GABRIEL Software Package for Qualitative Data Measurement

This technical report introduces GABRIEL, a Python library designed to utilize large language models (LLMs) like GPT for measuring attributes in qualitative data. The report outlines the capabilities of GABRIEL in quantifying qualitative aspects of various data types, including text, images, and audio. It discusses the accuracy and directness of GPT as a measurement tool, validated against human-annotated tasks. The authors present findings that indicate GPT's performance is comparable to human evaluators across multiple domains. Additionally, the report details the application of GABRIEL in analyzing trends in Congressional remarks, social media toxicity, and educational curricula. A significant application highlighted is the study of technology adoption, revealing a notable decline in the time lag from invention to adoption over the industrial age. The report emphasizes the importance of accuracy and directness in LLM measurements, addressing potential biases and measurement errors, and provides a roadmap for future research and application of GABRIEL.