Background
Between the 18th and the 19th of July, the African Union (AU) endorsed the African Union Strategy on Artificial Intelligence (AI) in Accra, Ghana, during the 45th Ordinary Session. This endorsement has been met with considerable applause from several actors in policymaking, governance, data science and development spaces within and outside of Africa. This acclaim pointed towards the timely release of this strategy that allows Africa to join the global AI conversation, particularly with respect to being on top of planning for its people to both benefit in a development perspective but also be protected from the harms these systems pose as a whole.
As captured in the Strategy’s title, “Harnessing AI for Africa’s development and prosperity”, the position that Africa takes in the integration of AI systems within the continent is seen in its elaboration throughout the strategy- a bold, hopeful and almost idealistic vision for a more developed Africa whose growth has been largely spurred by the potent of automation and smart systems.
However, beyond this utopian thinking or similarly anti-utopian sentiments that different players have been advancing, Pollicy intends to bring to the conversation a more total inquiry that is not rooted in redactive binaries and that centres the common African. Our hope is for this line of inquiry to allow for more incisive decision making by all players in the AI space in Africa based on this Strategy.
Understanding the Foundations of the AU-AI Strategy
- Africa’s Continental AI Vision
AI, like any other technologies, has been integrated within societies world over basing on distinct although not always unrelated reasons. This reasoning process may be traced to the ancient idea of technologies inherently being ways of revealing or ways of “bringing forth” whether as a means to an end or simply as human activity. For example, in different societies, AI has been envisioned to bring forth the revolutionising of industry, firmer state control through extensive integration of AI systems across social systems, facilitation of the manufacturing of AI systems, to be leaders in the governance of AI systems and to solve particular societal problems among other reasons.
According to the Continental AI Strategy, Africa appears to look at AI systems as a way of bringing forth the development of its economies and the subsequent prosperity of its people as envisioned in the continent’s commitments to the AU’s Agenda 2063. Growth, development and opportunities are the themes at the center of this vision. The question then begs, do African people outside the policymaking space share this vision at present? Or is that something we shouldn’t concern ourselves with? And if not development, what is the ordinary African primarily concerned with or interested in when it comes to AI? Now, of course there is no category as “the ordinary African” as divisions exist in all societies across socio-economic class, urban-rural living, religious affiliation, education levels et cetera. An intersectional lens is critical therefore in ascertaining through mass public participation what different communities’ sentiments are in terms of their hopes, dreams and visions with these technologies. This exercise in thought would lead to envisioning which answers the question, what do African people hope AI brings forth for them?
Further important to pay attention to under the vision are the meanings that arise around terms used in our envisioning process. This is so that the narratives and images they evoke match our practicable realities. For instance, what does the vision for AI for development in Africa as hinged on the attainment of Agenda 2063 and the 2030 Sustainable Development Goals (SDGs) practically give rise to on the continent? Is it actual economic progress for all people, or just a few players? Does it factor in socio-political dimensions of prosperity including living free of oppression and domination across the board for all Africans? This kind of questioning is meant to prompt a full understanding of what development means for our context especially as a way of avoiding the AI inevitability conscious or unconscious tropes rooted in techno-solutionism as mainly seen in the idea of technology as a transcendence of the political, social and cultural yet it can serve the purpose of maintaining the status quo. With evidence already showing how AI on the continent is worsening existing structural inequalities such as with unequal distribution of resources and power, it becomes imperative to capture in policymaking the full breadth of development including aspects such as liberation of people which is scarcely explored in the strategy.
Lastly, another useful approach in vision refining would be defining what Africa sees as its chief contribution to the global AI landscape. Currently, as we are majorly consumers of AI technologies from elsewhere, what does that reality impose on Africans? Having a clearly spelled out vision of how we envision contributing to the global AI landscape rooted in supporting the betterment of our current political and socio-economic contexts would therefore allow for a more emancipatory AI integration approach which is not exploitative such as in terms of labour and also allows people decisional autonomy and agency et cetera. Not defining this allows external forces define that for us — ie whether Africans will predominantly play the role of invisible labour to AI systems being developed all over the world or be talent to be nurtured and utilised or be unprotected data sources or have our nations be used as carbon sinks or extractive mineral hotspots for powering AI technologies and so forth. This unique positioning of Africa would be central as a guiding ideology for regulation of AI in Africa particularly.
