Ilanit Sobol & Emmy Abitbul

Domestic violence is a widespread problem impacting millions globally. In the United States, one in four women experience serious partner violence in their lifetimes. Recognizing the importance of early detection in mitigating harm and providing support, this research leverages advancements in Natural Language Processing (NLP) and Large Language Models (LLMs) like GPT-4.

This study involves creating and annotating a high-quality dataset derived from the Reddit relationship advice forum. We applied annotation strategies based on the “Michal Sela” risk questionnaires and had multiple annotators annotate each example. Our study demonstrates the promising application of NLP models in identifying signs of domestic abuse, categorizing them into different risk levels, and interpreting the predictions using predefined confounders. The results, reveal the potential effectiveness of LLMs in detecting potentially abusive relationships, marking a significant step in utilizing AI for social welfare.