Polymorphic Decision Protocol (PDP)
A normative candidate for high-uncertainty decisions with AI panels and human decision authority.
A staged protocol specification with limited mathematical components; end-to-end operation and empirical effectiveness remain unvalidated.
Theoretical Substrate
PDP is a normative protocol candidate for organizing high-uncertainty decisions involving AI panels.
Rather than computing simple majority votes or uncalibrated consensus, PDP discounts agent agreement by input provenance, subjects confidence to controlled adversarial pressure, and keeps authority with the human.
The repository includes a prospective research plan that has not been executed. The author-reported development case is not independently verified.
Primary Research Questions
- How can consensus among large language models be discounted when models share underlying pre-training provenance?
- What testing protocols empirically invalidate overconfident AI predictions under stress?
- How should human operators own decision weights, stakes, and emergency exit boundaries?
Primary Claims & Invariants
Normative specification for staged panel deliberation, pressure testing, human custody, and outcome audit.
Unexecuted Test 002 research plan with prospective capture and falsification requirements.
Limited mathematical routines and UI behavior with local tests; no complete operational system.
Software, Substrates & Datasets
PDP Protocol Specification v1.3.0
Normative candidate specification and limited, tested software components. Effectiveness is unvalidated.