A multimodal AI-based decision support framework for precision agriculture: Enhancing accessibility for low literate farmers

Background : Due to limited literacy, smallholder farmers in agriculture in developing regions continue to encounter challenges, hindering their ability to leverage AI-driven and digital technologies. Most of the conventional platforms are dependent on text-based interfaces, excluding low-literate users from accessing essential decision support services. Methods : A multimodal AI framework is proposed in this study that aims at enhancing the usability and accessibility for smallholder farmers in rural regions. Different accessibility features like localized voice prompts, AI-powered advisory modules incorporating computer vision, and culturally adapted iconography and recommendation engines are integrated for crop disease identification, market price forecasting, and fertilizer scheduling. A mixed-method methodology was conducted using ISO 9241-11 usability standards and participatory design methodologies involving 100 farmers (50 illiterate and 50 semi-Illiterate). Results : The results showed obvious improvement in user experience and inclusivity that integrated multimodal interfaces. Cross-task comparative results indicate improved usability, with task success rates of 62% for illiterate and 82% for semi-literate users. The AI modules achieved strong disease detection performance, with an average accuracy of 87.7%. Conclusions : The results showed the effectiveness of the described framework to fill literacy gaps of farmers in the AI agricultural domain.