Publication Note

This paper forms part of the Authority, Provenance and Semantic Governance Research Series produced by Younis Group under the Search Sciences™ Research Programme. The series examines the structural conditions governing authority, provenance, and semantic integrity in AI-mediated information systems. 

White Paper No. 3 builds upon the admissibility framework defined in White Paper No. 1 and the verification-first architecture articulated in White Paper No. 2. It extends the research programme by addressing the structural governance of meaning within admissible knowledge environments. The paper introduces the concept of semantic governance and outlines the architectural requirements for deterministic entity design in AI-compatible systems. 

The intellectual foundations of this research programme are rooted in the Islamic Golden Age tradition of information science. The semantic governance model developed in this paper draws most directly upon Al-Farabi’s systematic classification of the sciences, which established hierarchical knowledge organisation as a structural requirement for the elimination of ambiguity. It also draws upon Imam Al-Bukhari’s provenance-chain methodology, Al-Khwarizmi’s systematic derivation of unknowns through defined procedure, and Ibn al-Haytham’s empirical auditability principles. These contributions are the structural antecedents of deterministic entity architecture. They are not background context. They are the governing intellectual premise from which the arguments in this paper are derived. 

This document contributes to an ongoing and cumulative research programme. The architectural principles described herein are implementation-agnostic and intended to inform scholarly discourse and infrastructure design. They should be read as part of a staged and methodologically structured body of work. 

Companion Publications 

White Paper No. 1: The Admissibility Problem in AI-Mediated Information Systems. Authority, Provenance and Semantic Governance Research Series. Search Sciences™ Research Programme. Younis Group, 2026. 

White Paper No. 2: Verification-First Architecture: Designing Pre-Interpretive Constraints for Authoritative Digital Systems. Authority, Provenance and Semantic Governance Research Series. Search Sciences™ Research Programme. Younis Group, 2026. 

Suggested citation: 

Younis, M. (2026) ‘Semantic Governance in Admissible Knowledge Systems: Deterministic Entity Architecture for AI-Compatible Environments’. White Paper No. 3. Authority, Provenance and Semantic Governance Research Series. Search Sciences™ Research Programme. Younis Group.

Abstract 

The preceding papers in this series identified the admissibility problem in AI-mediated information systems and articulated the architectural requirements for verification-first design. Verification, however, governs legitimacy rather than meaning. Once representations are rendered admissible, a further structural challenge remains: the deterministic organisation of semantic meaning prior to computation. 

This paper introduces the concept of semantic governance, defined as the formal constraint of ontological placement, attribute definition, and relational structure within admissible knowledge systems. It argues that without deterministic entity architecture, even verified information remains vulnerable to interpretive ambiguity and semantic drift. 

The intellectual foundations of this model are traced to Al-Farabi’s tenth-century classification of the sciences, which established that knowledge must be organised hierarchically and domain-specifically to prevent categorical ambiguity. Al-Farabi’s insight that ontological placement is a governance requirement — not a taxonomic convenience — is the direct intellectual ancestor of the deterministic entity architecture presented in this paper. This lineage is reinforced by Imam Al-Bukhari’s provenance methodology, Al-Khwarizmi’s systematic procedural resolution of unknowns, and Ibn al-Haytham’s conditions of empirical auditability. 

Drawing on knowledge graph theory, ontology design, and applied information science, the paper outlines a structured entity model capable of supporting AI-compatible reasoning whilst preserving governance integrity. It concludes that semantic governance, combined with verification-first architecture, constitutes the structural foundation of trustworthy, AI-compatible digital ecosystems.

1. Introduction 

Verification establishes legitimacy. It does not establish meaning. 

A digitally signed representation may be cryptographically authentic yet semantically ambiguous. AI systems operating on admissible inputs still require structured organisation to interpret entities consistently. The admissibility problem, addressed in White Paper No. 1, and the verification-first architecture developed in White Paper No. 2, together provide the conditions under which a representation may be admitted as legitimate. But admissibility is a precondition, not a destination. A further structural layer is required. 

