Abstract
Analyzing texts such as open-ended responses, headlines, or social media posts is a time- and labor-intensive process highly susceptible to bias. However, large language models (LLMs) are promising tools for text analysis, using either a predefined (top-down) or a data-driven (bottom-up) taxonomy, without sacrificing quality. Here, we present a step-by-step tutorial to efficiently develop, test, and apply taxonomies for analyzing unstructured data through an iterative and collaborative process between researchers and an LLM. Using personal goals provided by participants as an example, we demonstrate how we used this method to write prompts to review datasets and generate a taxonomy of life domains, evaluate and refine the taxonomy through prompt and direct modifications, and apply the taxonomy to categorize an entire dataset with high intercoder reliability, while achieving high levels of human-LLM intercoder agreement, reducing analysis time by approximately 87.5%. This test offers a proof of concept, suggesting that with the right procedures LLMs can be used to generate reliable bottom-up categorizations. We discuss the possibilities and limitations of using LLMs for text analysis.