Read in:
ENITESFR

Ontology-First Computational Entities for Decision Systems

An ontology-first approach to designing decision systems where knowledge is structured as computational entities—typed, executable objects integrating definitions, variables, constraints, and decision logic.

Keywords
Document type
Research Paper
References
  • Gruber, T. (1993). A translation approach to portable ontology specifications. Knowledge Acquisition.
  • Newell, A. (1982). The knowledge level. Artificial Intelligence.
  • Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach. Pearson.
  • Pearl, J. (2000). Causality: Models, Reasoning, and Inference. Cambridge University Press.

A Formal Framework for Structuring Knowledge into Executable Decision Objects


Abstract

This paper introduces an ontology-first approach to designing decision systems, where the fundamental unit is not a page, document, or tool, but a computational entity—a structured, executable representation of knowledge that integrates definitions, variables, constraints, transformations, and decision logic.

Traditional information systems separate content from computation, leading to fragmentation between knowledge representation and actionable outputs. We propose a unified framework in which ontological structures directly define computational behavior, enabling deterministic, auditable, and scalable decision processes.

The architecture formalizes entities as typed objects with embedded rules, enabling automatic generation of calculators, evaluators, and decision interfaces. This paradigm supports multi-domain applications, including financial modeling, engineering calculations, policy compliance, and AI-assisted decision systems.


1. Introduction

Most digital systems treat:

  • Content as static text
  • Computation as separate logic

This separation leads to inconsistency, duplication, lack of traceability, and weak decision reliability. We propose a paradigm shift:

From pages to computational entities

In this new paradigm, each unit of knowledge is structured, typed, executable, and composable.


2. Problem Statement

Let traditional systems be defined as:

System = Content + Code

With weak coupling between the two. We redefine the system as:

System = {E_1, E_2, ..., E_n}

Where each E_i is a computational entity:

E = (D, V, C, R, T, O)

Where:

  • D: definitions
  • V: variables
  • C: constraints
  • R: rules
  • T: transformations
  • O: outputs

3. Computational Entity Model

3.1 Formal Definition

A computational entity is a function:

E: (V, C, R) → O

subject to:

C(V) = true

and:

O = T(R(V))

3.2 Properties

Each entity must be deterministic, auditable, versioned, composable, and context-aware.

3.3 Entity Schema (Example)

{
  "entity_id": "loan_payment",
  "type": "financial_calculation",
  "inputs": ["principal", "rate", "term"],
  "constraints": ["principal > 0", "rate >= 0"],
  "formula": "PMT = P * r / (1 - (1 + r)^-n)",
  "outputs": ["monthly_payment"]
}

4. Ontology-First Design

4.1 Ontology Layer

This layer defines entities, relationships, types, and hierarchies.

Example:

FinancialEntity
 ├── Loan
 ├── Mortgage
 └── Investment

4.2 Execution Layer

Transforms the ontology into calculators, decision engines, and evaluators.

4.3 Separation of Concerns

LayerResponsibility
OntologyMeaning
RulesLogic
EngineExecution
InterfacePresentation

5. Architecture

Figure 1 — Ontology-Driven System

Ontology Layer
     ↓
Entity Definitions
     ↓
Rule Engine
     ↓
Execution Engine
     ↓
Outputs (Calculators / Decisions)

Figure 2 — Entity Composition

Entity A (Inputs)
     ↓
Entity B (Transform)
     ↓
Entity C (Decision)

6. Decision Pipeline

We define a pipeline where each stage is governed by entity rules:

Input → Validation → Transformation → Evaluation → Output

7. Relation to Existing Paradigms

7.1 Knowledge Graphs

Represent relationships, but typically lack executable logic.

7.2 Rule-Based Systems

Executable, but often not semantically structured.

7.3 Machine Learning Systems

Predictive, but generally not interpretable.

7.4 Our Contribution

We unify ontology, rules, and execution into a single, cohesive structure.


8. Integration with Decision Systems

This framework integrates with Dual-Evidence Systems, scoring engines, and optimization pipelines. Each entity can independently produce features, evaluate constraints, and generate outputs.


9. Use Cases

  • Financial Calculators: Loans, ROI, taxation.
  • Engineering Systems: Structural calculations, safety checks.
  • Policy Engines: Compliance, regulation.
  • Decision Intelligence Platforms: Scenario simulation, optimization.

10. Formal Properties

10.1 Composability

E_3 = E_2(E_1(x))

10.2 Idempotence (Optional)

E(E(x)) = E(x)

10.3 Determinism

E(x) = y ⇒ E(x) = y

11. Limitations

  • Ontology design complexity
  • Rule explosion risk
  • Need for strict governance
  • Potential rigidity

  • Ontology Engineering: Gruber (1993) — Ontologies as specifications of conceptualizations.
  • Knowledge Representation: RDF / OWL frameworks.
  • Rule Systems: Expert systems, Prolog.
  • Decision Systems: Decision theory, Multi-criteria decision analysis.

13. Future Work

  • Automated ontology generation
  • Integration with LLMs
  • Hybrid symbolic and statistical systems
  • Self-updating entities

14. Conclusion

We propose a shift:

From content-driven systems to ontology-driven computational systems

This enables unprecedented consistency, scalability, traceability, and decision reliability across complex domains.


References

  • Gruber, T. (1993). A translation approach to portable ontology specifications. Knowledge Acquisition.
  • Newell, A. (1982). The knowledge level. Artificial Intelligence.
  • Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach. Pearson.
  • Pearl, J. (2000). Causality: Models, Reasoning, and Inference. Cambridge University Press.