Automating Classification of Moral Reasoning in Natural Language

Molly Crockett (Psychology), Spring 2023

We are building a machine learning classifier to identify and distinguish modes of moral reasoning in historical natural language data. We seek to explore the emergence of different forms of moral reasoning over time and relate them to other cognitive signatures in historical texts. As part of our project, we are building a large, multi-source, historical dataset of moral arguments spanning centuries. We will have expert coders label a sample of the data and use this information to fine tune a large language model. The automated detection of moral reasoning will allow social scientists to answer questions about the prevalence, variation, and impact of different modes of moral reasoning across time periods, cultures, and communication mediums.