Productive Value-Productive Power (PV-PP) framework
PV-PP AI Decision Architecture White Paper Series
A four-paper series for program managers, business owners, decision-system architects, and technically interested readers responsible for the decisions AI agents are increasingly being asked to make.
The series begins with the basic fact that AI agents increasingly occupy decision positions once held by people. It explains how those decisions are translated into rules, thresholds, scoring systems, and automated judgment. It then examines when a scalar model faithfully preserves the intended decision, when it becomes difficult to engineer, and when a broader decision architecture is required. The final paper turns that architecture choice into a bounded implementation method.
Available now
White Paper 1
How AI Agents Make Decisions
From Simple Rules to Scoring Systems and Automated Judgment
Explains how AI agents are assigned decisions, how simple rules differ from judgment, and how judgment is often converted into weighted scores and scalar aggregation. It also identifies what program managers and business owners are implicitly approving when they accept an agent decision model.
What You Don’t Know About Scalar Aggregation Can Hurt You
Examines when a one-number model faithfully represents the decision the organization intended to delegate and when it preserves only the appearance of correctness. It distinguishes reproducing a winner from preserving the full ranking and the actual decision process.
Development draft under review.
Forthcoming
White Paper 3
Before You Build the Utility Function
A Practical Test for Easy, Hard, and Impossible Scalarization
Provides an architecture-selection test for deciding whether to use a simple score, engineer a more complex scalar model, enrich the representation, or move to a broader structured decision architecture.
Development draft under review.
Forthcoming
White Paper 4
How to Design a PV-PP Decision System
From Business Rules to Structured Governance, Sidecar Control, and Runtime Execution
Describes a bounded path from business requirements to structured decision artifacts, sidecar governance, testing, replay, and the longer-term goal of inline decision control with persistent state and feedback.