Design of Stress Relieving Conversational Agents with Different Genders and Stress Recognition Levels
Abstract
In modern society, the increased daily stress level has a negative impact on personal health and production performance. To regulate stress, social sharing is a preferred method. However, people are not always able to find the right person to communicate with their stress event. To provide timely stress relief assistance for people, this paper concentrates on the design of stress-relieving conversational agents, and focuses on their traits of stress recognition levels and genders. An experiment (N = 76 participants) was conducted using the “Wizard of Oz” method, employing a 2 (CA genders: male and female) × 3 (CA stress recognition levels: without stress recognition, with stress recognition, and with stressor recognition) between-groups design, by eliciting the daily stress of participants using the method of recalling the telling of stressful events. According to the ANOVA results of participants’ subjective evaluations and speech behavioral responses, CA should be able to recognize stress, even stressor, and should be designed preferentially as female voice. Mediation analysis results show that the CA stress recognition level positively affects user satisfaction through user-perceived CA perspective taking, fantasy, and empathic concern level. Meanwhile, CA gender affected user satisfaction through the user-perceived CA perspective taking level, with female CA eliciting higher user satisfaction than male CA. Users’ psychological states and physiological signals indicate that sharing and communicating with CA about stressful events has a positive impact on users’ stress relief. The findings of this paper provide a theoretical basis and design reference for CA voice interaction design for daily stress relief.
Fields of Study & Subjects
Emotion and Mood Recognition
Social Robot Interaction and HRI
Digital Mental Health Interventions
Publication details
- Journal / source
- International Journal of Human-Computer Interaction
- Pages
- 1–23
- Publication date
- 12 March 2026
- Language
- EN