Provenance-Enhanced Statements in Knowledge Graphs

2026-06-13 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Knowledge Representation & Semantic Web · Depth: Expert, quick

Summary

The DEC framework is introduced to address limitations in current knowledge graph provenance models, which often treat attributed claims as semantically neutral, hindering reasoning over interpretations and hypotheses (capta) rather than observer-independent facts (data). DEC interprets provenance predicates as indicators of epistemic stance, organizing provenance-homogeneous statements into "cognitive worlds." By drawing on cognitive modal logics (doxastic, epistemic, conjectural), DEC characterizes locality, rationality, and controlled permeation between these cognitive worlds and a "factual core" or "reality." This approach enables principled reasoning over attributed content without collapsing disagreements into inconsistencies. The framework formalizes a DEC interpretation for RDF datasets, ensuring conservatism over RDF 1.2 semantics, and clarifies intensionality and identity, including the Superman paradox. It is illustrated using common Semantic Web representations like named graphs, quoted triples/RDF-star, and reification. A prototype DEC reasoner, implemented as a Fuseki dataset module, supports controlled factualisation and explicit detection of disagreements and delusions.

Key takeaway

For research scientists building knowledge graphs with attributed claims or interpretations, the DEC framework offers a robust method to manage epistemic uncertainty. You can integrate provenance as a semantic indicator of stance, allowing principled reasoning over disagreements without collapsing them into inconsistencies. Implement DEC's "cognitive worlds" approach to formalize how claims relate to a factual core, enhancing analytical capabilities and detecting delusions.

Key insights

DEC interprets knowledge graph provenance as epistemic stance, enabling principled reasoning over attributed claims and disagreements without collapsing into inconsistencies.

Principles

• Provenance predicates signal epistemic stance.
• Group similar claims into "cognitive worlds."
• Reason over attributed content using modal logics.

Method

DEC interprets provenance predicates as epistemic stance indicators, grouping statements into "cognitive worlds." It applies cognitive modal logics to define locality and permeation rules between these worlds and a factual core, enabling principled reasoning over attributed claims.

In practice

• Formalize DEC interpretation for RDF 1.2.
• Apply to named graphs, RDF-star, reification.
• Use reasoner for factualisation, detecting delusions.

Topics

• Knowledge Graphs
• Provenance Models
• Epistemic Stance
• Cognitive Modal Logics
• RDF Semantics
• Semantic Web Reasoning

Best for: AI Scientist, Research Scientist

Related on AIssential

• How Big Is A Dragon? — Knowledge graphs should manage assertions with context, not just facts, to reflect varying truths.
• The Open World/Closed World Conundrum — The semantic web is shifting from Open World to Closed World Assumptions, driven by practical system demands.
• Strikingness-Aware Evaluation for Temporal Knowledge Graph Reasoning — Evaluating Temporal Knowledge Graph Reasoning requires distinguishing rare, "striking" events from common ones.
• Formalizing the DATASUS RTS: An Ontological Model for a Resource Description Framework Knowledge Graph — Transforming relational health data into a W3C-compliant knowledge graph improves interoperability and semantic querying.
• VeriGraph: Towards Verifiable Data-Analytic Agents — Explicitly constructing heterogeneous evidence graphs enables verifiable reasoning in LLM-based data-analytic agents.
• R-Mod: Minimal Structural Revision of S5 Epistemic Models — R-Mod minimally structurally revises S5 epistemic models to accommodate new knowledge, addressing limitations of belief revision.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.