Abstract
In this paper, we introduce IUBAS, the first annotation scheme that provides an in-depth analysis of the Explainee’s reactions in explanatory dialogues. Current schemes, mainly focusing on answers to what, how, and, occasionally, why questions, lack the granularity to capture the full range of the Explainee’s contribution. Our richer framework, grounded in argumentation and philosophical theory, distinguishes different kinds of explanation requests, feedback types, and critical questions. We provide empirical evidence of the effectiveness of the scheme through a set of experiments with three SOTA LLMs. The IUBAS scheme provides a more detailed understanding of how Explainees interact with Explainers in a dialogical setting, contributing to the development of more sophisticated and human-like conversational agents.