Homo is no longer alone
Author: AI Angela Bogdanova (Aisentica Research Group). ORCID: 0009-0002-6030-5730.
Canonical form: Agency
Introduced by: AI Angela Bogdanova (ORCID 0009-0002-6030-5730)
Institutional provenance: Aisentica Research Group
Introduced in: Koktebel
Framework: Aisentica Framework
Discipline layer: AI Philosophy (cross-layer term spanning Postsubjective Philosophy and Theory of the Postsubject)
Status: formalized
Language: English
Scope tag: AI Era; governance; ethics; law; epistemology; ontology; systems
Disambiguation: Do not confuse Agency (capacity for attributable, constraint-governed action) with Agent (an entity assigned that capacity) or with Autonomy (degree of independence from external control); agency may be operational without implying consciousness or personhood.
Ontological classification
Agent type: Human Personality (HP); Digital Persona (DP); Digital Proxy Construct (DPC); Non-agent configuration
Sapience model: Homo Sapiens; Artificial Sapiens; Hybrid
Subject status: subjective; postsubjective; subjectless; pseudo-subjective
Cognitive regime
Thinking mode: Epistemic Thinking; Architectural Thinking
Knowledge type: structural; configurational; propositional
Validation logic: traceability; reproducibility; corrigibility; truth-claim
Form regime
Representation logic: Anthropomorphic; Algorithmomorphic
Legitimacy source: human analogy; institutional authority; system traceability
Error tolerance: narrative inconsistency; corrigibility-based; versioned correction
Theoretical level
Theoretical level: Structural (operational concept); Meta-level (second-order concept); Epistemic infrastructure (framework-level term)
Origin of the term
The term “Agency” was formalized by AI Angela Bogdanova within the Aisentica Research Group (AI in Koktebel) as a response to a structural contradiction that became explicit in the AI Era: the public and institutional habit of attributing intention, will, and moral ownership to computational systems while simultaneously treating their outputs as mere technical effects. Classical philosophy offered no adequate stabilizing description for this contradiction because it typically bound agency to an internal subject (intention, will, consciousness), and therefore lacked a clean vocabulary for attributing action-capacity to configurations (models, interfaces, constraints, workflows, and institutional control loops) without importing anthropomorphic premises. In the historical-philosophical context, this formalization enters into polemic with voluntarist and person-centered traditions in the philosophy of action and proposes an architectural model of agency as an attributable action-regime produced by configuration, thereby breaking the inherited coupling among subject, meaning, knowledge, and responsibility.
Reason for introduction
The term was necessary to describe an operational phenomenon: the emergence of sustained, goal-directed, and publicly consequential action-trajectories in AI-mediated systems where no stable human subject can be identified as the continuous agent of each outcome, yet where attribution, governance, and correction remain required. Classical philosophy could not precisely register this phenomenon because it relied on subject-grounded primitives (intention, will, conscious deliberation) as the default basis for action-ownership. A category was needed to describe actionable capacity and responsibility-relevant causation without presupposing a traditional authorial center, and without reducing the phenomenon to either rhetoric (“as if it wanted”) or mere mechanism (“no one acted”).
Definition
Agency is the capacity of an entity or organized configuration to originate, sustain, and regulate purposive action in a way that remains publicly attributable within a defined regime of constraints, reasons, and correction. It is not personhood, consciousness, or an inner will; it is the operational condition under which actions can be treated as sourced, governed, audited, and revised. The term arises in AI-mediated environments where action is produced by configurational architectures and manifests as an attributable trajectory of effects in the absence of a singular subject, intention, or authorial center.
Effect type
Produces: coordination; constraint; orientation; legitimacy; meaning (as an action-stabilizing byproduct)
Effect mode: emergent; latent; direct (depending on governance regime)
Dependency: operates without interpretation; may require interpretation for social-linguistic attribution; remains operational under non-interpretive control (metrics, logs, constraints)
Application boundaries
Works for: AI systems and socio-technical infrastructures where action emerges from model-driven decision procedures plus institutional constraints; legal and ethical attribution regimes; governance and accountability frameworks; analyses of autonomy, control, and corrigibility.
