In a continent often framed in a state of perpetual crisis, development seems like an unattainable endeavor. Indeed, recent military coups and armed conflicts such as the ones in Sudan and Gabon do not give spectators of African affairs, particularly those in the West, any assurances that Africa is on the rise, a motto that was once tooted around by global opinion movers and shakers such as The New York Times, The Economist, and others.

Development seems to be in crisis and in need of quick resuscitation. To put things in perspective, Official Development Assistance (ODA) totaled USD 185.9 billion in 2021, according to numbers from the Organisation for Economic Co-operation and Development (OCED). On the other hand, the dismal development outcomes show the ineffectiveness of international development. For example, poverty rates for most countries receiving international assistance have increased since 2019, leaving between 50% to 70% of their populations living under the poverty line. Globally, the situation is not encouraging. The number of people living in extreme poverty increased in 2022 to more than 700 million people around the world, according to The World Bank. The latest SDG 2023 progress report from the UN paints a grim picture. The progress has been weak and insufficient on more than 50% of the targets. Even worse, progress has either stalled or gone into reverse on more than 30% of the SDG targets. This includes key targets on poverty, hunger, and the climate. Furthermore, on an extremely alarming note, the report concludes that more than half of the world is being left behind, and the majority of those who are left behind, as you may have guessed, live in the Global South.

As the global development agenda struggles, artificial intelligence (AI) is being framed as an effective tool for expediting development goals and targets and fixing the broken model of international development. New AI for development (AI4D) programs have been deployed by international development agencies and local partners across several African countries in Sub-Saharan and West Africa among other locales. Ostensibly, this seems like a logical and worthwhile undertaking given the global hype around AI. However, AI initiatives in Africa are underpinned by the deficit model of development. This deficit logic emphasizes the failure of development in the Majority World as an imminent consequence of the lack of human and technological capacity.

Machine learning is often understood as a technology that will shift the agency from the human to the more reliable non-human systems and machines. This assumption, which drives AI development in the West as well, manifests in different ways in AI development initiatives in Africa.

AI initiatives in Africa are underpinned by the deficit model of development.

The efficient allocation of resources and productivity outcomes is prioritized over the real needs of the people and their visions for what it means to build their own communities and have them prosper. For example, several AI4D projects are focused on solving issues related to agriculture by trying to increase crop production and predict crop diseases using machine learning such as FarmSpeak Technology in Nigeria. Similarly, other projects related to environmental sustainability are creating machine learning models to predict environmental stresses and their impact on health and global warming. The Modelling Early Risk Indicators to Anticipate Malnutrition (MERIAM) project by Action Against Hunger partnered with other development agencies and academic institutions to tackle this challenge by monitoring and predicting drought conditions. Another application related to energy management by the Responsible AI Lab (RAIL) in Ghana, is trying to embed effective energy distribution models in the grid to optimize the availability of electricity. Perhaps one of the most promising applications of AI in the region is in the area of natural language processing (NLP), where there is an attempt to build language models for indigenous African languages such as Igbo, Hausa, Yoruba, Twi, Akan, and others by emerging start-ups using development funding programs such as the Lacuna Fund. These models can be embedded in other applications in areas such as healthcare and education. The benefits of these programs and applications might be apparent given the local conditions in most African countries.

However, the reality on the ground is that most AI development in Africa is configured by practices of international development agencies and corporate social responsibility programs of multinational corporations. These programs, enacted in partnerships with Big Tech, and local actors, including scientists and practitioners, are overly focused on generating local African datasets and building technical solutions, with the hope that one day they will become dazzling stars in the global tech scene. Significant energy and investment are going into collecting local datasets to revamp machine learning models for predictive analysis based on the local context. However, there is very little discussion about the aims and uses of these AIinitiatives, and which communities and social groups benefit from them? And how will these technological solutions be adopted in the local context? In short, there is lack of serious engagement with the political imaginations of the various local communities about their visions for what a technological future with AI looks like for them.

The Development Model of Responsible AI

Since the Paris Declaration on international development and aid effectiveness by members of the OECD and the Accra Agenda for Action in 2008, a lot of efforts and energy have been spent on improving development models towards greater ownership, harmonization, and alignment with local goals of development partners. On the surface, this looks like a significant departure from the conventional development economics wisdom that places an emphasis on domestic industrialization or international trade. However, critics of development ownership point out that while this model of development appears to be plausible at the outset, in practice, there is increased pragmatism and self-interest of development actors which render its implementation ineffective. Additionally, Takiyah Harper-Shipman, African studies scholar at Davidson College, argues that development ownership serves to perpetuate the central role of multilateral organizations and international development donors in African development while affording very little autonomy for local partners. Despite this criticism, various development programs targeting the Majority World, such as the UN Agenda 2030 and many of its regional variations seem to be adopting this development approach for achieving the SDGs.

Leading international development agencies in Africa such as Canada’s International Development and Research Center (IRDC), the Swedish International Development Cooperation Agency (SIDA), and Germany’s Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) play a key role in the dissemination of AI technology in the continent, at the moment. Their focus has been on local capacity building and ensuring that the ethical dimensions of AI development are not overlooked in a continent with deep-rooted colonial legacy and a long history of exploitation.

There is lack of serious engagement with the political imaginations of the various local communities about their visions for what a technological future with AI looks like for them.

