Retrieval Is Not Enough: OIDA Framework Introduces Epistemic Infrastructure for Organizational AI Agents

arXiv·medium signalXBlueskyLinkedInCopy link

Bottino et al. (arXiv 2604.11759, April 13) argue that the ceiling on organizational AI agents is epistemic fidelity, not retrieval fidelity. Their OIDA framework structures organizational knowledge as typed Knowledge Objects with epistemic class, importance scores with class-specific decay, and signed contradiction edges. The key innovation is QUESTION-as-modeled-ignorance — a primitive with inverse decay that surfaces what an organization doesn't know with increasing urgency, a mechanism absent from all surveyed RAG systems. Full-context baseline achieved EQS 0.848 vs OIDA's 0.530 at 28x fewer tokens.