Shared Representation Companion Micro-Demo

This deterministic Python CLI micro-demo shows that identical representational input does not guarantee identical downstream meaning. Each agent receives the same interaction-derived ExperienceEntry and the same typed TVS representation layer, but interpretation diverges when memory/inference context diverges and converges when memory/inference context aligns.

Companion Paper

This repository is a deterministic companion micro-demo for the associated paper. The paper provides the broader architectural framing; this repository isolates and demonstrates one narrow claim:

identical structured representation does not guarantee identical meaning across agents

Paper: https://zenodo.org/records/18943706

Architecture Diagram

WORLD_OBJECT_A
      ↓
Experience Entry
      ↓
Shared TVS Representation
      ↓
Agent-Specific Memory / Inference Context
      ↓
Interpretation Output per Agent
      ↓
Interpretation Divergence Matrix

File Overview

• main.py: Runs Run 1 and Run 2, prints shared representation, interpretations, divergence matrix, and comparison summary.
• models.py: Dataclasses for TVS, ExperienceEntry, and Interpretation.
• agents.py: Deterministic agent interpretation logic for NPC_A, NPC_B, and NPC_C.
• divergence.py: Euclidean pairwise interpretation-distance matrix and average pairwise divergence.
• README.md: Project framing and usage.
• LICENSE: Creative Commons Attribution-NonCommercial 4.0 International license.

How to Run

python3 main.py

What This Demo Proves

• Shared representation can remain identical across agents and runs.
• Interpretation is a downstream agent-relative layer shaped by memory/inference context.
• Interpretation divergence is measurable even when representation is identical.
• A meta-observer can model second-order interpretive divergence/convergence across agents.

Run 1 vs Run 2

• Run 1 applies identical representation with divergent memory context, producing higher interpretation divergence.
• Run 2 applies identical representation with aligned memory context, producing lower interpretation divergence.

Representation vs Interpretation Boundary

TVS in this demo is the shared representation layer, not belief or meaning. ExperienceEntry is the delivered event record upstream of interpretation. Interpretation outputs are downstream and agent-relative. Therefore, divergence reported by the matrix is interpretation divergence, not perception divergence.

Scope Boundary

This repository is a simplified companion micro-demo aligned with architecture papers. It is not a full TVS implementation, full EEC implementation, governance harness, conformance suite, memory promotion engine, field dynamics simulator, or normative proof/break harness.

Contact

Stephen A. Putman
Email: putmanmodel@pm.me
GitHub: https://github.com/putmanmodel

License

Licensed under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. See LICENSE.