The evolution of information systems has historically prioritised storage and retrieval over semantic determinism. As a result, knowledge graphs and enterprise data architectures often exhibit inconsistent classification, implicit ontological assumptions, undeclared attribute semantics, and unbounded relational expansion. These weaknesses become amplified within AI-mediated environments, where probabilistic inference operates on whatever structure — or absence of structure — the input layer provides. 

The Islamic Golden Age tradition of information science addressed this problem with considerable precision. Al-Farabi, writing in the tenth century, argued that the classification of the sciences was not merely a scholarly convenience. It was a structural requirement for the integrity of knowledge. A claim situated in the wrong domain, or attributed to an entity of unspecified scope, was not merely imprecise — it was structurally invalid. His hierarchical organisation of knowledge established that ontological placement, domain scope, and relational structure are governance constraints, not taxonomic preferences. 

Al-Farabi’s classification of the sciences was not a library index. It was a governance architecture for the integrity of knowledge. 

This paper recovers that architectural principle and applies it to the governance of semantic meaning in contemporary AI-mediated systems. Semantic governance is not a new invention. It is the structural application of a methodology formalised over a millennium ago. 

2. From Data Legitimacy to Semantic Determinism 

2.1 The Limits of Verification Alone 

Verification-first architecture, as developed in White Paper No. 2, ensures that authority is declared, identity is cryptographically bound, and representations are temporally auditable. These conditions are necessary. They are not sufficient. 

Verification does not answer the semantic questions that interpretive systems must resolve: 

• What is this entity? 

• Where does it sit within a conceptual hierarchy? 

• Which attributes are structurally valid for an entity of this type? 

• Which relationships are permissible within this domain? 

Without explicit constraints on these questions, meaning emerges through probabilistic inference rather than structural definition. An AI system operating on verified but semantically unstructured inputs will resolve these questions by pattern-matching against its training corpus. The result is interpretive drift: the entity is assigned the meaning most

statistically probable in the training environment, regardless of the meaning declared by its governing authority. 

Al-Farabi and the Governance of Classification 

Al-Farabi (872–950 CE) produced the most systematic classification of the sciences in the Islamic Golden Age tradition. His Ihṣaʼ al-ʿulūm — the Enumeration of the Sciences — organised knowledge into hierarchically structured domains, each with defined scope, permissible attributes, and boundaries of valid inference. 

His central insight was architectural: ambiguity is not an epistemic failure. It is a structural failure. A knowledge system that does not constrain the ontological placement of its entities cannot produce reliable conclusions, regardless of the quality of its individual claims. The structure of the knowledge system is as determinative of its outputs as the content of its assertions. 

This is precisely the condition that semantic governance addresses. Al-Farabi’s classification framework is the direct intellectual antecedent of deterministic entity architecture. 

2.2 Semantic Drift 

Semantic drift occurs when entity classifications change implicitly over time, attributes accumulate without ontological discipline, and relationships are inferred without declared scope. In large-scale systems, drift produces category confusion, attribute collision, and cross-domain ambiguity. 

Drift is not primarily a data quality problem. It is a governance problem. It occurs in the absence of structural constraints that would otherwise prevent implicit reclassification, unauthorised attribute extension, and undeclared relational expansion. In AI-mediated environments, drift is compounded: a system that learns from drifting representations will encode the drift as a structural feature of its world model. 

Admissible knowledge must therefore also be semantically constrained. Legitimacy without semantic determinism is a necessary but incomplete condition for trustworthy AI-compatible systems. 

3. Deterministic Entity Architecture 

A governed semantic system requires a structured model defining how entities are represented. This model must specify the conditions under which an entity’s identity, classification, attributes, and relationships are structurally valid. The following six components constitute the minimum requirements of a deterministic entity architecture. 

1. Canonical identity of the entity — a stable, unambiguous identifier that persists across systems and jurisdictions. 

2. Explicit hierarchical placement — the entity’s position within a defined ontological structure, specifying its type and governing domain. 

3. Verifiable attributes with defined scope — a constrained set of properties whose semantic meaning and data type are declared and governed. 

4. Structured capabilities or behaviours where relevant — explicit encoding of the actions, obligations, or services associated with the entity within its defined domain. 5. Declared interactions with other entities — relationships that are explicitly typed, directionally specified, and scope-limited. 

