Keywords

Introduction

The accelerated acceptance and deployment of digital technologies (especially Artificial Intelligence-AI) occasioned by the COVID-19 pandemic (Chang 2020; Kumar et al. 2021) is a pointer to the role technology has to play in our societies today. Public and private entities significantly increased the use of or employed a number of AI tools, digital platforms, big data and robotics as public service delivery tools, education platforms or work-based solutions during the global crisis. This clearly demonstrates that AI and other disruptive technologies are fast becoming critical foundations that enable human flourishing. AI is a major element and driver of what many have described as the fourth industrial revolution which is the functional convergence of AI, robotics, the internet of Things (IoT), 3D printing, genomics, quantum computing, blockchain and other disruptive technologies. Conceptualised by Schwab (2016), the idea of Fourth Industrial Revolution builds on the preceding industrial revolutions characterised by water and steam power, electric power and electronics and information technology. Schwab’s argument was rooted in the fact that the convergence of disruptive technologies is blurring lines between the physical, digital and biological spheres as well as transforming every industry through unprecedented velocity, scope and systems impact. And at the heart of this epochal revolution is AI—a technology increasingly misconceived, misunderstood and mischaracterised. The focus on AI in this chapter is informed by the desire not to confuse its impacts with other emerging technologies.

AI offers exciting possibilities for African societies, promising among other things to alleviate poverty, reduce economic inequalities and improve access to public and private services in health, transportation and education. Whereas there is still a huge gap between Africa and other developed parts of the world in terms of capacity to develop datasets, AI systems development and deployment, AI is gaining traction in many aspects of life in Africa. One can say that, as many parts of Africa are yet to benefit fully from the first three industrial revolutions (electricity, mechanisation of production and automation of industries), a critical element of the Fourth Industrial Revolution (4IR), AI has made inroads into African socio-economic fabric. This raises the question of whether Africa should focus more on AI capacity building rather than embracing transferred technologies from the Global North in order to ensure that the full benefits of the 4IR are obtained for all sections of the society. On the other hand, AI not only creates unprecedented opportunities with a real bearing on people’s lives, it raises fundamental questions on fairness, human rights, privacy, bias, security and the future of work among others. This chapter provides insights and perspectives on the AI landscape in Africa and a snapshot of how the future of AI development and deployment should look like. It explores the question; how can stakeholders in Africa ensure that Africa has sufficient capacity for Responsible AI? In this chapter, we present views on how future discussions of Responsible AI in Africa can be shaped. It starts with the presentation of the current landscape of AI deployment in Africa, highlighting potential socio-cultural impacts, ethical and legal impact of AI considering the unique cultural dimensions in Africa and providing recommendations of how Africa can achieve Responsible AI. We conclude with mapping the roles Africa can play in the global discourse on Responsible AI. This contributes to the emerging debate on AI ethics, regulation, policy and governance in Africa.

The Meaning of AI

The meaning of AI is often contested across different disciplines. Whereas it was established in the 1950s as a field of study, narratives of intelligent machines have had a very long history (Cave et al. 2018). However, since Turing (1950) posed the question, “Can machines think?” AI has grown significantly as a major branch of computer science concerned with the design and application of machines capable of performing tasks that normally require human intelligence. Alan Turing’s paper on Computing Machinery and Intelligence can be said to have laid the fundamental goals of AI and in 1956 John McCarthy organised a conference in Dartmouth where the term artificial intelligence was first adopted. McCarthy and Marvin Minsky, acknowledged as two of the pioneers of the field, subsequently co-founded the Artificial Intelligence Project (now the MIT AI Lab) in 1959 to explore the potential of AI. The term has since been defined differently in the research and innovation ecosystem. Minsky (1968, p. v) defined AI as “the science of making machines do things that would require intelligence if done by men”. This definition reflected an earlier definition of AI proffered by McCarthy et al. (1955) as the problem of “making a machine behave in ways that would be called intelligent if a human were so behaving”. In their work, Artificial Intelligence: A Modern Approach, Russell and Norvig (2002, p. viii) in answering Turing’s question described AI as “the study of agents that receive precepts from the environment and perform actions”. They went further to expose four different approaches that have characterised the history of AI including agents that think humanly, think rationally, act humanly and act rationally.

