‘Emotion aware’ chatbot offers transformative potential for mental health care

An AI-powered chatbot developed by computer scientists at Keele University has the potential to “transform” mental health care by removing existing barriers to mental health support, the researchers have said.

Led by Dr. Baidaa Al-Bander from Keele’s School of Computer Science and Mathematics, the research involved developing a new chatbot which is more “emotionally aware” than traditional AI-powered chatbot models, offering the potential for timely and accessible mental health care for patients in need.

Their findings are published in the journal Cognitive Computation and Systems.

According to the World Health Organization, mental health disorders have a global prevalence of 25%, a statistic which is exacerbated by factors such as stigma, geographical location, and a worldwide shortage of mental health practitioners.

Mental health chatbots have been developed to address these barriers, but many of these often lack key features such as emotion recognition, personalization, multilingual support, and ethical appropriateness, meaning they are severely limited in both the care they can offer and the amount of people they can help.

To combat this, the researchers developed a new emotion-aware psychological first-aid (PFA) chatbot, which integrates an emotional distress detection system trained on Google’s BERT language model and fine-tuned on OpenAI’s GPT model.

The PFA chatbot utilizes a type of artificial intelligence known as deep learning models; artificial neurons which use data to learn and solve problems and has been fine-tuned on anonymized transcripts from real therapy sessions to improve its ability to detect emotional distress in users, and provide appropriate responses and suggestions to help them.

The PFA chatbot developed using this model is multilingual, and when tested by the researchers it was found to accurately identify emotion in just over 83% of cases, as well as providing psychological first-aid that was considered to be ethical and appropriate.

The researchers say that although the chatbot is by no means a substitute for proper treatment and care by a medical professional for those experiencing mental ill health, it could be a useful tool for overcoming existing barriers to mental health support by providing timely and accessible care in conjunction with existing health services.

Lead author Dr. Baidaa Al-Bander, from Keele University, said, “Our research focuses on creating an AI-powered chatbot designed to provide emotional support and mental health assistance.

“Using advanced language models, the system can detect emotional distress in conversations and offer empathetic, personalized responses in multiple languages. This work highlights how AI can be used to make mental health support more accessible and help people navigate challenging emotional experiences in a safe and ethical way.”