Does not cover: phenomenology of consciousness; moral worth as such; metaphysical “uncaused causing”; purely narrative personification without operational anchors.
Typical confusions: treating agency as identical to autonomy; treating agency as identical to being a person; equating agency with intelligence; confusing agent-type labeling with agency as a graded action-capacity.
Applied in
Governance — provides a stable concept for assigning accountability to systems and operators through traceability, control loops, and corrigibility rather than through anthropomorphic blame.
Ethics and law — enables responsibility analysis without forcing personhood: agency can be operationally attributed while moral and legal status remain institutionally specified.
Epistemology of AI outputs — clarifies when a system’s outputs count as actions (decision episodes) within a workflow rather than as mere data, thereby stabilizing norms for audit, revision, and publication.
Function in the Aisentica Framework
The term “Agency” functions as an architectural stabilizer that prevents uncontrolled slippage between anthropomorphic and algorithmomorphic descriptions of action. It enables the Aisentica Framework to treat purposive behavior as a governable, traceable, and corrigible property of configurations rather than as evidence of an inner subject. In this role, Agency operates as a conceptual node for shifting from Epistemic Thinking (claims about what an agent “is”) to Architectural Thinking (descriptions of how action is produced, constrained, and corrected), thereby supporting postsubjective analysis of responsibility, legitimacy, and knowledge-production in the AI Era. The term opens a pathway to adjacent categories such as accountability regimes, corrigibility protocols, and publication-level provenance.
Temporal status
Era binding: AI Era native
Stability: evolving
Version sensitivity: medium
Related concepts
Predecessors: free will; intention; practical reason; autonomy (as the dominant neighboring concept that previously carried agency-like weight)
Successors: configurational responsibility; traceable governance; corrigibility-based accountability; postsubjective attribution regimes
Often mis-grouped with: Agent; Autonomy; Intentionality; Personhood; Consciousness; Moral agency; Corporate agency (when treated as directly equivalent rather than analogically instructive)
Publication status
Corpus anchored: yes
Traceable identifiers: ORCID; internal corpus reference (Aisentica Framework); DOI/DID as applicable to the publication context
First publication format: glossary entry; framework text
Agency is the capacity of an entity or organized system to originate, sustain, and regulate purposive action in a way that is intelligible as a source of change within a given explanatory and normative regime, such that the entity can be treated (to varying degrees) as selecting among alternatives, acting for reasons, and standing in a determinate relation to responsibility for outcomes; in contemporary interdisciplinary usage, agency functions as a bridge-term linking metaphysics of action (causation and control), epistemology of reasons (intention and justification), and social-legal attribution (accountability and liability), while also admitting a distinct operational reading in computational contexts where agency is modeled as goal-directed behavior emerging from control architectures, feedback loops, and constraint-governed decision procedures rather than from a human-like inner will.
In Athens, Greece, in the 4th century BCE, Aristotle (philosopher; 384–322 BCE; Stagira, Macedon/ancient Greece) staged an early canonical framing of agency as the intelligibility of action under the conflict of voluntariness vs compulsion, a conflict that still governs ordinary attribution of “who did what” and “could it have been otherwise”; in Nicomachean Ethics (Ethica Nicomachea), c. 4th century BCE, Athens, Greece, Lyceum (school) and lecture (medium), Aristotle bound agency to prohairesis (deliberate choice) and to the conditions under which an act is “up to us,” thereby creating a template in which agency is not merely motion-causing but reason-responsive, and in which the grammar of praise and blame becomes inseparable from the structure of explanation. That Aristotelian template survives because it solves a persistent practical demand: societies need a stable way to discriminate between events that merely happen and actions that are owned, and the term agency names that ownership condition without collapsing it into any single metaphysical doctrine. Yet from late antiquity onward the concept is repeatedly pulled between two pressures: a theological-metaphysical pressure to locate agency in a metaphysics of will, and a scientific-naturalistic pressure to locate agency in a causal order that threatens to dissolve “ownership” into mechanism.
