Demo: An Agentic Multi-Persona Generative AI System for Mental Health Companionship

Yuekai Wang

Keywords: Agentic AI, Mental Health, LLM, Multi-Persona, Generative AI

Abstract: Loneliness among older adults is a significant public health concern associated with depression, cognitive decline, dementia, and premature mortality. Existing generative AI chatbots for mental health support often rely on a single persona, have limited memory, depend solely on pre-trained knowledge, and lack design considerations for older users. This demo presents HeartBond, an agentic AI system that supports multi-persona interaction with persistent memory and tool augmentation. HeartBond enables users to configure and interact with multiple custom AI personas that emulate the voices, personalities, and interests of people they care about. The goal of this demo is to showcase the working system, highlight how personalized generative AI can foster more natural and emotionally resonant companionship, and spark discussions on the potentials and challenges of applying generative AI to companionship and mental health.

Submission Number: 33