However, following an understanding of human intelligence as a “product of many factors and subject to innumerable influences” (Wechsler 1975) or as Gardner (2000, pp. 33–34) defined it, the “biopsychological potential to process information…to solve problems or create products that are of value in a culture”, there is a question of whether a machine can achieve full human intelligence. Can a computer/machine acquire full socio-cultural, psychological intelligence like humans do? These questions are at the heart of the categorisation of AI. AI applications that can only perform specific tasks are generally known as having artificial narrow intelligence (ANI) or weak AI (Shane 2019). On the other hand, some have described AI systems—artificial general intelligence (AGI) that can be able to “reason, plan, and solve problems autonomously for tasks they were never even designed for” (Kaplan and Haenlein 2019). Or as Searle (1980) described; designed in a way that the “computer is a mind, in the sense that computers can be literally said to understand and have other cognitive states”. Kaplan and Haenlein (2019) also raise the possibility of a third category of AI called artificial super intelligence (ASI) “which are truly self-aware and conscious systems that, in a certain way, will make humans redundant”.

For the purposes of this paper, the use of the concept of AI remains at the level of ANI which we define here as systems or applications that have the ability to interpret and learn from data for the performance of identified tasks in an agile way. Big data, machine and deep learning are critical drivers of AI which are increasingly applied in many aspects of our lives. Whereas AGI and ASI remain only possibilities and not currently available, access to ANI applications is becoming ubiquitous and pervasive even in many parts of the world including the developing economies of Africa. This conceptual clarification of AI is critically important to the present discourse because a mischaracterisation of what current AI applications can do, can affect its design, implementation, regulation and overall governance. Responsible AI governance should be rooted in a clear understanding of the nature, scope and potentials of AI. There is no attempt to engage with the conceptual tensions surrounding AI in this chapter but it was necessary to provide a clear view of what we mean by AI.

The Current Landscape of AI Deployment in Africa

There is an increasing level of AI (as defined in this chapter) deployment in Africa. Sectors where AI has been employed include; healthcare, education, transportation, financial services, agriculture, public services, security, business management and telecommunications. This chapter highlights some of the AI systems deployed in these fields. It is important to note that many of these systems are developed by local experts but most influenced, sponsored or controlled by big tech companies from the Global North.

Healthcare

AI is being deployed for a number of services in healthcare in Africa. For instance, MinoHealth AI LabsFootnote 1 in Ghana is using AI for automated diagnostics, forecasts and prognostics. BareAppFootnote 2 (also developed in Ghana) combines AI and skin expertise to diagnose black skin diseases and make recommendations for best treatment. In Nigeria, RxAllFootnote 3 is enabling pharmacies and patients to avoid buying counterfeit medicines online through a Deep Learning-Hyperspectral IoT platform for authenticating drugs in real time. VinsighteFootnote 4 has developed an AI-powered system that detects eye diseases at an early stage and aids the visually impaired to read books and navigate their environments independently. InStratFootnote 5 (also developed in Nigeria) uses AI to detect and predict possible disease outbreaks by electronically collecting and analysing clinical and non-clinical data. In Uganda, Chil AI Lab GroupFootnote 6 has developed a system that combines AI and other emerging technologies to enhance management of female chronic diseases. The overall aim, however, is to provide accessible and affordable chronic disease prevention and management to women. AI is also being deployed for non-clinical healthcare purposes such as insurance. Deployed in countries such as Nigeria, Ghana and South Africa, CuracelFootnote 7 uses AI to optimise health insurance claims.

Financial Services

Financial services sector is another field where AI is being deployed in Africa. In Kenya, M-ShwariFootnote 8 has developed a system that relies on AI to review online loan applications, helping it to consider applications from customers who live far from bank branches. Similarly, In Egypt, CassbanaFootnote 9 relies on AI to create digital identities for underserved communities so as to integrate them into the banking system. It also manages their financial requests and builds a behaviour-based scoring system for them. In Nigeria, Debtors AfricaFootnote 10 and KudiFootnote 11 are two of the AI applications deployed in the financial sector. Debtors Africa uses AI to automatically update its independent, searchable database for recalcitrant and delinquent debtors, providing status of debtors in real time. Kudi on the other hand is a chatbot that responds to financial requests and allows users to send money and pay their bills.

