VBPM-Core: Value-Based Personality Model (Prototype)

A minimal computational model of value‑based action selection

---

📘 Overview

VBPM (Value-Based Personality Model) is a minimal, computational framework
that models personality as:

Value Dimensions × Weights → Internal Consistency → Action Selection

This repository provides:

• A clean, minimal implementation of the VBPM model
• A demonstration using a “baby personality” example
• The full research paper describing the theoretical foundation

The goal of VBPM-Core is not to simulate full human personality,
but to provide a simple and extensible starting point
for exploring value‑based behavior modeling.

---

🧠 Core Idea

VBPM assumes that:

• Individuals hold multiple value dimensions (e.g., safety, novelty, comfort)
• Each value has an associated weight
• Actions are evaluated by internal consistency:

$$ Consistency(a) = \sum_i w_i \cdot V_i(a) $$

• The selected action is the one that maximizes this value‑weighted score:

$$ a^\star = \arg\max_a Consistency(a) $$

This minimal structure allows VBPM to reproduce
value‑consistent action selection in a transparent and interpretable way.

---

👶 Demonstration: Baby Personality

The prototype includes a simple demonstration using a “baby personality”:

• High weight on safety and comfort
• Moderate weight on novelty and sociality
• Low weight on autonomy

Two scenarios are evaluated:

• Reaching for a colorful toy
• Staying close to the parent

The model computes internal consistency for each scenario
and selects the action most aligned with the given value priorities.

Run the demo:

python prototype/baby_personality.py

You can modify:

• prototype/values.json
• prototype/scenarios.json

to observe how behavior changes with different value structures.

---

📁 Repository Structure

VBPM-Core/
├── README.md
├── LICENSE
├── docs/
│   └── paper.pdf          # Full research paper
└── prototype/
    ├── baby_personality.py
    ├── scenarios.json
    └── values.json

---

🔬 Scope of This Repository

This repository intentionally focuses on the minimal theoretical core:

• No learning mechanisms
• No perception model
• No multi-agent interaction
• No LLM integration

These are discussed as future directions in the paper,
but are outside the scope of this prototype.

VBPM-Core is designed to be:

• Simple
• Transparent
• Extensible

A foundation for researchers exploring value‑based behavior modeling.

---

📄 Paper

The full VBPM-Core research paper is available here:

👉 docs/paper.pdf

---

📚 Citation

SwayMagpie (2026). VBPM-Core: Value-Based Personality Model (Prototype).
GitHub: https://github.com/SwayMagpie/VBPM-Core

---

📄 License

• Code (prototype/) — MIT License
• Paper & Documentation (docs/) — CC BY‑SA 4.0

See each directory for details.

---

❤️ Support the Research

VBPM-Core is the first step in a broader exploration of
value‑based personality modeling.

This repository provides the minimal theoretical core.
Future directions—such as perception‑driven evaluation functions,
value–perception interaction models, and multi‑agent simulations—
require significantly more computational resources.

If this framework contributes to your research,
or if you would like to support the next stage of development,
you can do so through GitHub Sponsors.

Your support helps fund the GPU resources needed
to explore more advanced versions of this model.

I make no promises for support,
but every contribution helps move the research forward.

---