Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex

Andrew Miracle; Phyl Georgiou; Christine Okoth; Ese Monday

cs.AINov 3, 2026

Toward a Compositional Knowledge Architecture for a Shared Organizational Cortex

Authors: Andrew Miracle, Phyl Georgiou, Christine Okoth, Ese Monday

Organizations: Robin Inc. · Nosen

Abstract

Organizations accumulate evidence in conversations, documents, decisions and relationships, but they act on interpretations: judgements assembled from several observations, resting on assumptions, and revised as circumstances change. Existing retrieval, knowledge graphs and agent memory keep the records and lose the interpretation. We propose a four-layer architecture, Signals, Domains, Claims and Dimensions, in which interpretations persist with their justifications and standing, a truth-maintenance network revises that standing as evidence changes, and validated, versioned Dimensions position interpretations on named axes. We call the result a shared organizational cortex. In a registered retrieval experiment, the Dimension term gains 0.028 nDCG@10 (p = 0.002). In system-level comparisons, the full retrieval configuration improves nDCG@10 by 0.081 over Signals-only and 0.053 over graph-augmented Signals-only retrieval (both reported p < 0.001). The architecture gives organizations a persistent, revisable basis for shared judgement.

Figures & tables

Figure 1: The architecture at a glance. Entries supply source material from which multiple Signals preserve atomic evidence and provenance. Domains provide contextual scope. Multiple Signals converge through justification into persistent Claims carrying assumptions and standing; Claims may also depend on prior Claims and are positioned on named Dimensions. Truth maintenance operates over justification, conflict and supersession dependencies at each maintenance cycle. People and agents consume the maintained interpretation with its grounds and geometric positions.

Figure 2: Signal admission and Domain maintenance. Only span-entailed candidates become Signals. Supersession changes support; reclassification, revision, splitting and retirement change scope or lifecycle. Affected Claim homes are reindexed, and staged state becomes visible at atomic commit.

Table 1. The four layers and the section that specifies each.
Layer	Unit	What it holds	Role
Signals	One atomic proposition extracted from one Entry	The proposition, its span and provenance, an embedding, its Domains	The evidence the rest of the system interprets; entailed by its Entry alone
Domains	One persistent semantic boundary	A name and description, a classifier threshold, an active set of Dimensions	Semantic routing: classify Signals and scope queries
Claims	One proposition no single antecedent entails	Justifications, context, standing, home Domains	Kept interpretations, relabelled as their evidence changes
Dimensions	One named, versioned axis	A value space, prototypes, a projection convention, a version	Position Claims and queries for reranking and for the contradiction check’s candidate filter

Table 3

Figure 3: Claims begin proposed, become accepted by review or rule, become contested under live conflict, retract on support loss or rejection, and become superseded on revision; a propagated retraction restored to support returns proposed. Panel A contains retrievable proposed, accepted and contested Claims, with penalties on proposed and contested; panel B contains retracted Claims kept for audit and superseded Claims reachable from successors.

Figure 4: Claim support structure. Within justification

j_{1}

, Signal A and Signal B are jointly required (AND);

j_{2}

, supported by a prior Claim and Signal C, is an alternative support path (OR). Unevidenced assumptions remain in context outside the evidence dependency graph. Claim C retains its interpretation, standing and history. The Claim is supported when every antecedent in at least one justification is supported; conflict and overrides then determine standing.

Figure 5: Dimension qualification, activation and projection lifecycle. A candidate supplies the seven-field contract and passes schema and qualification checks before validation. Useful or pin activation, inactive validation, replacement review, parking and human retirement remain distinct. Version edits either stale projections for exclusion and backfill or retain values while coefficients are refit.

