Paper Number

ICIS2026-2993

Paper Type

Completed

Abstract

Conversational agents are becoming an important source of support on mental health peer-support platforms, where engagement is critical to realizing support value and emotional support is a central support type. Unlike human supporters, AI agents’ emotional expressions are designable. We conceptualize this emotional affordance through a valence–complexity taxonomy that captures emotional tone and nuance. Using data from a leading online mental health community, we find that negative emotional expressions, which may convey validation and emotional resonance in support contexts, are more effective than positive expressions in eliciting user engagement. In particular, negative secondary emotions, which convey greater nuance and sophistication, outperform alternative emotional categories. These findings suggest that users value resonance and emotionally attuned responses from AI supporters. Our study highlights emotional expression as a key design feature of AI support agents and offers a framework for understanding how agent emotions shape user outcomes in mental health contexts.

Comments

18_IoT_publicIS_infrastructures

Download

Share

COinS
 
Dec 14th, 12:00 AM

Configured to Care: How AI Emotional Expression Shapes User Engagement in Mental Health Support

Conversational agents are becoming an important source of support on mental health peer-support platforms, where engagement is critical to realizing support value and emotional support is a central support type. Unlike human supporters, AI agents’ emotional expressions are designable. We conceptualize this emotional affordance through a valence–complexity taxonomy that captures emotional tone and nuance. Using data from a leading online mental health community, we find that negative emotional expressions, which may convey validation and emotional resonance in support contexts, are more effective than positive expressions in eliciting user engagement. In particular, negative secondary emotions, which convey greater nuance and sophistication, outperform alternative emotional categories. These findings suggest that users value resonance and emotionally attuned responses from AI supporters. Our study highlights emotional expression as a key design feature of AI support agents and offers a framework for understanding how agent emotions shape user outcomes in mental health contexts.