Keynote Speakers

Keynote: Beyond Black Boxes: Building Transparent, Reliable, and Locally-Relevant AI for Africa

Speaker: Prof. Tegawendé F. Bissyandé

Keynote Abstract: 

In this keynote, Prof. Tegawendé F. Bissyandé explores the critical challenges and opportunities at the frontier of AI research with special relevance to the African context. Beginning with insights from his team’s recent PEARL methodology—a novel approach for detecting memorization in Large Language Models through input perturbation analysis—Prof. Bissyandé expands the discussion to broader implications for developing trustworthy AI systems.
The presentation examines how current AI evaluation paradigms often fail to consider the unique data landscapes, linguistic diversity, and application priorities across Africa. Prof. Bissyandé demonstrates how techniques like PEARL reveal fundamental limitations in how we assess AI systems and proposes a framework for more comprehensive evaluation that accounts for cultural context, data provenance, and fairness across diverse user populations.
Drawing from case studies spanning healthcare, agriculture, and education sectors, the talk highlights both the promise and pitfalls of deploying advanced AI systems in African contexts. Prof. Bissyandé introduces a vision for “Transparent AI Development”—a methodology that combines technical approaches for model inspection with participatory design principles that center African stakeholders in the AI development process.
The keynote concludes with a roadmap for building AI research capacity across Africa, emphasizing the importance of local data sovereignty, cross-disciplinary collaboration, and investment in computational infrastructure. Prof. Bissyandé presents concrete strategies for how the Pan-African AI community can lead innovations in AI transparency and reliability while addressing region-specific challenges and leveraging unique opportunities for transformative applications.