Table 2. Dimension qualification matrix. The criteria test distinct properties of a candidate Dimension and its projection procedure. Passing them establishes operational consistency under the stated tests.
Criterion	Qualification procedure	Initial pass bar or timing	Caveat or consequence
Measurable	Project $S$ independently using different prompt phrasings or the two projector models named under Stable	$r_{s}\geq 0.7$	Tests rank agreement; the interval spacing assumed by $\delta_{d}$ and betweenness is untested.
Covered	Measure the share of $S$ with relevance $r_{d}\geq\tau_{r}$	At least $20\%$	Below the floor the candidate fails qualification.
Distinct	Compare the candidate definition embedding with every existing Dimension definition in the Domain	Cosine similarity $<0.9$	Prevents near-duplicate definitions
Non-redundant	Correlate positions with each active Dimension on Claims relevant to both	$\|r_{s}(\pi_{d},\pi_{d^{\prime}})\|<0.8$	Failure against exactly one axis creates a replacement candidate for human review.
Stable	Re-project with a different model at validation and after time or corpus growth; repeat after any version bump that stales projections	Meet Measurable at validation and again after 30 days or $20\%$ Claim-count growth, whichever comes first	Tests agreement across projector and time changes

Table 7

Figure 6: Complementary relations. Panel A contrasts embedding proximity with separated funding-availability positions. Panel B adds positional Dimension features to Domain-scoped base retrieval and subtracts the proposed-or-contested standing penalty.

Table 3. Essential model responsibilities and output constraints.
Name	Responsibility	Input	Permitted output	Constraint
Extractor	Propose atomic evidence from an Entry	One Entry	Candidate propositions with source spans	A proposal is stored only after the Labeller confirms span entailment; Extractor recall is not tested
Deriver	Compose an interpretation from several antecedents	A request with retrieved Signals and Claims	A Claim proposition with justifications and context	It cannot set standing
Labeller	Judge entailment, support or contradiction	A span or object with a candidate proposition	Entailed or not; supports, contradicts or unrelated	It cannot write propositions
Projector	Position a Claim or query on one Dimension	One proposition or query and one $\Pi_{d}$	One value in $V_{d}$ and one relevance	It cannot produce prose or take several Dimensions per call
Domain classifier	Assign contextual scope	One Signal or query and declared Domain descriptions	One probability per declared Domain	It cannot name an undeclared Domain

Table 9

Figure 7: Foundation X lifecycle. Signals justify and position the Claim at

T_{0}

; an agent reuses its grounds, assumptions, accepted standing and position at

T_{1}

. At

T_{2}

, independent contradiction creates conflict while supersession removes support. The

T_{3}

commit contests the first branch and retracts the second if no justification survives; only retraction propagates support loss. At

T_{4}

, consumers receive the updated standing.

Table 4. The systems this section compares, on the properties of our design claims.
System	Knowledge unit	Invalidation and standing	Multi-axis reranking	Persistence and update model
Retrieval with per-request reasoning (the alternative)	Retrieved Signals and documents, nothing derived is kept	None: there is no persisted statement to invalidate	The base retrieval score only	Interpretations are derived per request and discarded
GraphRAG and LightRAG	An entity graph; claims extracted from text units; community summaries	Extracted claims are produced once with a status fixed at extraction, and are not revised when the source changes	No Dimension term	Batch indexing; summaries pregenerated for global questions
Zep (Graphiti)	Temporal fact edges in a knowledge graph	Edges are invalidated on overlapping contradictions	No Dimension term	The graph is updated incrementally as facts arrive
This architecture	Signals, and Claims with justifications over them	Standing is relabelled by propagation through justifications: conflict contests, support loss retracts, a revision supersedes	The Dimension term over validated, versioned Dimensions, plus the standing penalty	Interpretations persist and are maintained as evidence changes

Table 11

Figure 8: Evaluation scope and evidentiary status. Pilot observations characterize decision-call behavior and annotation variability. The registered retrieval experiment provides the reported empirical results, supplemented by a post hoc facet-only comparison. The remaining registered components are specified in the evaluation protocol but fall outside the empirical scope of this paper.

