How a Clinical Judgment Becomes Community Intelligence
Follow a single piece of medical reasoning—from one physician's mind, through verification, into circulation, until it lives as part of a global community's collective judgment.
By Chao Li
Streptococcal pneumonia, as a worked example
↓ Begin ↓
i. The Encounter
A Tuesday morning. One patient.
Eight minutes of attention. A throbbing headache, fever, productive cough. The physician listens.
Clinical Note · Dr. M., São Paulo
"Streptococcal pneumonia is a respiratory disease caused by Streptococcus pneumoniae bacteria. It presents acutely with fever, productive cough, and bilateral infiltrates. Treatment requires antibiotics, and penicillin remains the first-line agent — unless the patient is allergic."
Date
Tue, June 17 · 09:42
Site
Primary Care · São Paulo
Patient
F · 34 · presenting cough + fever
Physician
Dr. M. · CRM-SP licensed
Window
~ 8 minutes
Prior knowledge
ICD-11, antibiotic stewardship
Decision pending
Treatment selection
A piece of community-validated judgment is about to enter, and another is about to leave. Watch closely.
ii. Extraction
Prose into structure, without loss.
The same clinical knowledge in two forms. One travels through the human mind; the other travels through a community.
Form A · Human-Readable Prose
Streptococcal pneumonia is a respiratory diseasecaused by Streptococcus pneumoniae bacteria. It presents acutely with fever, productive cough, and bilateral infiltrates. Treatment requires antibiotics, and penicillin remains the first-line agent — unless the patient is allergic.
DCG · extract
Form B · DCG Tuples— machine-precise · network-routable
Hover any phrase on the left to find its DCG counterpart on the right. The translation is not summary. It is structural equivalence — and one of these forms can travel through a network at the speed of inference.
iii. Verification
Two strands. One verifier.
No DCG enters the network alone. Three independent confirmations must concur — and the human is not the inspector. The human is the rule of pairing itself.
The DCG is a double helix.
Two technical strands check the same assertion from opposite directions. The human verifier is what locks them.
Strand I · Symbolic
Logical well-formedness. The arity of the tuple. The internal coherence of @domain indexing across tuples.
Asks: is this assertion structurally legal?
Strand II · Neural
Semantic alignment. Does the structured form actually capture what the prose meant? Does the extraction reflect real clinical intent?
Asks: is this assertion semantically faithful?
Pairing · Credentialed Human
A licensed practitioner — Dr. M. — confirms that this DCG represents the clinical judgment she actually holds. Her credentials, her institution, her professional standing are the binding rule.
The human is not an overseer of the helix. The human is the rule by which the strands pair.
iv. Micro-loop · seconds
The encounter is itself a metabolic cycle.
Within the eight-minute consultation, validated community judgment arrives, is applied, and the new judgment formed in this encounter is staged for verification.
Assistant queries the network with the extracted symptom DCGs. The community returns: relevant validated judgments from Lima cluster, last month.
09:42:14 · t = 14s
A boundary marker surfaces: in this specific symptom triad, one common first-line treatment shows reduced effectiveness. Dr. M. registers it.
09:48:30 · t = 6m 30s
Dr. M. confirms diagnosis, adjusts plan to account for the boundary, discusses with patient. Decision is made — informed by what was learned everywhere last month, in a form she could trust.
09:50:00 · t = 8m
Encounter closes. The new clinical judgment formed today — including any local refinement — enters the verification queue. The micro-loop is complete.
Cycle time
8 min
the time of one consultation
Knowledge accessed
~ ∞
whatever the community holds, surfaced in seconds
Patient experience
Same
eight minutes, a doctor who listened
v. Meso-loop · weeks to months
The institution refines what the encounter discovered.
A pattern that one physician noticed becomes a hypothesis the institution tests. Twenty-three peer clinics weigh in. The community arrives at consensus.
Month 0
Dr. M.'s clinic accumulates DCGs. The medical director, reviewing the quarter, notices: a community-validated medication adjustment shows different effectiveness here than elsewhere.
Month 1 · week 1
Neural analysis suggests correlation with a demographic feature absent from the original validation domain. Symbolic verification confirms statistical significance. Physicians review and concur.
Month 1 · week 2
Clinic encodes this as a refined DCG: original judgment + new boundary marker, indicating the demographic context where it applies differently. The refinement enters the network as a proposal.
Month 1 · weeks 3–5
23 peer clinics in similar populations test the proposal against their own data. 18 confirm. 3 partial match. 2 no match.
Month 2
Consensus. The original DCG stands, with an added domain marker for the population where the refinement holds. The refinement is now community intelligence.
Cycle time
~ 8 weeks
institutional learning, federated validation
Verification
23 / 5
peer clinics tested · consensus threshold met
Outcome
Refined
a sharper boundary, the community's now
vi. Macro-loop · continuous
The community itself metabolizes, without center.
A boundary discovered in Lagos sharpens treatment in Lima. A refinement validated in São Paulo updates clinical practice in Mumbai. No central authority. No data crosses borders. Only structured judgment.
from multi-institutional consensus to global propagation
Sovereignty preserved
100%
no patient data crosses institutional borders
Drift
None
each propagation requires consensus; nothing imposed
vii. A Closer Look
The error is the boundary.
In the example you have just walked through, two DCGs sit side by side. One says "do this." The other says "but not here." They are equals — and the second is the more valuable resource.
Positive judgment · strategy
Penicillin is the first-line agent for streptococcal pneumonia.
{ from: Penicillin, rel: strategy, domain:@Treatment@FirstLine, to: Streptococcal_Pneumonia }
This is the kind of knowledge medical literature publishes, guidelines codify, and physicians remember. It is well-served by the existing system.
Boundary marker · contrasts_with
— except where the patient has a penicillin allergy.
{ from: Penicillin, rel: contrasts_with, domain:@Treatment@Contraindication, to: Penicillin_Allergy }
This is the kind of knowledge the existing system loses: buried in patient records, lost when the physician retires, never aggregated as a community asset. In DCG, it travels as a peer of the positive rule — not a footnote, not an afterthought, but its own first-class structure.
Negative knowledge — the precise edge at which a treatment fails — is the most fragile resource any community holds. By treating it as a boundary marker rather than as contamination, the DCG protocol makes it first-class for the first time.
— · The Ground Beneath the Path · —
This is what general intelligence looks like when it is the community's own: judgment that travels, survives error, improves with use, and remains ours.