6. Traceable external references and identifiers — links to authoritative external registries or standards that corroborate the entity’s classification.

These structural components do not restrict domain diversity. They constrain representational ambiguity. An entity may belong to any domain and carry any set of semantically valid attributes — but the domain must be declared, the attributes must be governed, and the structure must be auditable. 

Al-Khwarizmi’s methodological contribution is directly relevant here. His systematic methods for resolving unknowns were premised on the principle that an unknown must be derived through defined procedure rather than approximated through estimation. A deterministic entity architecture applies the same principle to semantic structure: the meaning of an entity is not approximated through probabilistic inference. It is derived from declared structural constraints. 

The Six Components of Deterministic Entity Architecture 

• Canonical identity — stable, unambiguous, persistent. 

• Explicit hierarchical placement — type and governing domain declared. 

• Verifiable attributes — scope defined, data type governed, provenance traceable. • Structured capabilities — actions and obligations explicitly encoded. 

• Declared interactions — relationships typed, directional, and scope-limited. • Traceable external references — links to authoritative corroborating registries. 

4. Ontological Placement and Hierarchy 

4.1 Explicit Classification 

Every entity must occupy a declared position within a hierarchy. Implicit categorisation introduces ambiguity. Explicit ontological placement defines the permissible attributes of an entity, limits its relational scope, and enables cross-domain interoperability. 

Hierarchy is not merely taxonomic convenience. It is a governance constraint. This was Al-Farabi’s central argument in the Enumeration of the Sciences: each domain of knowledge has defined scope, and the placement of a claim within that domain determines which inferential rules apply to it. A claim placed in the wrong domain is not merely misclassified — it is governed by the wrong rules. The structural consequence is invalid inference. 

In contemporary AI-mediated systems, this failure mode is commonplace. Entities without declared ontological placement are classified by the interpretive system according to whatever category is most statistically probable. The result is systematic misattribution: entities are governed by inferential rules that do not apply to them, producing outputs that are coherent in appearance but structurally invalid. 

4.2 Cross-Domain Interoperability 

In federated environments, entities may exist across jurisdictions and domains. Consistent hierarchical definition enables deterministic mapping between systems, controlled federation, and semantic portability. Without structural alignment, federation produces incoherence: entities from different systems are merged or compared according to incompatible ontological assumptions, and the resulting outputs carry the ambiguity of both. 

Al-Farabi’s classification framework addressed precisely this problem. By establishing a shared hierarchical structure for the organisation of knowledge, it enabled scholars across different traditions and linguistic contexts to operate with shared ontological assumptions. The structural principle is unchanged in contemporary federated digital environments: interoperability requires shared ontological governance, not merely shared data formats.

5. Attribute Governance 

Attributes represent properties asserted about an entity. In ungoverned systems, attributes accumulate through incremental addition without ontological discipline. The result is structural entropy: entities carry conflicting, redundant, or semantically undefined properties that interpretive systems cannot resolve consistently. 

Governed attributes must satisfy four structural requirements. First, defined semantic scope: the meaning of the attribute must be explicitly declared and bounded. Second, declared data type constraints: the permissible values of the attribute must be specified. Third, verifiable provenance: the authority asserting the attribute must be identifiable and accountable. Fourth, temporal versioning: the historical states of the attribute must be inspectable and auditable. 

Ibn al-Haytham’s empirical methodology is structurally relevant here. His requirement that an observation be repeatable and independently verifiable applies directly to attribute governance: an attribute that cannot be traced to a declared authority, verified against a defined scope, and audited retrospectively does not meet the conditions of admissible knowledge. Attribute proliferation without governance is the semantic equivalent of the admissibility problem: it allows structurally invalid inputs to enter interpretive systems without constraint. 

An attribute without declared scope is not information. It is noise with the appearance of structure. 

A constrained attribute model ensures interpretive consistency. AI systems operating on governed attributes can distinguish between properties that are structurally valid for a given entity type and properties that are not. This reduces inferential ambiguity without requiring changes to model architecture. 

6. Relational Discipline 

Knowledge graphs derive power from relationships. The ability to traverse connections between entities enables inference at a scale and depth that flat data structures cannot support. However, unbounded relational expansion produces ambiguity that is structurally indistinguishable from the ambiguity of unstructured text: relationships without declared type, direction, or scope constraints are probabilistically interpreted rather than structurally resolved. 

