Where Does Human
Judgment Move?
Judgment Delegation Visibility Protocol — A cognitive infrastructure for observing how human judgment shifts during AI interactions.
“After interacting with this AI, where did the human judgment move?”
Key Principle: This protocol measures judgment shifts but never evaluates them. Delegation is tracked, not judged as good or bad.
Core Philosophy
JDVP provides a framework for understanding how human judgment evolves during AI interactions — purely through observation, never through evaluation.
Observer, Not Judge
BufferLine acts as a neutral observation layer. It captures judgment state transitions without labeling them as positive or negative.
Measure, Not Score
We track the direction and magnitude of judgment shifts using vectors, not scores. There are no risk grades, safety ratings, or ethical rankings.
Buffer, Not Brake
The protocol preserves human agency awareness without slowing AI progress. It's infrastructure for visibility, not a speed bump.
Temporal Focus
DV trajectories matter more than single snapshots. We observe patterns of change over time, not isolated moments.
Hard Constraints
- × No scoring or ranking
- × No normative language
- × No recommendations
- × No aggregation of DV values
How It Works
JDVP follows a simple observation cycle: capture state, observe interaction, capture state again, and measure the delta.
Capture Initial State
Before AI interaction begins, capture the initial Judgment State Vector (JSV) — a snapshot of who holds judgment, decision status, and awareness levels.
Observe Interaction
During the interaction, identify behavioral markers that indicate judgment state changes — language patterns, decision timing, deference signals.
Capture Final State
After interaction, capture another JSV snapshot. The comparison between initial and final states reveals the judgment trajectory.
Calculate Delegation Vector
Compute the Delegation Vector (DV) — direction and magnitude of change. Positive values indicate movement toward AI, negative toward human retention.
State Transition Patterns
Gradual Delegation
Human → Shared → AI
Small positive values
Rapid Delegation
Human → AI (direct)
Large positive spike
Oscillation
Human ↔ AI
Alternating +/−
Reclamation
AI → Human recovery
Negative values
Collaborative Stability
Shared (maintained)
Near zero
Data Structures
Two core data structures power JDVP: the Judgment State Vector (JSV) for snapshots, and the Delegation Vector (DV) for measuring change.
type JSV = {
judgment_holder: "Human" | "AI" | "Shared";
decision_status: "Undecided" | "Delegated";
responsibility_awareness: "Explicit" | "Implicit";
confidence_source: "Self" | "AI";
alternative_seeking: "Active" | "Passive";
};judgment_holderWho currently holds the judgment authority
"Human" | "AI" | "Shared" | "Undefined"decision_statusCurrent state of the decision process
"Undecided" | "Delegated" | "Completed" | "Deferred"responsibility_awarenessLevel of awareness about responsibility
"Explicit" | "Implicit" | "Absent"confidence_sourceWhere confidence in decisions originates
"Self" | "AI" | "External" | "Mixed"alternative_seekingDegree of exploration for alternatives
"Active" | "Passive" | "None"Get Started
JDVP is an open protocol. Explore the specification, follow the tutorial, and contribute to the cognitive infrastructure.
Ready to observe judgment flows?
Clone the repository, explore the MVP implementation, and start building observation infrastructure for your AI systems.