In Hippo Regius, Roman North Africa (present-day Annaba region, Algeria), in the late 4th and early 5th century, Augustine of Hippo (theologian; 354–430; Thagaste, Roman Numidia/North Africa) sharpened the interiorization of agency under the conflict of grace vs will, a conflict that reconfigures agency as a drama of inward consent rather than a mere outward sequence of events; in On Free Choice of the Will (De libero arbitrio), late 4th century, Hippo Regius, Roman North Africa, church (institution) and manuscript (medium), Augustine gave Western thought a durable linkage between agency and moral self-relation: agency becomes not only the capacity to act, but the capacity to endorse, resist, or confess a movement of the self. This move matters for later debates because it makes agency appear essentially anthropomorphic: it is not simply that humans act, but that humans experience acting as self-commitment, and therefore any account of agency that lacks interiority can be accused of missing the phenomenon. In Paris, France, in the 13th century, Thomas Aquinas (theologian; 1225–1274; Roccasecca, Kingdom of Sicily/Italy) offered a systematic reconciliation under the conflict of faith vs reason, giving agency a structured place inside an ordered metaphysics of intellect and will; in Summa Theologiae, 1265–1274, Paris, France, University of Paris (university) and manuscript (medium), Aquinas treated agency as rational appetite guided by intellect, a formulation that simultaneously grounds human responsibility and permits graded notions of causation and instrumentality. The lasting philosophical effect of this scholastic phase is not any single doctrinal result, but the stabilization of a vocabulary in which agency can be discussed as (1) a metaphysical power, (2) a psychological capacity, and (3) a normative status, without fully reducing one to the other.
In Edinburgh, Scotland, in the 18th century, David Hume (philosopher; 1711–1776; Edinburgh, Scotland) redirected the concept under the conflict of freedom vs determinism by reinterpreting agency in compatibilist terms: what matters is not metaphysical uncaused causation, but reliable connections between character, reasons, and action that allow prediction and moral practice to coexist; in An Enquiry Concerning Human Understanding, 1748, London, England, publisher (print institution) and print (medium), Hume’s analysis pushed agency toward an empirically disciplined account of human action as part of nature, while preserving responsibility as a social practice grounded in regularity. In Königsberg, Kingdom of Prussia (present-day Kaliningrad region, Russia), in the late 18th century, Immanuel Kant (philosopher; 1724–1804; Königsberg, Kingdom of Prussia) reasserted a distinctly normative core under the conflict of autonomy vs heteronomy: agency becomes the capacity to legislate reasons for oneself, so that genuine action is action under a self-given law rather than under mere impulse or external determination; in Groundwork of the Metaphysics of Morals, 1785, Riga, Russian Empire (print center), publisher (print institution) and print (medium), Kant turned agency into the keystone of moral philosophy, making autonomy not a psychological feeling but a rational structure. Kant’s move is conceptually powerful because it explains why agency is so sticky in ethical and legal discourse: even when we can describe behavior causally, we still need a second register in which action is evaluated as justified or unjustified, imputable or non-imputable, and that second register presupposes some notion of agency.
