With Olalena Akintande and Joseph Keller
If the international community is to effectively coordinate effective governance of artificial intelligence (AI), it’s critical to expand representation in discussions of regulation, standards, and policy. Currently, African nations lack sufficient representation in most high-level collaboration around AI. While inclusion is improving, it must be prioritized. Failure to do so risks perpetuating historic and systemic inequalities and harms, limiting the integration of the needs and lived experiences of certain groups into emerging AI innovations, and undercutting democratic values such as equity and fairness.
Last Spring, Kigali, Rwanda played host to one of the major international AI conferences (the International Conference on Learning Representations, ICLR 2023), drawing academics, industry scientists, and civic sector actors to the African continent in an unprecedented convening. While this event was notable in its attempt to increase global inclusion in machine learning (ML) research, it may signify an important component of a broader strategy to enhance the global governance of AI.
Global AI governance needs help
Recent interest in the global governance of AI has been driven by a variety of actors. National governments, international organizations, and civil society groups have constructed frameworks concerning AI development, historically driven by well-resourced tech companies with global reach. In this context, we consider governance approaches to encompass the rules of collective decision-making for a plurality of stakeholders, where few formal control systems dictate the relationships between actors. These regulatory mechanisms coordinate as well as navigate expected conflict, offering pathways for societal and state actors to engage with issues of societal concern. The global governance of AI implements a collection of tools, including “hard and soft” leverage, strategic economic investments, alliances protecting cherished values. AI development and regulation are relevant to societies across the world, as recent events in the US and at the international level indicate.
Governance initiatives targeting AI can generate public policy proposals and efforts enabling positive societal outcomes. Advances in AI have increased access to many, potentially heightening exposure to harm and disruption. While AI may cultivate competition between countries, it also presents opportunities for partnership and collaboration. These efforts provide chances to capture benefits, reduce harms, and guide the management of risks that shape society.
In this context, African nations and researchers must assume empowered roles to amplify their voices within the global AI landscape. Discussions related to global governance and regulatory strategies are concentrated in industrialized, wealthy, and largely Western countries. The Global South lacks sufficient representation in these high-level discussions, missing vital contributions from the African continent, Latin America, and the Caribbean. This uneven distribution is particularly glaring when it comes to Africa. The Organization for Economic Co-operation and Development (OECD) tracks national AI endeavors, revealing a striking contrast between Africa and the rest of the globe.
Informing a global AI audience
Most African states are battling with a host of issues and challenges (such as insecurity, famine, flooding, poverty, etc.) that AI tools can help solve. But many still lack the essential AI infrastructure and human capital to effectively power this sector. Unfortunately, the majority of African leaders have not yet fully realized the importance of this particular opportunity.
But convenings such as the conference in Rwanda have already yielded promising societal developments in local communities. For example, a South African ML researcher shared a compelling partnership with the local government to employ AI in natural language processing tasks, creating translations, and increasing access to serve the diverse languages in the region. This project, to improve tools for and availability of data for local languages or low resource languages, illustrates tangible efforts being undertaken worthy of an international platform consisting of the largest audiences of AI practitioners and researchers. Future research opportunities could accelerate with greater visibility.
While AI may cultivate competition between countries, it also presents opportunities for partnership and collaboration, providing chances to capture benefits, reduce harms, and guide the management of risks that shape society.
In addition, AI meetings on the African continent hold space for timely discussions of regional policy and governance challenges. Workshops held in Kigali provided a forum to discuss the local impacts of climate change and opportunities for the public sector to leverage AI tools and research approaches for mitigation and adaptation. The conference coincided with fatal environmental disasters – including torrential downpours, flooding, and landslides – which devastated lives in the host country, Rwanda, “Land of 1000 Hills.” This unique forum acknowledged the importance of place, and that the negative impacts of a warming planet cause more frequent extreme weather events – people and environments in low- and middle-income countries suffer disproportionately. It lays bare Africa’s relatively small role in contributing to global emissions. Certainly, it highlighted local societal issues that relate to everyone in attendance.
African AI research confronts challenges
Even though the event held promise, attendance from underrepresented AI researchers was underwhelming. Despite the fact that ICLR 2023 was held in Africa, persons of African origin stood at only 299 out of over 6000 participants. It begs the question, how many Africans in Africa conduct ML research on African soil? One would expect a higher population given that the conference was held on the continent. This population is rising significantly, yet there is an inverse relationship between the growing number of African researchers in ML and the willingness of African governments to fund ML research. The majority of funding for research comes from external groups, a phenomenon consistent across other technical disciplines. For example, less than 20 percent of the funding for climate change research in Africa is awarded to institutions located in Africa. Low levels of state-sponsored funding create an enormous impediment to advancement, pertinent to international institutions leading governance collaboration.
Access to conferences such as ICLR remain unobtainable for many. More than 80% of these attendees (persons of African origin that attended the ICLR 2023) were non-African residents and were either sponsored by foreign organizations or universities outside of Africa. Researchers gestured to the high cost of attendance. The established and expanding lines of communication between organizers and researchers offer ways to improve upon recent progress. This is a challenge for future ML conferences in Africa, as a financial burden to attend can be prohibitive.
African ML researchers also grapple with stifling systemic friction in academic research. There are several issues which pose great constraints to capacity building of an average African scholar. Funding, infrastructural deficiency, and institutional global ranking, among others are some biggest problems for an average ML researcher working in Africa. In addition, immigration difficulties perpetuate discrimination as African researchers may be denied authorization or refused a visa to attend vital AI/ML conferences outside of the continent. Participation in these activities is often crucial to academic promotion. Many also experience stigmatized global classification, where evaluative processes may result in relatively lower ratings by foreign counterparts. Negative associations and perceptions for African researchers must continue to be improved.
Take steps forward
AI conferences in Africa, supported by global industry and academic communities, can capitalize on recent promising developments. ICLR 2023 offered a positive example and similar meetings provide opportunities to further inclusive global AI policy and frameworks. These convenings elevate timely AI and policy challenges that deserve attention. No African country is in the top 50 for AI preparedness, and twenty-one out of the 25 lowest scoring countries were African. Greater awareness and acknowledgment of pressing social issues for African ML researchers and AI solutions for public sector problems may better inform higher-level discussions, still wanting of more diverse perspectives and partners. Scholars often ask, “Who benefits?” If we are to achieve effective global governance, AI research and development must take forward steps without leaving communities behind. (