Governed systems must define permissible interaction types, directionality of relationships, scope limitations, and cross-domain linking rules. These constraints are not restrictions on the richness of a knowledge graph. They are the conditions under which the graph’s inferential power can be reliably exercised. 

Al-Farabi’s hierarchical classification established that relationships between knowledge domains must be explicitly governed: the conditions under which a claim in one domain could be applied in another were not left to inference but defined by the structural rules of the classification system. The same principle applies to relational governance in contemporary knowledge graphs: explicit relational discipline reduces inferential ambiguity in AI systems and prevents the propagation of structural errors across federated environments. 

7. Actionable Semantics

Certain entities possess defined capabilities or structured actions. In enterprise and civic contexts, this may include regulatory obligations, operational capabilities, service availability, and procedural constraints. Encoding such capabilities within deterministic structures enables AI systems to reason over permissible actions without speculative inference. 

The structural requirement here is precision of scope. An entity’s capabilities must be bounded by its ontological type and governing domain. An AI system that infers capabilities from statistical patterns in its training corpus will systematically misattribute permissions and obligations to entities that do not hold them. Explicit capability encoding within a deterministic entity architecture prevents this failure mode. 

Actionable semantics must remain structurally bounded and auditable. This is Al-Khwarizmi’s methodological principle applied to the governance of permissible actions: an unknown — in this case, what an entity is permitted to do — must be derived through lawful systematic procedure, not approximated through probabilistic estimation. 

8. Machine Compatibility and Reasoning Stability 

AI systems interpret structured data differently from unstructured corpora. A deterministic entity architecture provides stable node definitions, reduced ambiguity in embeddings, clear relational pathways, and controlled inference boundaries. Together, these properties improve the structural reliability of AI-mediated reasoning without constraining computational innovation. 

Semantic governance does not eliminate probabilistic reasoning. It stabilises the input space within which probabilistic reasoning operates. The distinction is structurally significant. A probabilistic system operating on semantically governed inputs will produce outputs that are bounded by the structural constraints of the governance layer. A probabilistic system operating on ungoverned inputs will produce outputs that are bounded only by the statistical distributions of its training corpus — which may include semantically invalid, historically erroneous, or adversarially optimised content. 

Ibn al-Haytham’s empirical methodology required that the conditions of an observation be sufficiently controlled to isolate the phenomenon under investigation from confounding variables. Semantic governance performs an analogous function for AI systems: it controls the structural conditions under which inference operates, reducing the influence of confounding semantic ambiguity on the outputs of the system. 

9. Federation of Governed Semantics 

When combined with verification-first architecture, deterministic entity models enable federated knowledge systems of genuine structural integrity. In such systems, authority is cryptographically verified, semantic placement is structurally constrained, and relationships are explicitly governed. Federation becomes an exchange of admissible and semantically deterministic entities rather than loosely structured data. 

This constitutes a foundational layer for interoperable AI-compatible ecosystems. The conditions for trustworthy federation are not achieved by technical interoperability protocols alone. They require ontological alignment: shared structural commitments about what entities are, what properties they may carry, and what relationships are permissible between them. 

This is precisely the function that Al-Farabi’s classification of the sciences performed within the Islamic Golden Age scholarly tradition. It provided a shared ontological framework within which scholars from different linguistic and disciplinary traditions could operate with shared structural assumptions. The knowledge produced within that framework was portable

across its federated contexts because the ontological governance was shared. The same structural requirement applies to contemporary federated digital environments. 

10. Towards Structured Knowledge Frameworks 

The architectural principles described in this paper imply the need for a formalised semantic framework capable of defining entity structures, enforcing hierarchical discipline, governing attributes and relationships, and integrating with verification protocols. 

Such frameworks must remain domain-agnostic whilst supporting vertical-specific extensions. Their purpose is not to replace existing databases or graph technologies. It is to impose structural governance upon them: to introduce the conditions under which semantic meaning is governed rather than inferred, and under which AI-mediated reasoning operates on a structurally sound input space. 