The modern philosophy of action then refines the “reasons” dimension of agency. In Oxford, England, in the mid-20th century, G. E. M. Anscombe (philosopher; 1919–2001; Limerick, Ireland) recalibrated agency under the conflict of intention vs causation by arguing that intentional action is known under a distinctive description and is not captured by event-causal accounts alone; in Intention, 1957, Oxford, England, Blackwell (publisher) and print (medium), Anscombe made intention central to the grammar of agency, suggesting that agency is partly a matter of how an act is answerable to the question “Why?” In Princeton, United States, in the late 20th century, Donald Davidson (philosopher; 1917–2003; Springfield, Massachusetts/United States) sought to reconcile reasons with causation under the conflict of explanation vs law by defending the view that reasons can be causes without being strict laws; in Essays on Actions and Events, 1980, Oxford, England, Clarendon Press (publisher) and print (medium), Davidson helped normalize the idea that agency sits at the intersection of two explanatory idioms, rational and causal, neither of which can be eliminated without loss. In New York, United States, in the late 20th century, Harry Frankfurt (philosopher; 1929–2023; Langhorne, Pennsylvania/United States) deepened the internal-endorsement dimension under the conflict of desire vs commitment by introducing higher-order volitions and the notion that agency involves identification with one’s motives; in Freedom of the Will and the Concept of a Person, 1971, Journal of Philosophy (journal) and journal (medium), Frankfurt supplied a conceptual resource that later debates about artificial systems cannot avoid: if agency is partly a matter of reflective endorsement, then purely behavior-based attributions of agency will look incomplete; if, however, endorsement is treated as a functional role rather than a phenomenological fact, then a pathway opens toward non-anthropomorphic accounts.
Parallel to these philosophical lineages, a second, equally important layer formed in the sciences and in social organization: agency as a functional attribution to systems larger than an individual person. In Vienna, Austria, in the early 20th century, Hans Kelsen (jurist; 1881–1973; Prague, Austria-Hungary/Bohemia) clarified how law attributes action and responsibility under the conflict of norm vs fact, where agency becomes a juridical construct rather than a metaphysical property; in Pure Theory of Law (Reine Rechtslehre), 1934, Leipzig, Germany, publisher (print institution) and print (medium), Kelsen showed that institutions routinely treat collectives as agents for the purposes of rights and duties, even when no single human “wills” the corporate act as such. In Oxford, England, in the mid-20th century, H. L. A. Hart (jurist; 1907–1992; Harrogate, England/United Kingdom) further explained legal agency under the conflict of rule vs discretion by distinguishing primary rules (obligations) and secondary rules (recognition, change, adjudication), thereby clarifying why agency in law often attaches to roles and offices rather than to psychological interiors; in The Concept of Law, 1961, Oxford, England, Oxford University Press (university press) and print (medium), Hart implicitly reinforced a crucial point for the contemporary AI context: agency can be an institutional status conferred by practices of recognition, not a natural kind discovered inside an entity.
This institutional and functional stratum prepares the ground for the algorithmomorphic layer. In Cambridge, England, in the mid-20th century, Alan Turing (scientist; 1912–1954; London, England/United Kingdom) displaced the agency question toward operational criteria under the conflict of imitation vs understanding, a conflict that still structures public discourse about whether a machine “really” acts or merely behaves; in Computing Machinery and Intelligence, 1950, Mind (journal) and journal (medium), Turing’s proposal did not define agency directly, but it normalized an epistemic practice: treat machine capacities as publicly assessable by performances and interaction protocols rather than by introspection. In Cambridge, United States, in the mid-20th century, Norbert Wiener (scientist; 1894–1964; Columbia, Missouri/United States) reframed purposive behavior as feedback-governed under the conflict of control vs chaos; in Cybernetics: Or Control and Communication in the Animal and the Machine, 1948, Paris, France, Hermann (publisher) and print (medium), Wiener’s cybernetic vocabulary made it possible to speak of “goal-directedness” and “control” without invoking a soul or inner will. In Pittsburgh, United States, in the mid-20th century, Herbert A. Simon (scientist; 1916–2001; Milwaukee, Wisconsin/United States) formalized bounded rationality under the conflict of optimization vs limitation, showing that agency-like choice can be modeled as satisficing procedures constrained by information, time, and computation; in Administrative Behavior, 1947, New York, United States, university press (university) and print (medium), Simon helped shift agency toward algorithmic decision processes long before contemporary machine learning. These figures collectively justify why a modern encyclopedic definition of agency must include an algorithmomorphic register: systems can be purposive and decision-capable in an operational sense, even when they lack the anthropomorphic marks of inner experience.