Security Services

Private and public security services are also deploying AI systems. To provide a solution to the well documented risks of misidentification, unfair discrimination and bias against black people in the use of biometric surveillance technology, a Ghanaian company BACE GroupFootnote 12 has developed BACE APIs that enables accurate identification of black people through facial recognition technology. This company provides secure identity verification as a security service to both the public and the private sectors. Similarly, in Kenya, the government has deployed AI-powered facial recognition technologies to complement policing efforts as part of the Safe city Project.Footnote 13 This technology was developed by the Chinese tech giant Huawei and concerns have been expressed regarding the dearth of regulatory provisions that can ensure the responsible use of this technology considering the well documented legal, ethical and socio-cultural concerns related to facial recognition (Feldstein 2019). This is because there have been reports of a similar technology being purchased by the Uganda government from Huawei to spy on political opponents (ibid.). Tabiri AnalyticsFootnote 14 (which is deployed in Rwanda, Kenya and Uganda) has also been used to provide continuous monitoring service to prevent cyber threats. This system uses cloud computing, machine learning and AI to automate human analysis of IT system log data to achieve cybersecurity. Global Auto SystemsFootnote 15 is also using data and system analytics to provide security systems for schools, colleges and universities in Uganda.

Education

In Uganda, M-ShuleFootnote 16 has developed and deployed a Toolkit that helps learners build academic and life skills with interactive, self-paced and personalised learning over SMS, measure progress and performance and also keep stakeholders up-to-date with awareness campaigns and situational response information. It also helps to collect data and insights from stakeholders via SMS to make real-time decisions. In Kenya, Eneza EducationFootnote 17 is providing primary and secondary school students with virtual tutorials on curriculum-aligned content in all subjects while in Nigeria TuteriaFootnote 18 uses AI to link qualified tutors to students within a particular area and budget. It also verifies tutors IDs, conducts background checks and evaluates tutors’ performance. In South Africa, Botlhale AIFootnote 19 solutions specialises in conversational AI. With a suite of Natural Language processing tools, this company ensures that those who speak African languages do not miss out on the benefits of technologies.

Transportation and Logistics

AI is also changing the transport sector in some African countries. In Egypt, SWVLFootnote 20 and Softech TechnologiesFootnote 21 have deployed AI applications making huge impacts. Swvl uses AI to coordinate a fleet of private buses, allowing commuters to bypass often congested public transit networks while Softech uses AI to help commuters to plan their itinerary by collecting and analysing data on transportation conditions. Softech provides both an AI-based digital planner and solutions for commercial fleets providing end-to-end visibility and command and control over B2B logistics, transportation and mobility operations. In Kenya, AmitruckFootnote 22 is being used to create a digital marketplace for trucking; connecting transporters and clients on a digital platform while Kamtar is used in Ivory Coast to connect shippers and carriers. Kobo360Footnote 23 is also using big data analytics and technology to reduce logistics frictions in Nigeria. The overall goals are to ensure efficiency and cost reduction in the supply chain.

Telecommunication

In Telecommunication, MTNFootnote 24 (a mobile Telecom operator which is in operation in Benin, Cameroon, Ghana, Guinea Bissau, Ivory Coast, Nigeria, Rwanda, South Africa Uganda and Zambia) has launched a chatbot that simplifies and enhances the quality of customers’ experience. Similarly, OoredooFootnote 25 partners with PI works (an AI company) to enhance customer experience, network coverage and connectivity in the MENA region particularly Tunisia and Algeria. SafaricomFootnote 26 has also introduced its AI Chatbot assistant to popular messaging service WhatsApp to perform telecom-related tasks as well as answer queries regarding M-PESA (mobile money service) in Kenya.

Public Service Delivery

The most prominent use of AI for public service delivery can be found in Rwanda where robotsFootnote 27 donated by the United Nations Development Program (UNDP) were introduced to help in the fight against COVID-19 pandemic as part of public health intervention. These robots were used for a number of tasks including temperature screening, detecting people not wearing masks in healthcare settings and delivering medicine, food and other essentials in place of frontline health workers. In South Africa AI is being used to detect gunshots. ShotspotterFootnote 28 is used to fight wildlife poaching in Kruger National Park.

