Towards Human-centered Proactive Conversational Agents

Recent research on proactive conversational agents (PCAs) mainly focuses on improving the system’s capabilities in anticipating and planning action sequences to accomplish tasks and achieve goals before users articulate their requests. This perspectives paper highlights the importance of moving towards building human-centered PCAs that emphasize human needs and expectations, and that considers ethical and social implications of these agents, rather than solely focusing on technological capabilities. The distinction between a proactive and a reactive system lies in the proactive system’s initiative-taking nature. Without thoughtful design, proactive systems risk being perceived as intrusive by human users. We address the issue by establishing a new taxonomy concerning three key dimensions of human-centered PCAs, namely Intelligence, Adaptivity, and Civility. We discuss potential research opportunities and challenges based on this new taxonomy upon the five stages of PCA system construction. This perspectives paper lays a foundation for the emerging area of conversational information retrieval research and paves the way towards advancing human-centered proactive conversational systems.

Introduction. With the advent of large language models (LLMs), the emergence and integration of conversational systems mark a significant leap forward in information retrieval (IR), which evolves many traditional interactive IR systems into conversational IR systems. For instance, Microsoft recently released a new version of Bing with its integration with ChatGPT [52] under the idea of conversational search. In the rapidly evolving field of conversational systems, proactive conversational agents (PCAs) [14, 30, 41] are emerging to revolutionize how systems interact with human users. In the literature [7, 15], the proactivity of a conversational system typically refers to the system’s ability of being aware of the long-term conversational goal and capable of taking initiatives to lead the conversation towards the goal. Recent years have witnessed a number of advanced designs that address proactivity on a range of conversational systems.

Discussion / Conclusion. This perspectives paper investigated proactive conversational agents from the human-centered perspective. We first proposed a new taxonomy concerning three key dimensions of human-centered PCAs, including Intelligence, Adaptivity, and Civility. According to this taxonomy, we re-interpreted existing literature on PCAs upon the five stages of PCA system construction (i.e., Task Formulation, Data Preparation, Model Learning, Evaluation, and System Deployment). In the light of the limitations, we envisioned future research agenda and prospects for achieving human-centered PCAs. Meanwhile, PCAs are advancing towards the realm of superintelligent AI, where maintaining a human-centered system is crucial to ensure these superintelligent AIs continue to serve human’s interests.