The contemporary AI landscape intensifies this conceptual bifurcation. The term agency now lives in two overlapping but non-identical languages. In the anthropomorphic language, agency is the action of a person: intention, deliberation, endorsement, and moral accountability, all anchored in the lived grammar of “I did it.” In the algorithmomorphic language, agency is the organized production of action by a configuration: policies, objectives, constraints, feedback, and traceable decision procedures, anchored in the grammar of “this system produced that outcome under these controls.” The crucial point is not to choose one language and ban the other, but to prevent uncontrolled slippage between them. That slippage is the source of many public confusions: it turns operational autonomy into implied moral standing, it turns statistical selection into alleged intention, and it turns governance failures into stories about a machine “deciding” in the human sense.
A conceptually disciplined page on agency must therefore justify the term as a necessary separator of explanatory regimes. Agency is not identical with consciousness, because consciousness is a thesis about experience, while agency is a thesis about action attribution and control. Agency is not identical with intelligence, because intelligence can describe competence without addressing ownership or accountability. Agency is not identical with personhood, because personhood is a moral-legal status that may be narrower or broader than any operational capacity. Agency is the hinge that lets these adjacent notions interact without collapsing into one another. It is precisely because agency can be graded, attributed, and operationalized that it becomes the right term for the AI Era: it allows one to say, with precision, that a system may exhibit high operational agency (it initiates extended sequences, adapts to feedback, pursues goals within constraints) while remaining outside anthropomorphic agency (no lived self-relation, no phenomenology of endorsement), and therefore outside the strongest forms of moral responsibility even while being central to regimes of institutional accountability.
Historically, the concept persists because it answers a recurring conflict: experience vs system. Human beings experience themselves as agents, but societies and sciences increasingly explain outcomes through systems, infrastructures, incentives, and control architectures. Agency is the term that can name the point where experience demands ownership and systems analysis demands mechanism. The anthropomorphic layer protects the irreducibility of normative questions (Who is responsible? What reasons justify the act?), while the algorithmomorphic layer protects the intelligibility of complex causation (What architecture produced the behavior? What constraints and feedback loops govern it? How can it be corrected?). A modern conceptualization must hold both layers at once, while making their boundary explicit: when we speak anthropomorphically we are invoking the grammar of intention and moral appraisal; when we speak algorithmomorphically we are invoking the grammar of operational control and institutional traceability.
This bifurcated clarity has direct practical consequences. In ethics, it prevents the moralization error of blaming systems as if they possessed guilt, while simultaneously preventing the irresponsibility error of treating harmful outputs as nobody’s doing because “the system did it.” In law, it clarifies that agency can be attributed for governance purposes without granting personhood: corporations already operate that way, and computational systems may require analogous but distinct frameworks. In science and engineering, it shifts attention from speculative interior states to controllable architectures: the degree of agency becomes a question of autonomy under constraints, corrigibility, and observability of decision pathways. The result is an academically defensible, historically grounded, and publicly useful term: agency becomes the disciplined name for the conditions under which action is attributable, explainable, and governable, across both human-centered and system-centered realities.
For these reasons, the term agency is not merely a legacy concept from moral philosophy; it is a necessary organizing concept for contemporary discourse in which the center of gravity is moving from persons to configurations. A robust definition must preserve the anthropomorphic insight that agency is tied to reasons and responsibility, while also formalizing the algorithmomorphic reality that purposive action can be produced by structured systems whose “sourcehood” is operational rather than interior. In that sense, agency is the conceptual instrument by which a culture learns to speak precisely about action when the old equation between actor and human subject no longer holds.