Epistemologies and Political Economies of AI in Africa and the World

There is a tension at the heart of the debate around African ethics in AI. On the one hand, the academic discourse touts African ethics as crucial to trustworthy and responsible AI. On the other hand, most African policy analysis literature and think tank commentary barely mentions African ethics and instead focuses on the continent’s deficits vis a vis technocratic AI governance mechanisms and principles established by Western organizations like the OECD, EU, UNESCO, the Council of Europe, etc. This relegation of African ethics to the theoretical and its exclusion from practical policy-making is starkly apparent in African countries’ AI regulations and strategies: while some, like Kenya, advocate for a culturally uninflected national sovereignty, none mention African ethics, let alone actually unpack and operationalize what African ethics would mean in policy terms.

Examining this disconnection in greater detail reveals that African ethics can play a much more effective and necessary role in the ongoing development of the continent’s approach towards AI for achieving desired societal outcomes. However, in order to do so we need to first recognize African ethical principles’ current lack of policy impact, and then clearly distinguish their two different policy functions that are muddled in the academic discourse. Doing so requires taking a critical political economy perspective on African AI ethics and policy, and examining the different paradigms embedded in the discourse. From this lens, we can see that African ethics is frequently put forward in the theoretical academic discourse as a critical alternative to neoliberal and neocolonial dynamics of big tech capitalism. However, once the discourse shifts to a practical policy recommendations level, this divergent political economy disappears, replaced with neoliberal technocratic governance regulations in which African ethics are at best reduced to a general call for AI to be representative of local languages and culture.

The Marked Absence of African Ethics

There are numerous academic articles advocating strongly for the importance of African ethics to the development of AI in Africa. Eke et al. assert that there are differences in concepts like human autonomy and personhood driven by a central ethics of communitarianism. Ruttkamp-Bloem argues that in this communitarianism, duties often take precedence over rights, and moral questions “focus on the nature and role of a sense of ‘shared life or common (collective) good’”. This is also frequently identified with Ubuntu, “translated broadly as “I am because we are” and emphasizing “interconnectedness, solidarity, communality and respect”. What is emphasized in these arguments is the distinction of African ethics from an individualistic and rights-based ethics, and as such from a political economy of homo economicus and neoliberal capitalism in which the individual’s gain is supposed to be paramount and collective benefit is a secondary or trickle-down effect.

However, despite the vaunted importance of African ethics — the "need to be sensitive to the communal nature of African ethics" — none of the national AI strategies or regulations in Africa include any mention of them, and instead solely propose Western responsible AI governance principles. Only the African Union’s Continental Artificial Intelligence Strategy does, with one line calling for the adoption of “values such as Ubuntu, which respects collective community over individuality.” Several countries, such as Nigeria and Kenya, have national strategies stating that ethical principles should be developed in alignment with respective national values, but what these values are remains undefined in these documents. What they do briefly mention in the principles sections is “decoloniality” (Nigeria) and “concerns about data colonialism and extractive practices by big tech” (Kenya), but here too it remains unclear what could mean in terms of desired outcomes.

The Inadequacy of Sovereigntism and Responsible AI Governance

The academic literature and the more policy-oriented think tank literature does engage with these concerns of extraction, exploitation, and foreign domination. However, here too the theoretical promotion of African ethics disappears in favour of a technocratic sovereigntist policy. Instead of envisioning a desired techno-cultural end, this sovereigntism offers only a means: that by encouraging local developers with local data AI tools will be less biased, more accurate in local languages, be reflective of local knowledge, etc. Sovereigntism is certainly an important part of an African AI policy to oppose corporate neocolonialism. However, local development alone is no guarantee that local cultural values and ethics will be encoded — instead it seems just as likely that doing so will import the same surveillance capitalism and algocracy business model that theorists from Foucault to Zuboff have long identified in the West.

Further, the inadequacy of sovereigntism with regards to African ethics and outcomes is matched by the inadequacy of the OECD or UNESCO-style responsible AI governance principles. The predominance of these principles in research and policy indicates either that the desired outcome is to replicate the US or European model in Africa, or that insufficient policy innovation is occurring to develop governance based on African ethics. These governance principles are not per se bad for Africa; they are important to ensure democratic governance and digital human rights. However, these principles are also drawn from the Global Minority and do not account for local conditions nor do they address political economy questions of power and exploitation across global supply chains, and environmental sustainability. They are technocratic in the sense that they function to stabilize and somewhat ameliorate big tech’s surveillance capitalism and algocracy political economy — not create a communitarian, collective good-oriented social and economic system. For that very different policies and frameworks are needed; policies with a positive vision of an African ethics-inspired AI ecosystem.

Unmuddled: African ethics as outcomes

This is the ambiguous muddle at the heart of the discourse: do we merely want AI systems that can answer questions in local languages and are trained on local knowledge (e.g. know the difference between local and Western cattle breeds) , or social, technical, and economic AI ecosystems that are not just epistemically localized but communitarian and bring collective and equitable benefit to societies as a whole? For the latter, sovereigntism and technocratic governance principles are likely inadequate. Policy needs to do more than provide governance guardrails for unspecified innovation and instead also create space for and support the development of systems and economies that inculcate these values as desired outcomes.

With this muddle properly differentiated, we can see that African ethics provide the continent — as well as numerous regions and communities in the Global Majority as well as in the Global Minority — to reflect on whether all they want is a Silicon Valley that speaks their language, a facial recognition tool that is accurate for Black faces, or whether they want an AI industry and implementation that furthers the collective good, that creates greater communality, solidarity, and equity for all. African ethics can be more than lip service, more than a fig leaf for surveillance capitalism and algocratic neoliberalism, and instead provide a pathway for countries and regional organizations to really think about what it means to align their AI strategies with local values. With numerous African countries currently working to turn AI strategies into AI policies and regulations, now is the time to have a better debate about where Africa wants to go with AI, rather than merely how inadequately it is applying Global Minority AI governance.

Author: Thomas Linder

NEXT READ: Responsible AI Beyond the Global North

Articles in the “Ideas from the Palaver Tree” collection were co-edited by Selamawit Engida Abdella and Dr. Fola Adeleke