/moral_vector_system

Moral Vector System. Because things.

Primary LanguageHTML

Moral Vector System

Overview

The Moral Vector System is a Python-based implementation that models moral decision-making for agents in various societal scenarios. Each agent is represented with a unique moral vector space that includes weights for different duties and inclinations. This system is inspired by Kantian ethics, with duties categorized as perfect or imperfect.

The project includes:

  • Moral Agents: Represented as agents with distinct social classes, weights, inclinations, resources, and time allocations.
  • Moral World: Represents different scenarios that agents must judge.
  • Story Generation: Generates a Markdown file describing the scenarios, agents' decisions, and the effects of their judgments.
  • Visualization: PCA-based visualization of agent weights and their changes over time.

Features

  • Agent Representation: Each agent has:

    • A set of weights reflecting their moral priorities for perfect duties.
    • Inclinations that influence their approach to imperfect duties.
    • Resources (time and money) and time allocation for various activities.
    • A description that summarizes the agent's character traits.
  • Scenario Judgments: Agents judge scenarios by calculating decision scores based on their weights and inclinations.

    • Perfect Duties: Strict duties that agents try to fulfill.
    • Imperfect Duties: Flexible duties that agents weigh based on inclinations.
  • Story Generation: Produces a detailed story that includes agents, their decisions, and resource allocations for each scenario. The output is available in both JSON and Markdown formats.

  • Visualization:

    • Moral Vectors in Reduced Space: Displays agent weights and scenario vectors in a PCA-reduced 2D space.
    • Weight Changes Over Time: Shows how each agent's moral weights evolve.

Installation

Prerequisites

  • Python 3.8 or higher
  • matplotlib
  • scikit-learn
  • scipy

Installation Steps

  1. Clone this repository:

    git clone <repository-url>
  2. Install the required Python packages using pip:

    pip install -r requirements.txt

Usage

  1. Ensure the data file (moral_scenario.json) is present in the project directory. This file contains agent and scenario data.
  2. Run the Python script:
    python moral_vector_system.py
  3. The script will:
    • Generate and display the moral story (moral_story_output.md and moral_story_output.json).
    • Show visualizations of agent decisions and weight changes.

File Structure

  • moral_vector_system.py: Main Python script containing all classes and logic for the moral vector system.
  • moral_scenario.json: Data file with details on agents and world situations.
  • moral_story_output.md: Markdown output of the generated story.
  • moral_story_output.json: JSON output of the generated story.
  • requirements.txt: List of Python dependencies.

JSON File Structure

The JSON file (moral_scenario.json) consists of two main parts:

  1. Agents: Defines each agent's properties, including name, social class, weights, inclinations, resources, and time allocation.
  2. World Situations: Defines various scenarios that agents must judge.

Example agent:

{
  "name": "Alice Johnson",
  "social_class": "upper",
  "description": "A wealthy individual who is family-oriented and community-focused.",
  "weights": [0.3, 0.5, 0.2, 0.1, 0.4, ...],
  "inclinations": [0.8, 0.4, 0.9, 0.6, 0.7, ...],
  "resources": {
    "time": 120,
    "money": 1000
  },
  "time_allocation": {
    "sleep": 56,
    "recreation": 28,
    "learning": 10,
    "subsistence_work": 14,
    "community_service": 14,
    "family_care": 28
  }
}

Visualizations

  • Moral Vectors in Reduced Space: A 2D PCA representation of moral priorities for all agents and scenarios.
  • Weight Changes Over Time: A line plot showing the evolution of agent weights.

Story Generation

The story generation component outputs detailed Markdown describing each agent's decision in each scenario. It includes:

  • Agent Descriptions: Their character traits, inclinations, and resources.
  • Scenario Analysis: Decision scores, chosen actions, and the effects of judgments for each situation.

Contributing

Contributions are welcome! Feel free to submit issues or pull requests.

License

This project is licensed under the MIT License.