This project explores context-engineering to bootstrap conversational agents for mental wellbeing, specifically in higher education.
| Academic Advisor | Prof. Simo Hosio |
| Topics | Mental wellbeing, multi-agent systems, conversational agents |
| Degree | (Industry) Ph.D. |
Abstract
Mental health stigma remains a significant barrier to care, preventing many from seeking support. The EMPOWER project harnesses multi-agent conversational AI to reduce stigma by fostering empathy and openness through authentic, crowdsourced narratives. Using open-ended crowdsourcing, we collect diverse human experiences to create relatable AI agents that facilitate non-judgmental dialogues. These agents are validated in real-world field studies, focusing on higher education students, a group particularly vulnerable to stigma. Collaborating with AI, psychology, and clinical experts, EMPOWER ensures ethical data practices and impactful solutions to destigmatise mental health globally.
Key Objectives
- Optimise conversational crowdsourcing to gather authentic, diverse narratives for stigma-reducing AI agents.
- Design multi-agent AI systems to foster empathy and reduce stigma through interactive, peer-like dialogues.
- Validate stigma reduction in a large-scale field study (N=1000) among higher education students.
- Share findings via high-impact publications and open-source tools to drive anti-stigma efforts.
Research Questions
- What factors improve the effectiveness of crowdsourcing authentic narratives for stigma-reducing AI agents?
- How do cultural factors influence the deployment of AI systems for stigma reduction?
- Which human factors (e.g., trust, perceived empathy) enhance user engagement with and the effect of multi-agent AI for stigma reduction?