Skip to main content

Overview

Welcome to the AI Social Researcher subsystem! This is a powerful tool designed to automate key stages of social science simulation research using Large Language Models (LLMs).

The core mission of AI Social Researcher is to free researchers to focus on high-level scientific conceptualization, experimental design, and interpretation of results, rather than spending a great deal of time on tedious protocol writing, environment configuration, and report generation. By transforming natural language research ideas into standardized simulation protocols and final research reports, this subsystem can significantly improve research efficiency and standardization.

Core Features​

This subsystem consists of two main core components, covering the entire process from research conception to publication of results:

  1. Environment Design: Systematically transforms a vague natural language research topic into a detailed simulation protocol that follows the ODD (Overview, Design Concepts, Details) specification through multi-agent collaboration, and automatically creates the corresponding simulation environment directory structure.

  2. Report Generation: Reads the output data after the simulation ends (including the ODD protocol, scene information, metric data, and plots), and fully automatically generates a structurally complete and detailed research report in PDF format, supporting iterative optimization through multiple rounds of review.

Target Audience​

This subsystem is particularly suitable for the following users:

  • Social Science Researchers: Researchers who want to accelerate the simulation experiment process, especially in simplifying ODD protocol formulation and report writing.
  • Computational Social Science Students: Students who are learning and practicing social simulation methods and need a standardized tool to assist in completing course projects or thesis research.
  • AI Agent Developers: Developers exploring the application of LLMs in the field of research automation. This subsystem provides a complete end-to-end reference implementation.

Project Structure Overview​

To use this tool effectively, it is important to understand the organization of its key directories. All operations should be executed from the project root directory.

<project_root>/
├── config/
│   └── model_config.json      # Default location for LLM model configuration
├── src/
│   ├── envs/
│   │   └── <scene_name>/          # Root directory for a single simulation environment
│   │       ├── scene_info.json    # Core environment information (including ODD)
│   │       ├── code/              # Code required for the simulation
│   │       ├── metrics_plots/     # Plots generated by the simulation
│   │       └── research/          # All documents generated by AI Researcher
│   └── researcher/
│       ├── env_design.py        # Main entry script for environment design
│       ├── report_generation.py # Main entry script for report generation
│       ├── env_design/          # Functional module for environment design
│       └── report_generation/   # Functional module for report generation
└── ...

Next Steps​

  • To understand the overall workflow of the toolkit, please refer to Research Workflow.
  • To learn in-depth how to generate a simulation environment from an idea, please refer to Environment Design.
  • To learn how to automatically generate a research report, please refer to Report Generation.
  • To see specific command-line and code examples, go directly to Examples.

Documentation for YuLan-OneSim - A Next Generation Social Simulator with LLMs