AI for Analyzing Divergent Perspectives in a Conversation
Yair Neuman · 2025
When conversing, people express their perspectives. Sometimes, we may want to understand how these perspectives diverge or converge. This chapter presents a methodology for identifying and measuring divergent perspectives in conversations. First, I will show how to use large language models (LLMs) to enrich an utterance with information. Prompting the LLM, we may extract implicit dimensions from an utterance, such as the ambiance implied from the text, and use it to improve understanding. Next, I present a new measure to quantify the divergence between interlocutors’ perspectives. I illustrate this measure through the way Homer and Marge Simpson see each other. Finally, I apply the methodology using a detailed analysis of a diplomatic meeting between the United States and Chinese officials.