Table 5. Dimension authoring agreement: the five matched pairs of Dimensions that annotators A and B authored independently, with positions compared over the Claims both marked relevant. Dice share is per Domain and repeats for its pairs; alpha is Krippendorff alpha on positions. Pair abbreviations are DS, decision settlement; AM, adoption maturity; MC, mechanism continuity; DAO, direct-access openness; DAV, direct-access viability; SB, switching burden; and FC, financing constraint. Definitions in Appendix B.
Domain	Pair	Axis match	Dice	alpha	Spearman	Coverage (A / B)	n both relevant
PEP	DS / AM	same	0.8	0.86	0.94	0.93 / 1.00	28
PEP	MC / MC	same	0.8	0.68	0.67	0.80 / 0.97	24
Enron	DAO / DAV	same	0.4	0.38	0.48	0.47 / 0.97	14
Enron	SB / DAV	partial, reversed	0.4	0.74	0.83	0.33 / 0.97	10
Enron	FC / DAV	partial, reversed	0.4	0.38	0.56	0.50 / 0.97	15

Table 13

Figure 9: Relevance-grade agreement between annotators on the 36 query-Claim pairs per Domain: Cohen kappa, exact agreement and ordinal alpha on grades, for PEP and Enron. Definitions in Appendix B.

Table 6. Judgment noise on the two Domains, pilot observations only. Definitions in Appendix B.
Measure	PEP	Enron	What it describes
SD across queries of the nDCG difference between annotators	0.063	0.086	the judgment part of the per-query variance in the registered retrieval tests
Pooled per-item grade SD (0-3 scale)	0.471	0.540	relevance-grade annotation
Pooled per-item position SD, range over pairs	0.126 to 0.220	0.154 to 0.256	position and projection agreement

Table 15

Table 7. The Domain-classifier decision-call runs: two models and two call shapes over the pilot’s 12 queries and 177 Signals. Argmax errors count calls whose highest-probability Domain differs from the pilot’s assignment; saturated counts returned probabilities of exactly 0 or 1.
Model	Shape	Items	Argmax errors	Saturated
clef	forced two-way choice	12 queries	1	0 of 24
jev	forced two-way choice	12 queries	2	18 of 24
clef	per-Domain yes/no	12 queries	1	0 of 24
jev	per-Domain yes/no	12 queries	1	0 of 24
clef	per-Domain yes/no	177 Signals	0	0 of 354

Table 16

Table 8. The nine arms on the 272 pooled held-out queries: mean nDCG@10 with the per-query standard deviation in parentheses, mean Recall@50 and mean MRR. We added arm (ix) post hoc. Definitions in Appendix B.
Arm	nDCG@10 (SD)	Recall@50	MRR
(i) hybrid baseline	0.421 (0.199)	0.609	0.400
(ii) hybrid plus standing penalty	0.453 (0.193)	0.627	0.417
(iii) full reranker	0.481 (0.199)	0.656	0.441
(iv) permutation control	0.456 (0.213)	0.637	0.422
(v) Signals-only	0.400 (0.222)	0.582	0.378

Table 17

Figure 10: Retrieval ablation comparison on the 272 pooled held-out queries, using the mean nDCG@10 values reported in Table 8; higher is better. The bars show absolute scores for each arm. Facet-only is a post hoc attribution arm. Statistical interpretation, registered contrasts and uncertainty remain in Table 9. Definitions in Appendix B.

Table 9. The gains on the pooled held-out queries, nDCG@10: the arms compared, the query count, the difference of means, the interval and p-value as recovered from the record, the registered bar and the outcome. The Dimension gain’s interval is not centred on its difference and we could not re-derive it. Definitions in Appendix B.
Gain	Arms	n	Difference	Interval	p	Bar	Outcome
Dimension (primary)	(iii) minus (ii)	272	0.028	0.014 to 0.056	0.002	positive and significant, one-sided 0.03 per look	passes
Standing	(ii) minus (i)	272	0.032	0.004 to 0.052	0.011	Holm family	passes
System, Signals-only	(iii) minus (v)	251	0.081	0.058 to 0.112	$<0.001$	Holm family, both must pass	passes
System, graph-augmented on Signals-only	(iii) minus (viii)	251	0.053	0.033 to 0.083	$<0.001$	the same clause	passes
Domain scoping	(i) minus (vi)	272	0.049	0.024 to 0.076	$<0.001$	Holm family	passes