2. The Strategy’s Development
With statements indicating a lack of evidence on various assertions made in the Strategy when alluding to the prevalent African reality, one wonders how the massive projections made on the changes envisioned were ascertained? While a high level document should be ambitious, isn’t it equally important to plan with feasibility in mind as made possible with data obtained about our context first? Preliminary projections evidencing how for example AI is expected to revolutionise sectors like health and education would be an ideal start to present the assertions made in the Strategy where then we have these foundational assumptions which more realistically match our context’s underpinnings whether that be resource, capacity, power and so on. Going into phase one of the implementation of this strategy at National level, leading this process with evidence based thinking would be helpful in decision making. A truly people centric approach ought to start from having an in depth understanding of our own realities and needs and while comparison with the rest of global conditions is a helpful exercise in learning lessons from steps taken by them, basing our analysis of the African condition and aspirations for AI based on this is counterproductive as we ourselves ought to be the central point of analysis that informed the propositions advanced in the Strategy.
Central to the development process of the AI Strategy also is the collaborative process that was undertaken to ultimately inform the Strategy. While ideal collaboration which speaks to working across diversity in discipline and composition isn’t always fully attainable, it remains important to work towards it for the benefits it poses especially for a multidimensional realm like the AI space which puts across scientific, social, political and economic consequences of a wide range. The AI Strategy mentions being a result of the technical team which led the Study as well as other African experts from different member states, the Regional Economic Communities (RECs), AU Specialised institutions plus the virtual multi stakeholder consultations which were held in April whose findings were taken as a reflection of the aspirations of the African people. This might look like a sufficient enough base to have enough representative voices however some criticism has come up as to the nature of the consultations and how that hampers adequate participation of even stakeholders that presumably have a seat at the table like Civil Society Organizations (CSOs) including limited time, limited access and so on.
Even more so is the issue of the general public’s involvement. Here, there appears to be a need to step back from this “knowledgeable or intellectual” stance with which players in the AI space communicate which casts the illusion that the general public is unable to grasp the essence and meanings of these systems as they get deployed upon them. Instead, finding simple language that can be used to communicate to people whose interests are supposedly represented becomes even more critical towards engagement which fosters collaborative thinking and subsequently, taking on large scale action together .
Furthermore, as part of the development process of the Strategy and its implementation both at national and continental levels, the question of dependence on foreign players comes up especially with regard to agenda setting of priorities that may not fall in line with the African people’s priorities with what they want from AI. A number of initiatives have been cited in the Strategy as part of the ongoing work to ready Africa for its AI future for example towards inclusive growth and development or towards ethical and responsible AI. While these are much needed interventions, careful attention must be paid to making sure that these global partnerships do not become leaders of the AI conversation in Africa. How can a global big tech company that is offering solutions to a country’s public instead of its government be checked to ensure it respects the aspirations and will of the particular country? How can a legacy multilateral organisation funding government AI initiatives respect the independence of the nation in its workings without unnecessary interjections especially around guiding ideologies and so forth. Our nations already are and will have to contend with these questions especially at a continental level if we are to realise the Africa-centric and people-centric AI we are envisioning. We ought not to shy away from addressing deep rooted issues such as colonisation or imperialism but rather have them integrated in the broader AI for development discourse.
