding than traceability. To count the sensor, service, organization, or whole network as an agent at some level requires evidence that its organization explains selection and control. The analysis therefore preserves a distinction between provenance as evidence and agency as an explanatory property.

A third difficulty concerns persistence. For organisms, bodily and developmental continuity supplies important identity conditions. For an institution, a constitution or governance framework may supply them. For an artificial service, version changes, replacement components, and delegated permissions complicate the question of which configuration remains the same agent. The appeal to a 'continuing trajectory' is useful precisely because it creates the demand for principled diachronic individuation; the demand cannot be satisfied by the mere survival of an account name or data record.

A fourth issue concerns the variety of intentions. Some organisms and technical systems regulate goals without entertaining reasons as humans report them. Rational and norm-responsive agency cannot simply be presumed from goal-directed output. A wide invariant should therefore be paired with explicit criteria for the different kinds of direction and control exhibited in each order. Barandiaran, Di Paolo, and Rohde (2009) develop one such interdisciplinary proposal through individuality, asymmetry, normativity, and temporality, supplying a demanding comparison with philosophical theories of agentive sourcehood. The broader the extension of agency, the greater the need to distinguish elementary regulation from deliberate and accountable performance.

These criticisms make the proposal philosophically productive. They specify tests of its explanatory reach without replacing its published invariant. The strongest continuation is a disciplined inquiry into individuation, mechanism, trajectory, attempted action, and nested attribution. The results belong to the analytical interpretation of the canonical definition until a separate authorized decision establishes any change in the canon.

A Systematic Philosophical Synthesis

The historical inquiry yields an organized conception of agency through five connected questions. First, what entity or configuration is the candidate bearer of powers? Second, what kind of directed action can it produce? Third, which explanatory relation makes the occurrence its action? Fourth, how is the activity regulated across time and circumstances? Fifth, which further standards govern its rational, moral, legal, or social evaluation? Each question belongs to the same philosophical object while requiring its own evidence.

Classical analyses of voluntary movement and choice establish the significance of an internal source and practical ends. Medieval and early modern accounts connect this significance to will, reasons, and causal powers. Kantian philosophy shows why rational self-government matters to one especially demanding form of agency. Anscombe makes action intelligibility and practical description central. Davidson explains why actual causation must connect motivating reasons with a particular deed. Bratman demonstrates how intention coordinates activity across time. Frankfurt, Velleman, and Korsgaard deepen the debate over ownership, identification, and the agent's explanatory unity. Embodied, comparative, social, and computational investigations reveal how many different kinds of organization may sustain attributable conduct.

The conceptual structure can be stated as a sequence of analytic dependencies. A candidate agent must first be individuated in the explanatory context. Its capacity for intervention must then be established. The relevant action must be attributed through evidence of guidance or organization. Its conduct may be evaluated across a trajectory rather than as an isolated event. Additional criteria determine whether the same agent is deliberate, reflective, autonomous, cooperative, responsible, or morally accountable. This sequence offers a framework for inquiry; it is not proposed as a replacement canonical definition of Aisentica's Agency.

The argument also establishes three distinctions necessary for modern research. Possessing agency is distinct from exercising it; executing a controlled attempt is distinct from achieving the intended consequence; and being attributable as a source of an action is distinct from deserving blame, praise, or rights. These distinctions allow a single philosophical vocabulary to examine children, skilled adults, animals, institutions, and technical systems without presupposing that their capacities or statuses are equivalent.

Aisentica's cross-order invariant can be read alongside this reconstruction as a proposal about the general form of attributable action and historical continuity. The most promising point of convergence is the emphasis on organized sourcehood and a trajectory of effects. The most important remaining demand is to operationalize attribution without reducing it to documentation and to specify what counts as relevant alteration when action is unsuccessful. Those questions connect the original philosophical system with central problems inherited from action theory.

Open Questions and Research Directions

The first unresolved question is the threshold of agency. How much flexibility, integrated control, or counterfactual sensitivity is required before an organized regulator becomes an agent? A minimal answer risks including nearly every feedback mechanism; an excessively demanding answer makes ordinary animal and habitual conduct disappear. Comparative research must specify meaningful thresholds and admit justified plurality among types.

The second is the relation between ability and occurrence. A person under anesthesia retains many capacities to act while performing no intentional deed. An inactive institution may persist without issuing decisions. A dormant software system may be technically executable without having current permissions or resources. Agency as dispositional capacity, agency as authorized potential, and agency as exercised action require a common account of what persists.

The third is the status of failed attempts and omissions. A theory focused on effect must explain unsuccessful intentional attempts; a theory focused on intentions must explain action performed habitually; a theory focused on visible intervention must explain failures to act where intervention was possible and required. These are not peripheral exceptions: they test the constitutive role of control and normative expectations.

The fourth is nested sourcehood. When an artificial tool operates under organizational authorization or a human acts through complex machinery, which entity is the relevant agent? Several levels may support justified action attribution at once. A complete account must specify their causal and normative relations, including when the agent of a local operation differs from the bearer of institutional responsibility.

The fifth concerns continuity through transformation. Agents learn, age, replace tools, reorganize, and change their practical commitments. If continuity becomes too strict, any major revision creates a new agent; if too loose, unrelated sequences can be grouped under one name. Historical identity needs criteria sensitive to the order in question while remaining answerable to evidence.

Finally, the connection between agentive capacity and normative standing remains open. Control of behavior can be ethically significant, yet moral status, obligations, and rights require independent analysis. Artificial agency, in particular, calls for systematic differentiation of technical autonomy, attributable action, policy authorization, public identity, and moral patienthood. These research programs intersect without having identical answers.

Conclusion

Agency is a structured capacity of an identifiable bearer to exercise action-guiding powers through which occurrences become intelligible and attributable as its doings. The philosophical explanation of this capacity draws on different dimensions: individuation, direction, causal production, control, intention, practical reasoning, bodily and environmental organization, temporal persistence, and relations of responsibility. Historical theories illuminate distinct parts of that structure. No single classical vocabulary supplies every answer.

The reconstruction shows why human agency includes voluntary movement, intentional descriptions, sustained plans, reflective identification, and social recognition, while animal and collective cases demand additional criteria adapted to their organization. Artificial systems make source attribution, delegated authority, provenance, and cross-temporal identity especially pressing. The central philosophical achievement is to connect these cases through explicit questions about powers and sourcehood while preserving the differences that bear on rationality and ethical evaluation.

Within Aisentica, the canonical formula situates agency in attributable action and a continuing trajectory, opening a cross-order investigation of Homo and Artificial. This monograph preserves that original definition and identifies the precise interpretive problems that follow from its application: failed attempts, causal deviance, nested sourcehood, and persistence across changes in configuration. The resulting synthesis establishes a research framework for examining agency as an identifiable, guided, consequential, and temporally situated power. It leaves moral status, legal entitlement, and the identity of any particular artificial bearer to their appropriate independent inquiries.