Table 19

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Table 10. Metric definitions: one name per quantity, its definition, the population it is computed over, and where the paper reports it. Formulas marked “below” follow the table.
Quantity	Definition	Computed over	Reported in
Dice share of matched Dimensions	$(m_{A}+m_{B})/(n_{A}+n_{B})$ ; below	The Dimensions authored on both sides in each Domain; only same-axis pairs count	Table 5; Section 8.3
Krippendorff alpha on positions	$1-D_{o}/D_{e}$ ; level below	Each matched pair, over the jointly relevant Claims	Table 5
Spearman correlation on positions	Pearson correlation of the two rank sets, ties at mean rank	Each matched pair, over the jointly relevant Claims	Table 5; Section 8.3
Coverage	Share of the Claims positioned, not marked NR	Each annotator and Dimension, over the Domain’s 30 Claims	Table 5; Sections 5.4 and 8.3
Cohen kappa on grades	Unweighted; below	The 36 graded pairs per Domain	Section 8.3; Figure 9; Section 9

Appendix

Table 20

Knowledge Representation

Knowledge representation studies how to encode facts, concepts, rules, context, and uncertainty so machines can interpret and reason over them.

See all 131 papers

Retrieval Is Not Enough: Why Organizational AI Needs Epistemic Infrastructure
Apr 13, 2026Federico Bottino, Carlo Ferrero, Nicholas Dosio +1Knowledge Representation Epistemic Uncertainty
Kakashi Ventures Accelerator (KVA) · Massachusetts Institute of Technology
Knowledge-Centric Information Systems
Jul 1, 2026Mariano Garralda-Barrio
Independent Researcher Lleida, Spain
The Missing Knowledge Layer in Cognitive Architectures for AI Agents
Apr 13, 2026Michaël RoynardCognitive Architectures for AI Agents Agent Memory
Independant Researcher

Apr 13, 2026J/K move · Enter open · S save

Retrieval Is Not Enough: Why Organizational AI Needs Epistemic Infrastructure

Federico Bottino, Carlo Ferrero, Nicholas Dosio, Pierfrancesco Beneventano

Kakashi Ventures Accelerator (KVA) · Massachusetts Institute of Technology

Knowledge Representation Epistemic Uncertainty Retrieval-Augmented Generation Knowledge Graphs AI Agent Evaluation

Open paper AskPDF arXiv

Organizational knowledge used by AI agents typically lacks epistemic structure: retrieval systems surface semantically relevant content without distinguishing binding decisions from abandoned hypotheses, contested claims from settled ones, or known facts from unresolved questions. We argue that the ceiling on organizational AI is not retrieval fidelity but epistemic fidelity--the system's ability to represent commitment strength, contradiction status, and organizational ignorance as computable properties. We present OIDA, a framework that structures organizational knowledge as typed Knowledge Objects carrying epistemic class, importance scores with class-specific decay, and signed contradiction edges. The Knowledge Gravity Engine maintains scores deterministically with proved convergence guarantees (sufficient condition: max degree

< 7

; empirically robust to degree 43). OIDA introduces QUESTION-as-modeled-ignorance: a primitive with inverse decay that surfaces what an organization does not know with increasing urgency--a mechanism absent from all surveyed systems. We describe the Epistemic Quality Score (EQS), a five-component evaluation methodology with explicit circularity analysis. In a controlled comparison (

n{=}10

response pairs), OIDA's RAG condition (3,868 tokens) achieves EQS 0.530 vs.\ 0.848 for a full-context baseline (108,687 tokens); the

28.1\times

token budget difference is the primary confound. The QUESTION mechanism is statistically validated (Fisher

p{=}0.0325

, OR

=21.0

). The formal properties are established; the decisive ablation at equal token budget (E4) is pre-registered and not yet run.

Abstract

Knowledge Representation

Explore similar work

Retrieval Is Not Enough: Why Organizational AI Needs Epistemic Infrastructure

Knowledge-Centric Information Systems

The Missing Knowledge Layer in Cognitive Architectures for AI Agents