Hear You in Silence: Designing for Active Listening in Human Interaction with Conversational Agents Using Context-Aware Pacing

TL;DR

This work identifies pacing, including silence, as a critical but overlooked component of empathic human-AI interaction. By qualitatively analyzing ten active-listening cases, the authors establish a five-strategy framework for context-aware pacing and implement an LLM-based CA that adapts response timing accordingly. In a between-subject study across career and relationship support tasks, the context-aware CA outperformed a static-pacing baseline on perceived listening quality, affective trust, human-likeness, smoothness, and interactivity, with deeper user self-disclosure and engagement. The findings demonstrate that strategically timed pauses can enhance social presence and trust in AI, while highlighting design trade-offs with expectations of machine efficiency and the need for personalization and careful cross-domain validation. Overall, the work provides design principles and implementation guidance for integrating pacing as a core element of empathic, user-centered conversation agents.

Abstract

In human conversation, empathic dialogue requires nuanced temporal cues indicating whether the conversational partner is paying attention. This type of "active listening" is overlooked in the design of Conversational Agents (CAs), which use the same pacing for one conversation. To model the temporal cues in human conversation, we need CAs that dynamically adjust response pacing according to user input. We qualitatively analyzed ten cases of active listening to distill five context-aware pacing strategies: Reflective Silence, Facilitative Silence, Empathic Silence, Holding Space, and Immediate Response. In a between-subjects study (N=50) with two conversational scenarios (relationship and career-support), the context-aware agent scored higher than static-pacing control on perceived human-likeness, smoothness, and interactivity, supporting deeper self-disclosure and higher engagement. In the career support scenario, the CA yielded higher perceived listening quality and affective trust. This work shows how insights from human conversation like context-aware pacing can empower the design of more empathic human-AI communication.

Figures (9)

• Figure 1: Data analysis process and an analysis example for a video where the speaker faced dating anxiety.
• Figure 2: Pipeline and visual elements of the context-aware pacing CA. It consists of three core backend modules (Context Analysis, Response Generation, and Conversational Memory Module). After receiving user input, the Context Analysis Module would first select the most appropriate pacing strategy when the status bar of the CA shows "Thinking...". Then its output and user input would be fed into the Response Generation Module to generate responses and apply corresponding pacing strategies. Both modules are supported by the Conversational Memory Module to manage the conversational contexts.
• Figure 3: Overview of the user study procedure. The user first landed on the welcome page, and then interacted with the conversational agent under two scenarios, career and relationship support, in a counterbalanced sequence. After completing every scenario, the user completed a survey evaluating the corresponding experience. Finally, a semi-structured interview was conducted.
• Figure 4: Statistical analysis on questionnaire results in career-support scenario, with $G_n$ referring to the control group, and the context-aware pacing group $G_p$. These figures show the significant difference between the control and experimental group in terms of perceived listening quality, affective trust, human-likeness, smoothness, and interactivity.
• Figure 5: Statistical analysis on questionnaire results in relation-support scenario, with $G_n$ referring to the control group, and the context-aware pacing group $G_p$. These figures show the significant difference between the control and experimental group in terms of perceived human-likeness, smoothness, and interactivity.