Contextualizing Sexual and Reproductive Health (SRH) Information to Develop an AI-enabled Chatbot for Adolescents with Disabilities in Ghana: Challenges and Strategies

Abstract

Although artificial intelligence (AI) has proven to increase access to information on virtually any issue of human interest, not all content generated by AI may be appropriate for all consumers since socio-cultural and policy contexts vary. This paper demonstrates processes undertaken to contextualize sexual and reproductive health (SRH) data for adolescents with disabilities in Ghana for the purposes of training a chatbot to respond to their SRH information needs. Using design science research (DSR), and participatory approaches, medical and health promotion officers from the Ghana Health Service, who work on adolescent sexual and reproductive health, vetted 6591 question-and-answer pairs on adolescent sexual and reproductive health generated from OpenAI’s ChatGPT and Google’s Gemini. Employing content analysis, our findings after the vetting indicated that 1652 {25%} answers were modified, deleted, or expanded. This implies that without the review and validation, a notable amount of the SRH information would not have been relevant to the targeted end users. It is recommended that efforts to design AI solutions in SRH and other issues in Africa, which require the use of LLMs, should be deliberate about contextualizing the data generated to suit the targeted end-users.