More than a loading bar: Conversational silence and anthropomorphic latency cues in mental health chatbots

Abstract

Mental health chatbots increasingly rely on conversational interaction to support users' emotional well-being, yet little is known about how users interpret temporary silence while waiting for system responses. Existing research on social presence and anthropomorphic interaction has primarily focused on conversational content after a response is generated, whereas the waiting phase remains underexplored. Drawing on social presence theory, this study conceptualizes anthropomorphic latency cues as socially meaningful signals that may shape users’ emotional and interactional experiences during temporary conversational silence. Two studies were conducted. Study 1 developed and semantically validated three categories of anthropomorphic latency cues (empathic, effortful, and playful). Study 2 employed a within-subjects laboratory experiment with university students, comparing five latency-cue conditions (text, progress bar, empathic, effortful, and playful) using subjective evaluations, eye-tracking measures, and semi-structured interviews. The results showed that anthropomorphic latency cues were associated with lower anxiety and greater pleasure in several condition-level comparisons relative to non-anthropomorphic feedback. Perceived interaction continuity was significantly associated with reduced anxiety and enhanced pleasure. Within the anthropomorphic conditions, mediation analyses revealed significant indirect associations among perceived anthropomorphism, social presence, and emotional and interactional outcomes. Eye-tracking results indicated longer fixation durations on anthropomorphic latency cues during temporary silence. Qualitative findings further suggested that participants often interpreted latency cues as signals of attentiveness, responsiveness, and interaction continuity rather than merely as technical loading indicators. Overall, temporary conversational silence may function as a socially meaningful phase of AI-mediated interaction rather than a purely technical delay. The study extends social presence research into temporally silent phases of conversational AI interaction and provides preliminary design implications for chatbot-based support systems.

The authors' abstract, as published at the source. Computers in Human Behavior Reports, 2026 · DOI ↗