By extension to the above analysis is the realisation that, in approaching AI as a neutral scientific artefact we run the risk of failing to see the social, political and economic impacts it creates on the whole and not just in the obvious places we tend to look. While this may or may not be a result of the voices captured in the writing of this Strategy, having a multidimensional team may have helped with pointing out some complexities pertaining to AI generated implications. For example, the issue of extractive practices surrounding dataset collection in Africa seems to be ignored in view of the immense need for datasets for Africa which is approached through the lens of limitations of capital. However, while that is true, what could we anticipate with looking to fill that gap with foreign corporations or organisations who have interests that may not always align with the overall advancement of our societies- e.g. what good is helping in building datasets that we may have no control over? Another example is the issue of media and information literacy brought up in the Strategy which paints a picture of it being a solution to issues like hate speech, disinformation and misinformation- not beginning the analysis of the problem from its roots and how majority of these issues fall under the broader Technology Facilitated Gender Based Violence (TFGBV) which is as a result of patriarchal and sexist attitudes towards women means we are likely to deal with such an issue longer since we are not addressing it wholly. This kind of analysis which moves away from unidimensional problematization would be helpful in more effectively approaching AI-generated implications of today.
Finally here is the issue of AI divides in Africa especially as evidenced by AI venture capital investment in just about three to five countries leaving others out. Lessons can be picked from earlier ICT initiatives in governance such as the Malabo Convention on Cyber Security and personal data protection whose implementation including the ratification process was strained partly because of these national technological divides. Addressing this is a priority to achieving collective continental AI objectives where nations are moving more in tandem and are aligned in agreement with the AI visions without which this process simply becomes null. Even more, in line with being truly people centric, AI high level strategic planning should have the broader African citizenry at its heart and not just the minority tech savvy middle class which has been given priority thus far.
3. On Prioritisation in the Strategy
What guiding ideologies have been prioritised in the Continental AI Strategy? The prioritisation of putting in place the necessary conditions for the smooth running of markets over human impact eats away at the capacity to shape discussions on how to integrate these systems in our contexts. For example, where the standard of value for progress is positioned as levels of production or amount of investment in AI with little to do with how the common African is benefiting or how these systems have empowered them distorts us from more well rounded analyses that account for our societies in their fullness. Therefore, planning which is more alive to the reality as to why discrepancies exist in our contexts, for example why majority African countries are information poor as opposed to the global information rich North, would be more helpful in guiding more appropriate placement of our priorities in national policy making.
Additionally, the understanding of the African masses of the AI space as explored earlier in this essay is also a critical point of investigation of the priorities that have been set out by the Continental AI Strategy. Prioritising the public sector or the private sector and a minority tech savvy elite works to marginalise the needs and realities of the other part of the population not accounted for in high level planning. The question here begs, which groups of people are being prioritised in AI decision-making for Africa?
Finally, the Strategy broadly points to an Africa-centric AI ridden with African values especially towards an envisioned cultural renaissance enabled by AI. While this is a noble priority to have for Africa’s AI overarching vision, there is not enough definitiveness to what this means as its mostly broad statements offered on the whole. Natural Language Processing (NLP) for instance is one clear area where we have been made to understand that integration of our local languages in the online world will not only work to include us in there while communicating but that it’ll also aid in the digital preservation of our languages. Beyond that, statements such as creating African datasets will automatically lead to a cultural renaissance seems to be a blanket statement with little evidence to show for it. Still, the point of inquiry here is on the idea of an African AI stance; what does Africa-centric AI truly mean? What elements in the Strategy and subsequent National AI Strategies will capture this idea?
Conclusion
In taking part in the above exercise of inquiry along the foundational aspects of the AU-AI Strategy, the aim is for African people of different existences and policymakers to refine micro-visions under the broader vision proposed by the AU-AI Strategy such that they are able to better plan for integration of AI in their communities in a way that allows for their collective thriving. This rethinking as we get into implementation involves accounting for asymmetries of relations and society towards making these more equitable. Overall, to overcome the structures through which people are oppressed and dominated should be at the centre of the development ambitions of Africa aided by AI and as such, we get to acknowledge that science and tech provide fresh sources of power that also necessitate fresh sources of analysis and political action.
Author: Bobina Zulfa, Data & Digital Rights Researcher, Pollicy