However, the AI ethics agenda pushed by proponents of responsible AI development in Africa serves two aims. It frames the deployment of AI as a way to avoid both old development pitfalls as well as forms of algorithmic harm that is experienced in the West. Of course, there is nothing wrong with trying to engage in technological development and innovation in a way that does not exacerbate economic exploitation and reproduce prejudices, biases, inequality, and oppression in any society. However, the focus of the AI ethics agenda in Africa has been influenced by Western frameworks which take a universalist approach to conceptions of ethics, privacy, human rights, and so forth. Current efforts of AI governance in Africa attempt to adopt these approaches and build on the same normative claims about the benefits and risks of AI without adequately engaging with how AI technology is understood by Africans.

Additionally, most African governments seek responsible AI programs to attract further development funding that can help them in dealing with some of their fiscal issues and making good on unfulfilled promises concerning national science, technology, and innovation programs. However, there isn’t enough due diligence in the provisioning of these programs by national governments or ensuring their alignment with a broader national economic development strategy. Most of these programs appear to be relying only on technoscientific and commercial rationales that are aimed at attracting Big Tech and multinational technology consulting firms with a narrow focus on economic competitiveness in a globalized market system. By doing so, national governments allow more agency to international actors in shaping the AI development agenda in Africa.

In a world where development models are deeply entrenched in a neocolonialist capitalist world system, it seems more urgent than ever to try to decolonize AI if Africans were to reap any benefits out of the technology.

Today, there are numerous challenges to an AI development agenda in Africa. Many AI researchers are pointing out the difficulty in accessing the amounts of data required for AI models, limited data quality and storage capacity, and lack of regulatory frameworks and policies. Broadband connectivity is still a big issue despite the proliferation of mobile devices and connectivity in the continent. However, the focus on data collection without addressing the broader aspects and more urgent issues related to data infrastructure beyond connectivity perpetuates the dependency of African countries on Western and Chinese companies and limits their ability to influence the development agenda. For example, African universities do not currently have specialized AI programs that are institutionalized in meaningful ways that can support national development. They lack adequate funding to develop such programs to produce local talent and relevant local innovations. Most African governments do not have a national vision or strategy to capitalize on the opportunities that AI technology may present to their national development. Instead, they are completely reliant on international development funding. This is not a good way to build national technoscience and innovation strategies that can benefit the majority of their citizens. The absence of an effective role for the state leaves Africa, once again, vulnerable to the practices of data extractivism and digital exploitation.

Decolonizing AI

Decolonization has been central to many international AI development programs ascending on the continent in recent years. In a world where development models are deeply entrenched in a neocolonialist capitalist world system, it seems more urgent than ever to try to decolonize AI if Africans were to reap any benefits out of the technology. However, the focus on decolonizing AI seems to be centered around technology without adequately attending to the social, political, and economic conditions that shape and are shaped by technoscientific innovation.

The focus on data collection without addressing the broader aspects and more urgent issues related to data infrastructure beyond connectivity perpetuates the dependency of African countries on Western and Chinese companies and limits their ability to influence the development agenda.

Google’s DeepMind scholar and AI developer Shakir Mohamed and his colleagues William Isaac and Marie-Therese at the University of Oxford have, for instance, proposed a decolonial approach that focuses on sociotechnical foresight based on values and power to bring AI development closer to the community. Their approach, like other salient decolonial thinking in the field, seems to focus on how to apply decoloniality to AI and create a version of AI that can do more good than harm. Conversely, South African scholar, Rachel Adams, has asked the question, “whether AI can, in fact, be decolonized?”! Adams has pointed out that the decolonization agenda is preoccupied with building a superficial version of AI that seeks to be connected to the community, without fully addressing the underlying assumptions that underpin AI in terms of Western ethics, moral philosophy, and a Eurocentric understanding of human intelligence. She urges AI decolonial thinkers to start by asking the question of what does AI mean because of colonialism? In another way, there is an urgent need to reimagine AI from an African perspective. The decolonization discourse, in some sense, has been coopted by Western actors to characterize this version of international development as different than previous development models.

If African proponents of AI development and their international allies want to decolonize AI per se, they must first recognize that decolonization is not the same across the world. Also, it is not limited to collecting local datasets that feed machine learning models or building indigenous technological solutions based on Western models of innovation. Particularly in Africa, decolonization was understood as a political project that sought to both dismantle colonial institutions and structures and challenge the hegemony of European knowledge production systems. It was a project rooted in the political and economic autonomy and technoscientific sovereignty of postcolonial African countries. One of the ways that the protagonist of African independence pursued this vision is through a Pan-African agenda for economic development cooperation and collective political action. This resulted in the culmination of the Organization of African Unity (OAU) in 1965, the predecessor to the African Union (AU).

Admittedly, both the OAU and the AU have been marred by failures, ineffectiveness, and negative perception. Moreover, the region has been facing a long-standing crisis of democracy, with its institutions compromised and its struggles with political instability. However, there remain opportunities, and more importantly, an urgent need to consider a Pan-African agenda on AI where consensus-building might be possible. This will require national governments to prioritize technoscientific development and innovation in their national agendas in a way that ensures its detachment from Western terrains of development and progress.

Another key aspect is the need for national governments to assert their autonomy and sovereignty over large technological projects. With the increased influence of China on economic development in Africa through organizations such as the Forum on China-Africa Cooperation (FOCAC), and the struggle for global AI dominance, Africa will become, yet again, the battleground for powerful international actors and the continuous exploitation of both the data and people. Thus, a renewed national agenda on resisting datafication and extractivism and prioritizing national interest on large public data infrastructures becomes crucial for Africa to benefit from any technological innovations in AI. This will require strengthening regional arrangements and building regional organizations. These initiatives should not only be capable of engaging in technoscientific innovation in AI, they also must have the capacity to think through the process of transforming Africa’s rural economy and communities into more sustainable economic models that truly harness data and AI for the benefit of the African peoples.