Franco Beukes

In the absence of digital texts from minority groups, such as San from Southern Africa, public generative AI systems create responses that lack relevance and contextual understanding of local people. This study explores the relatability of AI-generated motivational stories from a Retrieval-Augmented Generation system accessing a curated knowledge base of previously collected real-life San youth experiences and San elder's wisdom proverbs. A focus group session was conducted with 46 members of the San community to collaboratively frame how AI could be used to generate culturally valid motivational stories to promote educational resilience. An AI story generation was simulated with San youth to determine relatability. Based on this, a fully functional prototype was developed, with a smart mirror interface, that detects basic emotions to trigger the construction of motivational stories for San youth. The system was evaluated through feedback sessions with youth and elders, focusing on emotional resonance, cultural appropriateness, and contextual understanding. Results showed that participants responded positively to stories grounded in their lived experiences and traditional wisdom, confirming the importance of contextual knowledge bases. The system was seen as a source of emotional encouragement; however, challenges such as limited literacy among users also highlighted the need for further development to make the tool fully inclusive and culturally grounded.

Climate change threatens African subsistence farming with erratic rainfall, droughts, and seasonal shifts causing devastating crop failures. Hydroponic farming, often recommended to mitigate these threats, requires constant monitoring and knowledge, which can be overwhelming for first-time growers. Research has explored using the Internet of Things (IoT) and Artificial Intelligence (AI) to automate and monitor hydroponic systems. Still, few studies have tested these systems in the field with farmers. Our previous work involved an eight-month deployment of an AI and IoT-enabled hydroponic system with rural subsistence farmers in South Africa. We conducted focus groups and observed the hydroponic systems using a qualitative approach. This paper discusses lessons learned from deploying the system in Namibia. We found that generalized AI models are unlikely to succeed due to differences in water quality, climate, participant readiness, and internet connectivity. Therefore, AI and IoT systems for rural Africa must be tailor-made to be functional and effective for crop growth.