From Division to Relation—and Beyond: Locke, Hume and the Reconstruction of Agency in the Age of Artificial Intelligence

Abstract

Modern Western thought is often narrated as a progressive movement toward analytical clarity: Separating fact from value, person from property, observer from observed and individual from community. This paper argues that these distinctions became more than philosophical descriptions. Through law, property, contract, accounting, administration and technical systems, they became institutional infrastructure. The resulting order privileged the discrete, self-identifying individual as the principal unit of ownership, agency and responsibility. Locke provided a powerful philosophical formulation of the proprietary individual; Hume articulated the distinction between descriptive and normative reasoning; and the scientific revolution supplied increasingly powerful methods for representing identifiable entities and their interactions. Their institutional consequences, however, cannot be reduced to their authors’ intentions. The colonial and commercial order in which these ideas developed included slavery and exclusion from full legal personhood.

Artificial intelligence now exposes the limits of this architecture. Consequential AI capability is rarely located in a model alone but emerges from relations among models, data, infrastructure, permissions, humans, institutions, other agents and environments.

The appropriate response is therefore not simply a shift from substance to relation, but ultimately from relation toward field: From isolated entities, through relational systems, toward the conditions within which entities and relations themselves arise. Drawing on Buddhist and Vedāntic philosophy, Indigenous relational traditions, commons scholarship, second-order cybernetics and non-classical logic, the paper proposes relational accountability as an institutional response to distributed and emergent artificial agency.

Introduction: When the Entity Is No Longer the Unit of Analysis

The contemporary question "What is the AI?" appears descriptive. Is it a tool, agent, autonomous system, product, or person? Yet classification increasingly allocates consequence. Describing an AI as a tool may locate responsibility in its operator; describing it as autonomous may shift attention toward the system itself.

The deeper problem is that the entity may no longer be the appropriate unit of analysis. An AI system’s consequential capacity can depend upon models, training data, prompts, memory, tools, APIs, permissions, human supervision, other agents, institutional objectives, feedback loops, physical infrastructure and legal authority. Remove those relations and one may no longer have the same operational system.

AI therefore exposes a problem that predates computation: modern institutions have become exceptionally effective at converting relationships into identifiable entities for purposes of ownership, measurement, attribution and control.

This movement is already beginning to acquire an architectural form in AI ethics. Wisdom as a Service (WaaS) proposes an explicitly pluralist framework integrating Ubuntu, Śrī Vidyā, Dao, Kabbalistic Sephirot, Sufi Tawḥīd and Dzogchen, together with Nyāya epistemology, to foreground relationality, contextual discernment and metaphysical pluralism [1]. Its significance here is methodological: Rather than asking which single ontology should govern AI, it asks whether an AI system can remain capable of discernment across multiple ontological and ethical frames.

From Person to Proprietor

The genealogy does not begin with Locke. Fifteenth-century papal bulls, including Dum Diversas (1452), Romanus Pontifex (1455) and Inter Caetera (1493), contributed to a legal-theological framework subsequently invoked in European colonial expansion. Contemporary Catholic authorities have rejected the modern use of a generalized "Doctrine of Discovery"; nevertheless, the historical instruments formed part of a European legal order in which sovereignty, land and the status of Indigenous peoples were contested through concepts of dominion and possession.

Locke subsequently gave the proprietary individual an influential philosophical formulation: "Every Man has Property in his own Person" (Second Treatise, §27). Locke’s relationship with the colonial order was not merely theoretical. He served as secretary to the Proprietors of Carolina, was associated with drafting its Fundamental Constitutions, held shares in the Royal African Company and served in colonial-trade administration. The Royal Society’s archives likewise document extensive institutional connections between Fellows and the Royal African Company, including the Society’s own investment in the company.

None of this has to establish that Locke’s philosophy was consciously designed to justify slavery. It establishes something more historically significant: The philosophy of the proprietary individual and the institutions of colonial property developed within an overlapping political and commercial environment.

Hume and the descriptive firewall

Hume’s famous discussion of the transition from is to ought identified a genuine philosophical problem: Normative conclusions cannot simply be smuggled into descriptive premises without explaining the transition.

The distinction becomes problematic when institutions treat descriptive classification as independent of the normative consequences produced by that classification. Historical claims that Indigenous peoples "are" uncivilised or that Africans "are" naturally inferior could appear as empirical descriptions while simultaneously performing normative work concerning land, sovereignty and personhood. Hume himself made an explicitly racist claim about Africans in "Of National Characters" (1753).

The argument here is not that Hume invented racism, nor that his is/ought distinction was designed to protect it. The stronger proposition is functional: A distinction between description and obligation can become institutional insulation when the descriptive category itself determines who is entitled to what.

The same problem now appears in AI. "The system generated the decision" appears descriptive; but once that description determines who bears responsibility, description has become allocation.

From Philosophy to Infrastructure

These distinctions became institutional rather than remaining philosophical. Property law turned relations into rights; contract law turned relationships into enforceable obligations; corporate law created enduring legal entities; accounting converted complex social processes into transactions and accounts; administration converted people and environments into classifications; technical architecture converted actors and permissions into data structures, logs and APIs.

