Synthesizing Knowledge

Series preview — how explicit information becomes something we can act on

MARKO ŽIVKOVIĆ Draft

This page is a series preview. Articles listed below are planned, not yet published.

Thesis

Knowledge synthesis is not Shannon information gain from re-reading the same captures. It is the deliberate ascent from data (committed events) through information (structured, provenance-carrying statements) to knowledge — rules, models, ontologies, and graphs ready to act on. Pattern-finding compresses raw detail and creates knowledge; tier and provenance track how far that synthesis is justified.

Genesis established why processing cannot invent past ground truth. This series establishes how organisations still learn — honestly — from what they recorded.

What this series will cover

The capture stack and atomicity

Develop the A3 — open mapping from Genesis Part III: L0 ≈ data, L1 ≈ information, L2 (often) ≈ knowledge, L3 ≈ presentation for different audiences. The facets (commitment, tier, uncertainty) cut across the stack — the rhyme is intentional, not a collapse.

Methods of synthesis

Method	What it does	Typical placement
Statistical discovery	Patterns, correlations, fitted models from captures	In-sample pattern → holdout; hypothesis from data
Analytic derivation	Laws, proofs, closed-form models from structured premises	Logical inference at L1/L2; tier 2 when locked to cited substrate
Knowledge organisation	Ontologies, knowledge graphs, typed relations over statements	L2 explicit synthesis; events and triples as inputs, not substitutes

Each method shares the same discipline: causal chain to L0, tier on every claim, no tier laundering from fit to fact.

From hypothesis to trusted model

Validation paths the series will treat in full: holdout and prospective test, replication, demotion on misprediction, and honest registration of HypothesisRegistered / model-version commits. Running those validations is experiment architecture — developed in Enabling Intelligence. Connection to Event Sourced Science and the knowledge thread.

Harness and infrastructure

Structure is what converts capability into trustworthy synthesis — schemas as epistemic commitment, event logs as institutional memory, RAG and graphs as grounded retrieval (not inference-as-fact). The epistemic harness is that structure generalised: not only an LLM wrapper (citations, review, version pinning) but socio-technical design — domain authority, tier routing, governance at the commitment boundary, experiment architecture — for human organisation, AI, and hybrids together. Overlaps with planned capture-stack articles from the long editorial roadmap; this series owns synthesis, not faithful recording alone. Enabling Intelligence carries harness design for reasoning and calibration.

Planned articles (working outline)

Capability without structure — why synthesis without epistemic harness produces confident decks, not knowledge
Data, information, knowledge — the three layers in practice; events are data until structured
Statistical synthesis — mining, overfitting, multiple comparisons, promotion criteria
Analytic synthesis — derivation, laws, tier-2 claims versus L-level placement
Graphs and ontologies — explicit knowledge organisation; weakest-link tier in traversals
Presenting knowledge — L3 rendering without concealing uncertainty
Collective memory — append-only institutional records versus CRUD knowledge loss

Order and titles will change as articles are written.

Where to start meanwhile

Genesis Part III — facets, L0–L3, uncertainty
What Information Theory Says — Shannon limits versus knowledge synthesis
Capture ladder map — data → information → knowledge
Threads: data, information, knowledge
Epistemic harness — structure for synthesis at org scale, not only LLM wrappers
VRIO framework — strategic Organized (O) as harness for knowledge advantage
Knowledge-based view (KBV) — knowledge as prime resource; intelligence as prime dynamic capability

Status: planned. First substantive article not yet published.