Politics

There is also documented evidence to indicate that AI tools have been employed in the African political landscape (e.g. in Kenya and Nigeria) as a tool for mis/disinformation. However, AI tools like the one developed by African check,Footnote 29 a South African organisation using AI technology to fact check political claims, are helping to fight misinformation. In the same vein, the protests that followed the #ENDSARS hashtagFootnote 30 against the Special Anti-Robbery Squad (SARS) in Nigeria in 2020 demonstrated the potential impact of AI systems in changing political narratives for the citizens. AI-powered digital tools are increasingly changing the way politics and public sector decisions are and while providing citizens with information on their rights.

Agriculture

AI Mozambique, Hello TractorFootnote 31 is helping farmers share equipment. It also leverages machine learning to predict crop yields and facilitates access to financing for farmers. Other AI tools used in Agriculture include Agrix TechFootnote 32 (in Cameroon) that helps farmers detect crop diseases and propose sustainable and environmentally friendly solutions to small-scale farmers and PlantVillage NuruFootnote 33 in Kenya that serves as a crop disease diagnosis tool. In Egypt, AbuErdanFootnote 34 uses deep learning neural networks and predictive analytics algorithms to forecast chicken future performance. Apollo AgricultureFootnote 35 in Kenya uses satellite data and machine learning to advise farmers on credit decisions and automated operations to keep costs low and processes scalable. Similarly, AeroboticsFootnote 36 is being deployed in South Africa to assist farmers/growers to make informed decisions. This application uses AI to collect and analyse data on crop yields and subsequently predict future performances.

The Future of AI in Africa

The current landscape of AI in Africa is dotted with the presence of big tech companies from the Global North including Google, Facebook, Alibaba Group, Amazon, Microsoft and IBM Research. These companies bring improved capacity and enabling infrastructure to make data more pervasive and valuable which can be leveraged by AI and other emerging technologies to drive large-scale transformation in Africa and make the continent more competitive. This will require increased sharing, interoperability of data processing systems and significant convergence of emerging technologies. AI presents good opportunities for many sectors to optimise solutions to Africa’s problems. The future of many public and private sectors in Africa will be intricately linked with the future of AI. In healthcare, AI can provide solutions to many challenges including in medical diagnostics, drug research and discovery, clinical trials, disease management, pharmacogenomics, improvement of patient outcomes, data management and clinical decision support tools. In banking and other financial services, AI can improve risk detection and management in addition to data management. There are also prospects of using AI as a new tool for counterterrorism in Africa. McKendrick (2019) and Ramanouski (2019) have described how AI can theoretically contribute to counterterrorism operations. Therefore, with the continued activities of terrorist groups such as Boko Haram and al-Shabaab in Africa, AI can provide a potent tool for counterterrorism. Although a joint report by UNICRI and UNCCT have also detailed the possible malicious use of AI for terrorist purposes (United Nations 2021).

Furthermore, AI holds the promise of enabling a revolution in how agriculture is done in Africa. From crop and soil monitoring, improved plant and crop disease diagnosis, crop yield prediction and price forecasts, intelligent spraying, pest control, drought prediction, to agriculture robots and genomic precision, AI has the potential to transform the Agriculture industry in Africa. In addition, Africa’s educational system has many challenges in personnel and facilities that AI can address in future. This includes among other things AI-powered virtual teaching assistants that can help both teachers and students in assessments and providing feedback. Considering the lack of physical infrastructure and insufficiency of training for teachers in Africa, the potential impact of AI on learning may be extensive. There is sufficient evidence to show that there will also be possible increases in the use of AI in politics, public service delivery, transportation (e.g. driverless cars) and the military in Africa in the next two decades as they are currently used in the Global North.

However, despite the potential benefits of this technology, the design and implementation of AI systems raise significant ethical, legal and socio-cultural challenges. There is a growing body of the literature to highlight that AI design and implementation are not only changing socio-cultural dynamics but also exacerbating existing societal inequalities, biases and stereotypes (Nelson 2019; Weber 2019). Many AI systems ranging from applications for predictive policing (McDaniel and Pease 2021), to facial recognition technologies (Raji et al. 2020) have shown to be inherently biased and discriminate against certain sections of the society. These biases can creep into AI through the underlying datasets or algorithms. Both the data and algorithms can include biased human decisions or reflect historical inequalities bordering on gender, race, social status and geographical location. Most importantly, black people have been shown to be disproportionately affected by unfair bias in current AI systems which puts Africa at a disadvantage. The impact of such biases inflicts hurt on those who are discriminated against. It brings mistrust and possible unacceptance which reduces its potential benefits to businesses and the society at large.

