Participatory AI for Agriculture: Co-Designing Hydroponic Pest Detection Models with African Farmers

Africa faces pressing agricultural challenges exacerbated by climate change, declining productivity, and limited infrastructure. While Artificial Intelligence and Internet of Things (AI and IoT) enabled hydroponic systems offer a promising solution for resource efficient, high yield food production, many existing technologies remain misaligned with the lived realities of African farming contexts. This research adopts a mixed methofs, community-based approach to co-design an AI and IoT hydroponic system tailored for resource-constrained regions.Through participatory engagement, farmers contributed directly to system development, particularly in refining a pest detection model. Drawing on farmers’ insights, the study pivots to a Convolutional Neural Network (CNN) model focused on detecting visible pest frass specifically as a more reliable indicator in real-world settings.This work forms part of the broader SmartHydro project, which aims to deliver an affordable, offline-capable hydroponic solution for African farmers. The development of the frass-based CNN pest detection model is a core contribution of this study.By using Co-design, local knowledge, and context-aware Artificial Intelligence, this study highlights the importance of user-led innovation in creating practical, resilient technologies. The resulting system promotes digital inclusion, enhances user  agency, and contributes to sustainable, scalable agricultural solutions rooted in African realities. Key words:  Artificial Intelligence, Internet of Things, Co-design, Convolutional Neural Network, Hydroponic farming.