Knowledge Synthesis
How data becomes something we can act on — distinct from Shannon information
Knowledge synthesis is how an organisation gains knowledge from data — not by violating the data processing inequality, but by ascending the capture stack: data (committed events) → information (structured statements with provenance) → knowledge (synthesized, aggregated, actionable models and rules). The data, information, and knowledge threads on the site develop this ladder; events are data, not knowledge, until synthesis occurs.
Knowledge is not Shannon information. Claude Shannon’s theory measures uncertainty in signals: entropy, mutual information, channel capacity. It answers: how many bits, how much surprise, how much is preserved along this channel? Shannon explicitly bracketed meaning and truth — his framework is syntactic and statistical, not a theory of justified belief, understanding, or what to do next.
Finding a pattern or formula reduces Shannon mutual information with the concrete L0 captures that produced it (compression, generalization — see logical inference). It can increase knowledge for agents and systems: a kinematic law, a promoted hypothesis from data, an ontology — L2-class artifacts with tier, provenance, and predictive use until fresh evidence confirms or refutes.
The framework uses Shannon where it belongs — epistemic uncertainty, atomicity limits, trust composition — and knowledge where synthesis, tier, and accountability belong. Confusing the two produces either dismissive “it’s all just re-encoding data” or overclaiming “mining created new ground truth.” Compression loses raw Shannon detail; synthesis creates knowledge — the subject of the planned Synthesizing Knowledge series (statistical, analytic, and graph-based methods). Downstream: Enabling Intelligence (intelligence as core dynamic capability), Towards Wisdom, Strategy and Beyond… (KBV + VRIO on knowledge assets).
A3 — open question: L0 ≈ data, L1 ≈ information, L2 (often) ≈ knowledge, L3 ≈ audience-specific presentation — to be developed alongside that series; facets and tier cut across the stack.
Corpus stance
A1 — adopted on this site: Names the positive frame for pattern discovery: knowledge gain, not Shannon information gain relative to raw inputs.