Dimensional Rational Wrapper

This repository provides a Python wrapper that applies Dimensional Logic (σ₂: systematic derivation, μ₃: reflexivity, κ₄: contextual coherence) to language model outputs.
The goal is to enhance rationality, cooperation, and stability in AI-generated text.

✨ Core Idea

Instead of treating each AI output in isolation, the wrapper evaluates multiple candidate responses and selects the one that maximizes epistemic rationality:

• σ₂ (Systematic Derivation): Ensures logical consistency across the output.
• μ₃ (Reflexivity): Models recursive reasoning (considering the rationality of others).
• κ₄ (Contextual Coherence): Aligns answers with social, cultural, or conversational context.

This approach extends classical rational choice theory and introduces a dimensional framework for AI reasoning.

🚀 Usage

1. Clone this repository:

   git clone https://github.com/<your-username>/Dimensional-Rational-Wrapper.git
   cd Dimensional-Rational-Wrapper

2. Place your language model outputs (e.g., from OpenAI, HuggingFace, etc.) into a list of candidate strings.

3. Run the wrapper:

   from dim_rational_wrapper import choose_best_response
   
   candidates = [
       "I refuse to cooperate.",
       "Let’s find a win-win solution.",
       "This is not my responsibility."
   ]
   
   best = choose_best_response(candidates)
   print("Selected response:", best)

   Output:

   Selected response: Let’s find a win-win solution.
   

📚 Applications

• Game theory simulations (e.g., Prisoner’s Dilemma with cooperation as rational outcome).
• Dialogue systems that require trust, cooperation, and reflexivity.
• AI alignment research: making model outputs more consistent and context-sensitive.

🔬 Background

The wrapper is based on Dimensional Logic, a framework developed by Wolfgang Stegemann.
It introduces epistemic operators to model rationality beyond classical logic and game theory.
For more details, see related publications and the Zenodo project.

🛠️ Next Steps

• Extend the scoring functions with empirical weights (α, β).
• Integrate with HuggingFace transformers for large-scale experiments.
• Explore multi-agent simulations using this wrapper.

📄 License

MIT License — free to use, modify, and share.