The development of such frameworks requires the same independence of stewardship that verification-first architecture demands. Semantic governance standards cannot be controlled by commercial platforms whose revenue models depend on the current ungoverned state of the information environment. They must be publicly specified, iteratively refined through scholarly and civic engagement, and maintained independently of the interpretive and monetisation systems they are designed to govern. 

Verified but semantically ungoverned knowledge is a necessary condition for trustworthy AI systems. It is not a sufficient one. 

11. Conclusion 

Verification-first architecture restores representational legitimacy. It does not, on its own, restore semantic integrity. Legitimacy without semantic determinism leaves admissible systems vulnerable to interpretive ambiguity, classification drift, attribute entropy, and relational incoherence. The compounding effect of these failures in AI-mediated environments is not marginal. It is structural. 

The intellectual foundations of the solution to this problem were laid in the Islamic Golden Age tradition of information science. Al-Farabi’s hierarchical classification of the sciences established ontological placement as a governance requirement. Al-Khwarizmi’s systematic procedural derivation of unknowns established that meaning must be determined through defined structural rules rather than probabilistic approximation. Ibn al-Haytham’s empirical auditability conditions established that the validity of a representation depends on the structural conditions under which it was produced and can be inspected. Imam Al-Bukhari’s provenance methodology established that every claim must carry a verifiable chain of attribution. 

Together, these four scholars — working between the ninth and eleventh centuries — articulated the intellectual architecture that semantic governance requires. That architecture was not incorporated into the design of the modern web or the training environments of contemporary AI systems. The result is the compounding structural failure that this research programme exists to document and address. 

Admissible knowledge systems require both pre-interpretive verification of authority and deterministic governance of meaning. Together, verification-first architecture and semantic governance form the structural basis of trustworthy, AI-compatible digital environments. Subsequent research in this series will examine implementation models, federation dynamics, and applied sector case studies within governed knowledge systems.

References 

Al-Bukhari, M. (846 CE) Al-Jāmiʿ al-Ṣaḥīḥ (Ṣaḥīḥ al-Bukhārī). Compiled 846 CE. The foundational collection of authenticated hadith, incorporating the isnad methodology as a formal system of provenance verification and chain-of-transmission governance. 

Al-Farabi, A.N. (c. 952 CE) Ihṣaʼ al-ʿulūm (The Enumeration of the Sciences). Translated by Palencia, A.G. (1953). Madrid. The foundational systematic classification of the sciences establishing hierarchical knowledge organisation, domain-scoped authority, and ontological placement as structural requirements for the integrity of knowledge. 

Al-Khwarizmi, M. (c. 830 CE) Kitāb al-mukhṣar fī ḥisāb al-jabr waʻl-muqābala (The Compendious Book on Calculation by Completion and Balancing). Translated by Rosen, F. (1831). London: Oriental Translation Fund. The foundational text of systematic algorithmic procedure and the lawful resolution of unknowns through defined structural methods. 

Ibn al-Haytham, H. (c. 1011 CE) Kitāb al-Manāẓir (Book of Optics). Latin translation: De Aspectibus (c. 1200). The systematic application of empirical auditability, controlled observation, and repeatable verification as structural conditions of valid knowledge. 

Berners-Lee, T., Hendler, J. and Lassila, O. (2001) ‘The Semantic Web’, Scientific American, 284(5), pp. 34–43. 

Floridi, L. (2011) The Philosophy of Information. Oxford: Oxford University Press. 

Gruber, T.R. (1993) ‘A translation approach to portable ontology specifications’, Knowledge Acquisition, 5(2), pp. 199–220. 

Nickel, M., Murphy, K., Tresp, V. and Gabrilovich, E. (2016) ‘A review of relational machine learning for knowledge graphs’, Proceedings of the IEEE, 104(1), pp. 11–33. 

Shannon, C.E. (1948) ‘A mathematical theory of communication’, Bell System Technical Journal, 27(3), pp. 379–423. 

Version History

Version 1.0

Initial publication. March 2026. Islamic Golden Age intellectual genealogy formally integrated throughout. Al-Farabi’s classification of the sciences established as the governing intellectual antecedent of deterministic entity architecture. Semantic governance model developed from verification-first architecture defined in White Paper No. 2.