AI also raises challenges for human rights in a number of ways. From the creation of autonomous and intelligent agents like driverless cars, neurotechnologies that could clearly disrupt people’s sense of identity and agency (Yuste et al. 2017), to the possibility of digital authoritarianism, AI challenges established perceptions of human rights which need attention. The possibility of deploying AI-based surveillance technologies by governments is a challenge that requires attention since human rights reports in Africa do not look good. However, the above concerns related to unfair biases and discrimination and human rights together with the high energy requirements for AI systems contribute to conclusions that AI can inhibit some of the UN sustainable development goals (SDG) targets (Gupta et al. 2021; Vinuesa et al. 2020). These and many more negative impacts and unintended consequences of AI applications call for a design and implementation of this technology in a way that is ethically responsible, legally compliant and socio-culturally acceptable (Wakunuma et al., 2022).This is the concept of Responsible AI. Realising the full capacity of AI for human flourishing depends on Responsible AI and therefore should be a major agenda for AI discourse in Africa.

What Can Africa Do to Achieve Responsible AI?

AI and other emerging technologies are characterised by features such as logical malleability, ubiquity, pervasiveness, interactivity, possibility of augmentation and potentially autonomy. ‘Logical malleability’ (Moor 1985) makes it difficult to predict how AI systems can be used or others discussed under the concept of interpretive flexibility (Doherty et al. 2006). Its ubiquitous, pervasive and interactive nature continues to be more pronounced as new use cases emerge. Together with the potential to achieve autonomy, these features mean that both the design and deployment require the consideration of ethical, legal and socio-cultural values and principles because of possible intended and unintended consequences. These features inform the many uses of AI including for improving processes and efficiency, social control and to promote human flourishing (Stahl, 2021). However, for AI to promote human flourishing especially in Africa, the principle of responsibility needs to be integrated into its design.

Responsible AI is about how AI can be sensitive to human values (which is shaped by cultural beliefs and systems) and to societal needs, expectations, hopes and fears. As Dignum (2017) opined, Responsible AI rests on three pillars; the willingness of stakeholders to accept responsibility for the impact of AI, the development of mechanisms that can enable AI systems to be sensitive to ethics and human values and appreciation of different impacts of AI in different cultures. These pillars shape initiatives in AI education, governance, regulation, risk assessment approaches and quality assurance. For instance, the promotion of AI governance comes from the willingness by public and private entities to address the impact of AI. However, the question for Africa is; do we have the necessary technical and socio-economic infrastructure to facilitate these? All available indexes point to the fact that Africa lags behind in the comparative global AI readiness. Recent reports on the current state of the art on AI in Sub-Saharan Africa confirms the evident lack of AI capacity in the AI ecosystem and suggest the need for greater capacity (Butcher et al. 2021; Gwagwa et al. 2021). One thing that is clear is that the role of ensuring that Responsible AI is achieved in Africa is firstly ours before any other person else. We are mainly responsible for ensuring that AI designed and deployed in Africa is sensitive to our socio-cultural contexts but not only ours.

The previous section has mapped the future of AI applications in Africa and their potentially historical impacts, we will now provide perspectives on how Africa can not only increase AI capacity but how to achieve Responsible AI considering the continued global discourse on AI ethics.

Framing the Role of AI for Africa

Africa is historically credited with contributing to the industrial revolutions of the eighteenth century. It was Karl Marx who wrote that the “turning of Africa into a warren for the commercial hunting of black-skins” contributed to the “rosy dawn of the era of capitalist production”. Eric Williams (1944) also echoed the role slavery had on capitalism in his work titled Capitalism and Slavery. According to Parvanova (2017) this industrial revolution became the force behind colonialism because it created the need for Europe to expand; increased production capacity required more raw materials to satisfy demands. However, a number of factors including colonialism impeded the spread of industrial revolutions in the dependent countries of Africa. Alam (2012) has provided empirical evidence to demonstrate the impact of colonialism on industrial Revolutions in Africa. Unfortunately, since the end of colonialism, Africa has not caught up with subsequent industrial revolutions characterised by digital technology and the results are in the social, economic and digital inequalities between the Global North and the Global South. The first three industrial revolutions have shaped the societies we live in today and it is safe to say that Africa has not benefited fully from them. About 40% of Africans still do not have access to electricity and fully automated productions are yet to be achieved. And with the 3rd industrial revolution came the challenge of digital divide. According to a 2021 report by the Ibrahim Forum, “89% of learners in sub-Saharan Africa do not have access to household computers. 82% lack internet access and at least 20 million live in areas not covered by a mobile network” (Mo Ibrahim Foundation 2021).

