AI Assistant for counselling of alcoholics: a case study with Kikuyu language in Kenya

Abstract

Social and economic health is greatly affected by how people feel. Anxiety disorders, depression, bipolar disorder and Post-Traumatic Stress Disorder (PTSD) are becoming more common in Kenya. Many of these issues are difficult for the country to manage, mainly because there is not enough access to mental health specialists. The treatment gap is still a major problem, even though there are not many psychiatrists. It is possible that generative AI and large language models can help more people access mental health assessments. Regardless, we still do not have wide options for languages like Gikuyu. For better rehabilitation of Kikuyu-speaking alcoholics living in outlying Kenya, this study recommends using generative AI in mental healthcare across the country. For individual care, assessments and connecting participants with mental healthcare, the study employed a transformer that had been taught on Kikuyu data. The research looked at whether AI could help alcoholics overcome the stigma by offering help that is both private and available to many. The research team built an AI assistant that responds to gaps by using language that is suitable for local and cultural needs. The accuracy of GPT-3.5 transformer when processing user inputs was assessed. Test runs of the model showed that it can accurately detect questions about mental health in Kikuyu, Swahili and English and produce suitable answers to user questions. Keywords: Alcoholics, Generative AI, low resource large language models, mental health, AI in healthcare, Natural Language Processing.