Towards Human-centered Proactive Conversational Agents

TL;DR

This paper argues that proactive conversational agents should be designed around human needs and ethics rather than solely advancing technical capability. It introduces a three-dimensional taxonomy—Intelligence, Adaptivity, and Civility—and reframes the PCA development lifecycle into five stages: Task Formulation, Data Preparation, Model Learning, Evaluation, and System Deployment. By reinterpreting existing literature across these dimensions, the authors outline concrete tasks, data practices, alignment methods, and evaluation frameworks that emphasize trust, boundaries, and user well-being. The work highlights multidimensional evaluation and human-centered deployment as essential for responsible PCA progress, especially as AI advances toward higher levels of intelligence. It also sketches a forward-looking research agenda focused on real human needs, collaborative data collection, and robust, customizable evaluation to sustain human-centricity in the era of increasingly capable conversational agents.

Abstract

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.

Figures (6)

• Figure 1: Three key dimensions of human-centered proactive conversational agents with representative abilities.
• Figure 2: Different types of proactive conversational agents in terms of three key dimensions of human-centered PCAs.
• Figure 3: Three types of human alignment approaches.
• Figure 4: Different user interface designs for user preference elicitation in conversational recommender systems.
• Figure 5: A simplified diagram outlining the different ways people rely on and trust proactive conversational agents.