With AI driving the 4IR, there is a potential to further divide humans on a class level and Africa may be disconnected or will not receive the same level of benefit from AI systems. Some have also pointed out that AI raises the risk of neo-colonialism with regard to data and the algorithms that shape AI. AI is as good as the datasets and the algorithm that shaped it. That means that there is inherent power and control in datasets. Allowing the Global North to own and control the datasets that shape AI systems developed for Africa amounts to what many have described as ‘data colonialism’ (Couldry and Mejias 2019; Viera Magalhães and Couldry 2021).

Beyond data there are also the algorithms that are not neutral (Mittelstadt et al. 2016; Warfield 2020; Stinson 2021). According to Mittelstadt et al. (2016), ethical issues related to algorithms include both epistemic and normative concerns; possibilities of unjustified actions, opacity, bias, discrimination, challenges to autonomy and informational privacy. Birhane (2020) has also written about the possibility of ‘algorithmic colonization’ to show that algorithms contain the biased interests and values of those who develop them which are often overshadowed by the hype around AI. The argument here is that Africa needs to move beyond the hyperbolic language surrounding AI to understand the true facts about datasets and algorithms and their inherent power in AI. Those who have the data and the algorithms will hold great power and influence; great power to improve processes and procedures, for social control and to acquire better human flourishing but in a way to favour them. An example is the development of COVID-19 vaccines. Despite the potential benefits of equitable access to vaccines, the greatest barriers to adequate vaccine supply remains intellectual property (IP) protection governing the production and access to vaccines (Erfani et al. 2021). Those who produced the vaccines and own the IP remain in the position of power and lack the political and moral will to waive their IP rights to facilitate equitable access to the vaccines. With these in mind, Africa needs to be deliberate in framing the role AI can play for the continent and in the development and definition of our data and our algorithm. Responsible AI in Africa means AI developed with African data and culturally sensitive algorithms. Policymakers in public and private sectors, researchers, industry players and all stakeholders need to decide on what kind of AI-driven society we want. The role of AI in Africa should be to provide fundamental solutions that can level up evident inequalities in production, in healthcare, education, gender and other spheres of life. AI in Africa needs to amplify our positive cultural contexts. Over reliance on importing AI systems driven by foreign values and principles can only exacerbate the risk of neo-colonialism. Africa’s attitude towards the 4IR, especially AI does not need to mirror how the other industrial revolutions were handled.

Another thing to say here is that due to the pervasiveness of AI technology and the interconnectedness of the African societies, it will take the whole continent to frame the goal of AI around improving and promoting the unique cultural contexts Africa possesses. Africa’s framing of the goals of AI needs to therefore rely on the fundamental cultural narratives of Africans, consisting of our stories, our beliefs, values, needs, expectations, fears and concerns. Values and power are central in AI design and implementation, however, inherently-power-driven AI narratives from the Global North should not determine the paths we chart for our AI journey. Africa should proactively frame how AI should be developed and deployed in our communities (Eke and Ogoh, 2022). Lessons from the Global North should shape such a framing to centre on creating an ecosystem where AI can thrive and on embedding critical African values into AI systems to ensure that AI in Africa can truly conform to the principles of science for and with the society.

Identification of Relevant African Values and Principles to Be Embedded into AI Systems

The centrality of data in AI means that AI aligns with human preferences, interests and values. But the critical question is what and whose values should AI applications align with. At the foundation of global discussion on AI alignment are established ethical traditions (Yu et al. 2018) such as utilitarianism (Roff 2020), deontology (Hooker and Kim 2018) or virtue ethics (Neubert and Montañez 2020). Each of these ethical traditions are without roots in Africa and emphasise the importance of AI to respect the objective interests of humanity or a particular group of people. Impliedly, ethically-aligned AI discourse has focused on the principles and values from cultural contexts from which these frameworks emerged—the Global North.

