Comparing the persuasiveness of role-playing large language models and human experts on polarized U.S. political issues

This repository contains the data and code to support the paper Comparing the persuasiveness of role-playing large language models and human experts on polarized U.S. political issues, authored by Kobi Hackenburg, Lujain Ibrahim, Ben Tappin, and Manos Tsakiris

Repository Content

Data

• final_dataset.csv in the data folder is the final processed dataset that is used in the analyses.
• The raw_data folder in the data folder contains the raw experiment data.
• The messages folder contains the messages generated by GPT-4 and political communication experts.

Code & Notebooks

The code folder contains all the Jupyter Notebooks that were used to generate the analyses and figures in the paper.

Running Analysis

System Requirements

• To install Python 3, follow these instructions.

• To install Pip, follow these instructions.

• To install Jupyter Lab/Notebook, follow these instructions. To run Jupyter Lab/Notebook, follow these instructions.

• To set up a virtual environment and use it in Jupyter Lab/Notebook, follow these instructions.

• To install requirements:

1. Clone this github repository
2. Download Python packages needed

pip install -r requirements.txt

Contact

Please contact Lujain Ibrahim or Kobi Hackenburg for any questions regarding this repository or the paper.