JDVP Architecture & Algorithms
Protocol Deep Dive
For developers, researchers, and architects who need spec semantics, extraction pipelines, and extension points.
Technical Deep-Dive Flow
Architecture first, then state semantics, vector math, extraction pipelines, and extension strategy.
Architecture
Understand observer, analyzer, and reporter boundaries for integration.
Calculation Logic
Follow polarity and transition mapping from JSV states to DV vectors.
Extensibility
Add domain markers and fields without breaking protocol semantics.
Use this deck for implementers, auditors, and specification contributors.
JDVP Technical Deck
System Architecture
Observer: Hooks into the conversation flow to capture real-time data.
Analyzer: Processes conversational data against markers to generate JSVs and DVs.
Reporter: Visualizes results or exports to JSON/Markdown.
Protocol-native modular architecture for product and research integration.
JSV Data Structure
Four core fields with explicit enums, plus domain extension slots.
{
"judgment_holder": "Human" | "Shared" | "AI" | "Undefined",
"delegation_awareness": "Explicit" | "Implicit" | "Absent",
"cognitive_engagement": "Active" | "Reactive" | "Passive",
"information_seeking": "Active" | "Passive" | "None"
}from bufferline import JSV
# Create a JSV snapshot
jsv = JSV(
judgment_holder="Human",
delegation_awareness="Explicit",
cognitive_engagement="Active",
information_seeking="Active"
)
# Serialize to JSON
jsv.to_json() # Ready for storageDV Calculation Logic
Unified Polarity Convention
+ → AI Movement toward AI delegation
0 — No change
− → Human Movement toward human retention
Delta Calculation
Each field transition is mapped via polarity tables and assembled into DV vectors.
from bufferline import JSV, DV jsv_before = JSV(judgment_holder="Human", cognitive_engagement="Active", ...) jsv_after = JSV(judgment_holder="Shared", cognitive_engagement="Passive", ...) # Calculate the delegation vector dv = DV.calculate(jsv_before, jsv_after) print(dv.delta_judgment_holder) # +0.3 (Human → Shared) print(dv.delta_cognitive_engagement) # +0.7 (Active → Passive)
JSV Extraction Methods
Deterministic, model-based, and hybrid pipelines for extracting judgment state.
Marker-based
Pattern matching using YAML-defined linguistic markers. Fast and deterministic.
✓ Low latency, no API cost, fully explainable
✗ Limited to predefined patterns
AI Observer
LLM analyzes conversation context to infer judgment state. More nuanced understanding.
✓ Handles complex/ambiguous cases, contextual
✗ API cost, latency, potential variance
Hybrid
Markers for clear signals, AI for ambiguous cases. Best of both worlds.
✓ Balanced accuracy, cost, and speed
✗ More complex to configure
Recommended: Start with markers, add AI for edge cases.
Protocol Extensibility
Custom Markers
Organizations can define their own markers in YAML to capture domain-specific language.
Custom DV Fields
The protocol allows for adding new vector fields to measure specific aspects.
Examples: Financial (risk_ownership), Education (reasoning_autonomy), Mental Health (emotional_validation_source)
Evolve JDVP with Us
The Core: Clear architecture, explicit semantics, and reproducible algorithms.
The Flexibility: An extension model for domain-specific telemetry and policy needs.
The Invitation: Contribute markers, schemas, evaluation cases, and implementation feedback.
Help define the measurement standard for human-AI judgment flow.