The forensic boundary is produced through this infrastructure. It identifies the unit to which causation, authority, liability and redress can be attached. In Spencer-Brown’s formulation, a distinction is constituted by drawing a boundary: The act of distinction produces the marked and unmarked sides that can subsequently be treated as separate objects of observation [2]. Modern institutions therefore do not merely describe boundaries; they construct operational boundaries. AI destabilises them [3].

From Substance to Relation—and Beyond

The relational alternative is considerably older than twentieth-century Western systems theory. Buddhist philosophy challenges the existence of an independently self-subsisting self through anattā and dependent origination. The person is not an isolated proprietor but a conventionally designated process arising through conditions [4].

Vedāntic traditions offer a different but complementary challenge. Advaita Vedānta does not merely replace substance with relation: It questions the ultimacy of separation itself. The apparent multiplicity of independently existing selves is grounded, ultimately, in Brahman. Viśiṣṭādvaita provides another formulation in which multiplicity remains real while existing in an inseparable relation to Brahman [5].

The distinction is important:

·         Substance asks: What thing is this?

·         Relation asks: Through what relations does it exist?

·         Field asks: What is the larger condition within which both thing and relation arise?

This is not a single philosophical tradition. Indigenous legal and ecological traditions, Buddhist philosophy, Vedānta, process philosophy, phenomenology, cybernetics and commons scholarship articulate different versions of this challenge [6-9].

Ostrom’s work, for example, demonstrated empirically that durable governance of common resources can emerge through rules, monitoring, collective choice and polycentric institutions rather than requiring either exclusive private ownership or central command.

Second-order cybernetics takes the next methodological step: The observer is not outside the system observed. Von Foerster’s distinction between the cybernetics of observed systems and the cybernetics of observing systems directly destabilises the assumption of an external, neutral classifier.

Non-classical logics such as Belnap–Dunn and tetralemmatic reasoning provide a complementary formal vocabulary where propositions need not be forced into a simple true/false binary. For AI governance, this matters because a system may be tool-like in one respect, agentic in another, both in a third and neither under a fourth description [10-13].

AI and the Reconstruction of Agency

AI makes the relational problem operational. Each agent may legitimately respond only to its local objective; yet their interaction can generate a capability no individual agent was expressly instructed to produce. The terminal actor is therefore not necessarily the complete cause [14].

The relevant question becomes: Who constituted the conditions under which the capacity to act emerged?

This is the central challenge to the conventional forensic boundary. Distributed agency does not imply disappeared responsibility. A relational principle of accountability follows:

Where consequential agency is constituted through relations among persons, institutions and artificial systems, accountability should not be determined solely by the entity producing the terminal act. Inquiry must extend to those who materially constituted, enabled, constrained, connected, delegated to, benefited from, or retained power to intervene in the capacity that produced the outcome.

This does not predetermine liability; it changes the unit of inquiry [15].

Conclusions

The historical argument is not simply that Newton, Locke and Hume were “wrong.” Their distinctions generated extraordinary intellectual and institutional power, but they also became embedded in an order whose categories of personhood, property and classification were entangled with colonialism, exclusion and racial hierarchy. The problem arises when historically situated distinctions are treated as complete descriptions of reality after the systems built upon them have exceeded their original unit of analysis.

Physics progressively destabilised the classical isolated object. Systems theory destabilised the isolated observer. AI now destabilises the isolated agent.

Physics had already undergone a comparable journey: From the classical description of discrete objects toward a quantum account in which observation, uncertainty, complementarity, probability and non-classical correlations challenged the adequacy of the isolated classical object. Just as Locke’s friend Isaac Newton gave way to Heisenberg, Bohr, Einstein and Schrödinger (and their re-conception via Eastern philosophy)): Physics had to reconsider the object; accordingly, AI now has to reconsider the agent.

Further, the movement is therefore not simply from substance to relation, but from substance, through relation, to field: From the entity, through the relations that constitute its appearance, toward the non-dual ground in which the distinction between entity and relation is ultimately dissolved.

What is emerging is best understood as a doctrine of recovery: Not the invention of relationality, but the recovery of intellectual resources in which the isolated, self-subsisting entity was never treated as the ultimate unit of reality.

AI does not prove Buddhism, Vedānta, or Indigenous ontology. It does, however, confront Western institutions with questions those traditions explored long before computation: What constitutes an agent, whether identity is independent, how action arises from conditions and whether the boundaries through which responsibility is allocated correspond to the systems that actually produce consequences.

The challenge for AI ethics is therefore not merely to decide whether artificial systems are tools, agents, or persons. It is to build institutions capable of recognising agency when no single entity contains the conditions that produced it.

Acknowledgement

During the preparation of this manuscript, the author(s) used Claude Sonnet 5.5 to check the language, readability, and grammatical accuracy of the text. After using this tool/service, the author reviewed and edited the content as needed and takes full responsibility for the final contents of the peer-reviewed publication.

Conflict of interest

The author declares no conflict of interest.