Integrating Stakeholder Engagement and Participatory Design in the Development of a Training Dataset for a Culturally Adapted Multilingual HIV Awareness Chatbot in Kenya

Abstract Globally, adolescents (15-19) and young people (19-24) are impacted by HIV and largely account for new infections. This age group suffers significant gaps in their access to HIV knowledge, including verified and accurate information on transmission, prevention, and treatment. The study answers the question "how can participatory design approaches ensure that an AI-enabled HIV awareness chatbot is inclusive and locally relevant in adolescent HIV programs?" A modified spectrum of public participation and participatory design was incorporated during the co-design and co-development of the HIV training dataset that was used to train an AI chatbot. Thirty-two multisectoral stakeholders and 91 end-users, including adolescents, health workers, counsellors, policy makers, university students and teachers, were engaged throughout the development process. Insights from the co-design process were used to iteratively shape the training dataset and consequently the chatbot's content, language, tone and functionality, ensuring relevance and cultural sensitivity. Stakeholders and users helped shape the training dataset by suggesting the inclusion of facilities that offer HIV care and counselling services for adolescents and young people. Stakeholders advocated the localisation of the training dataset, resulting in translation into Swahili with inclusion of slang phrases that youth identify with. They also suggested improvements to the developed chatbot through the inclusion of multimedia content and offline access. This research offers insight into inclusive, culturally adapted and locally driven development of a responsible AI solution in health. Inclusion and awareness of gender and inclusivity precipitated the development of a chatbot that was contextualised for the local population.