The PAIRSx Africa Webinar with a focus on pioneering new approaches to AI through participation was held on 31 March 2025. The forum created a space for African AI practitioners and researchers to advance participatory approaches to AI as a response to the organisation of the inaugural Global AI Summit on Africa, which took place in Rwanda on 3-4 April 2025. Put together by Rwanda’s Centre for the Fourth Industrial Revolution and the World Economic Forum, the Summit sought to “pioneer new approaches and forge transformative partnerships” to harness AI for Africa’s economic growth and facilitate a safe digital future for all.
But the Rwanda Summit was invite-only and mainly targeted corporate and regulatory players in AI, a point that Kiito Shilongo, convener of the PAIRSx Africa Webinar, noted. As a result, the Summit potentially excluded other stakeholders including civil society groups, researchers, and crucially users from a process that should ideally be anchored on broad participation. As an inaugural Summit, the Rwanda event is likely to set a precedent for future AI summits in and about Africa. Hence, the Webinar discussants emphasised the need to check exclusionary practices and ensure that no one is left out of discussions on AI development and governance.
By so doing, the webinar built on the Participatory AI Research and Practice Symposium that was part of the Paris AI Action Summit and served as a contribution to wider conversations and projects around Public Voices in AI.
Discussions on Approaches to AI through Participation
The central focus of the webinar was the importance of including African publics in AI design, deployment, and governance. Its objectives were to set a clear policy demand, shape how African policies around AI and data are implemented, and make a clear statement on the vital importance of participatory AI. The discussions centred around three themes:
- Participatory AI development.
- Participatory AI governance.
- Power and resistance.
On participatory AI development, Naftali Indongo from the Namibia University of Science and Technology spoke about inclusive approaches that his team implemented to benefit disadvantaged communities in Namibia. Examples include using a co-design approach to develop indigenous technologies for the Ovahimba and San communities and building multilingual machine translation systems for low-resource indigenous languages. What is vital here is trust, and Naftali made it clear people that will only trust AI systems if they know how it works.
Most of the presentations focused on participatory AI governance. For instance, Bobina Zulfa of Pollicy highlighted the governance gaps that exist in representation, coordination, and implementation. She also critiqued the ambitious prediction that AI will add $2.9 trillion to Africa’s economy by 2030 and the focus of the AU’s AI Strategy on “development and prosperity.” What she sees as valuable is the need to ensure that Africa’s workforce is empowered to participate in processes that register their needs in the development of AI for work. Related are questions on the geopolitics of AI and what Africa’s contribution should be to global AI policy debates.
Sebastian Calderón of Innovation for Policy built on the question of AI geopolitics by pointing to the way that AI exacerbates existing inequalities along political, economic, and social lines, especially for global majority countries. This connects to the AI arms race, with implications for Africa and its ability to shape AI policy discourse. What we then have is a situation where meaningful African participation in global AI conversations remains inadequate, leading to fears that the major AI systems impose largely Western ideas on Africa. Yossra Kallali of Niyel made this point, establishing why it is important to de-mystify AI for African publics, especially for those in Francophone Africa. Achieving this, Edrine Wanyama of CIPESA observed, will require addressing AI challenges and risks in areas such as academic plagiarism, mis/disinformation, and bias. His call is that policymakers should provide avenues that include civil society actors in AI policy discussions – highlighting the importance of civil society counterpower in AI design and development.
The final theme on power and resistance featured the work of Kauna Malgwi, a human rights advocate with African Content Moderators. She described the immense psychological and emotional harms that African content moderators face. This is especially so for women, who, in addition, suffer gendered forms of harm. The question is how to resist. Resistance, according to Kauna, should spring from collective demands by content moderators for better working conditions, including that non-disclosure agreements be scrapped. Equally vital is the need to impose a duty of care on AI developers, requiring them to pass a “no-human-was-harmed” test. This will ensure that content moderators and AI labellers are placed at the heart of safety practices in the training of AI models.
Participatory AI in Africa: Challenges and Prospects
Resistance in the way that Kauna described can be effective as part of a broader move towards strengthening participatory AI approaches. But participants at the webinar warned of the possibility of participation-washing and the co-opting of languages of resistance by Big Tech and government regulators. This underscores why it is important that stakeholders facilitate participatory AI approaches that are able to withstand co-option, by, for instance, building AI movements that are strong enough to serve as a check on Big Tech.
Movements such as these can be expanded to include not just civil society actors, but the wider population. This can be seen in a proposal for a Global Citizens’ Assembly that takes inputs from local community assemblies and aims to educate people and include them in decisions on AI design and governance. Such an Assembly can potentially be developed at national and continental levels in Africa to reinforce the principle of public participation in AI. African AI enterprises and startups can also be required to embed public participation in how they train and develop AI models, given that there is little evidence to show that public participation is a key part of what commercial AI developers do.
Going forward, the goal should be to ensure that participatory AI is embedded into not just AI development and governance in Africa, but also discussions about them in events such as the Rwanda Summit – and this was the aim of the webinar. One useful approach that can be adopted to realise this goal is the participatory AI for humanitarian intervention framework developed by Nesta. It provides a protocol for public engagement with AI based on four categories: consultation, contribution, collaboration, and co-design. Adapting such a framework will serve as a useful step towards entrenching participation in discourses and practices related to AI in Africa.
PAIRSx Africa was organised by the Participatory AI Research and Practice Symposium team, supported by CMI through Dr Susan Oman’s membership to the Programme Committee.
The Public Voices in AI project was led by the Digital Good Network at the University of Sheffield and is supported by CMI through Dr Susan Oman’s involvement.