Research on Theory-Oriented Interaction between Humans and Conversational Agents: A Concept for Systematization

Abstract

The current mainstream in conversational agent research centers on an engineering-driven approach, aiming to achieve advanced dialogue technology through the design and implementation of sophisticated models. However, in recent years, often referred to as the era of large language models (LLMs), it has become possible to implement a certain level of conversational functionality even without specialized expertise. In light of this context, research focused on building and applying theories to understand phenomena in human-conversational agent interactions, and to explain effective dialogue strategies and utterance designs for these agents, will gain increasing importance. Although such studies have been conducted on an individual basis, a systematized methodology has not yet been established, and research approaches are not well-defined. This presentation introduces a new concept: ”Theory-Oriented Interaction Research between Humans and Conversational Agents.” We will also present a concept for its systematization. Specifically, drawing upon the author’s previous research, this presentation will organize and discuss this field of research, highlighting its significance and distinguishing it from other research approaches.