AI-CARES - LLMs for chat support in critical mental care

About the project

AI-CARES focuses on enhancing online assistance within the welfare and mental health care sector in Flanders through AI technology. By developing and implementing AI-driven tools, the project aims to improve accessibility, quality, and efficiency in digital service provision, with a strong emphasis on ethics and safety.

AI-CARES explores the application of Large Language Models (LLMs) that, after fine-tuning with domain-specific data, can better support both help chat operators and people who seek trustworthy information. The project focuses on RAG for tailored information dissemination and to support operators in chat and email-based assistance.

Specifically, AI-CARES will:

1. Develop an AI language model optimized for the sensitive context of welfare and mental health services.
2. Create three AI assistants to provide information and support via websites, chat platforms, and email assistance programs.
3. Test and validate the AI tools in collaboration with welfare and healthcare organizations.
4. Promote knowledge transfer and skill development through learning networks and integration into educational programs.

IDLab role

IDLab (Internet Technology & Data Science Lab) at UGent plays a key role in the technological development and implementation of AI within AI-CARES. The team brings expertise in:

• Natural Language Processing (NLP) and Large language models
• GPDR in ethics for very sensitive personal data
• Design of a test environment to replay chat interactions between operators and help seekers

IDLab will focus on the technical development and optimization of AI tools, as well as their integration and validation within the specific context of the welfare and healthcare sector.