In his book, Human Compatible: Artificial Intelligence and the Problem of Control, Stuart Russell (2019), observed that aligning AI to human values is the crucial goal of AI value alignment. This is particularly important because of the critical potential characteristic of autonomy of AI systems. The values embedded into AI, therefore, are critical to its impact on society. As AI systems are developed for African societies, it is critical to ensure that the values embedded in these systems represent objective interests and beliefs in Africa. In essence, to design AI for Africans requires the positive action of integrating African values and principles in the design and implementation. The first part of this is to normatively understand what values and principles ought to be embedded in AI systems. Africa has rich moral traditions built around core values of interconnectedness, solidarity, communality and respect conceptualised in ethical frameworks such as Ubuntu (translated broadly as “I am because we are”) and or ujamma (the spirit of brotherhood). These and other relevant African value-systems should form the central focus of AI value alignment in Africa. AI discourse in Africa ought to focus not only on what AI might do for Africa but also how AI should be done for the benefit of Africans.

For AI to be truly for society, there must be an understanding that the technical design needs to reflect societal values, needs and expectations. Unlike the usual tick-box exercise evident in most research and innovation processes, this is about a proactive and continuous consideration of the social and ethical consequences and a conscious integration of values in the design as well as the deployment of AI systems. As Stahl (2021) suggested, this demands a fundamental rethink of the relationship of AI research and innovation and ethics. African ethical principles should be an integral part of AI’s scientific excellence in a way that promotes solutions tailored for African societies. Our values should be central to the intended consequences of AI in Africa; included in the risk assessments and form part of the evaluation of trustworthiness and responsibility of the AI systems. Interventions needed to address identified risks should reflect contextual African values. However, the identification and applications of these values and principles require the collaboration and contributions of diverse stakeholders from all AI ecosystems (Stahl 2021).

Increased Involvement of Relevant Stakeholders

In his book, Stahl (2021) highlighted the idea of AI ecosystems which is already established in the European Commission’s AI White paper (European Commission 2020), global recommendations from OECD (2019) and UNESCO (2020) and also in the UK’s Digital Catapult (2020). Using ecosystems as a metaphor, Stahl conceptualised AI ecosystems (consisting of individuals, organisations, innovation systems and landscapes) as examples of innovation ecosystems characterised by complex relationships between different and interdependent actors, willing to co-evolve and mutually learn as they drive change with openness. As he pointed out, there are elements of AI that are global but identifiable regional differences in the USA, China and Europe suggest separate ecosystems distinguished by geography, jurisdictions and other elements or environments within which the AI system is embedded. These environments range from “technical, policy, economic, legal, social, ethical and other aspects that closely interact with AI” (Stahl 2021, p. 93) and produce diverse stakeholders that can influence how societal impacts of AI are perceived, identified and can be addressed. It is important for Africa to identify these AI ecosystems (at the broader continental and national levels; relevant disciplines and sectors) and the stakeholders therein for a process of co-creation of frameworks for AI. There are diverse cultural contexts (languages, values, belief systems, etc.) and interests (educational, political, economic, legal, etc.) in Africa which need to be represented in the discourse on both what AI might do and how it should be done. The recognition of these contexts is the foundation of Responsible AI (Table 1).

Table 1 Roadmap to achieving Responsible AI in Africa
Full size table

The Role of Africa in the Global AI Discourse

As we have highlighted in earlier sections, every region, nation and cultural community has a role to play in shaping the discussions around Responsible AI. The approach to Responsible AI should be bottom-up rather than top-down. A clear conceptualisation of contextual values, needs and interests should precede acceptable international frameworks for AI governance and principles. Therefore, Africa nations in particular and the region in general, like other countries and regions in the Global North, have key roles to play due to the level of distribution of power that comes from global AI governance. We have interests, needs, hopes, fears, principles and values that need to be factored in the global consideration Responsible AI. This starts with having clear normative and epistemic understandings of unique African perspectives that AI design and implementation should align with. Such understanding should shape Africa’s roles in global discussions on AI ethics. Africa should not only be included in the global discourse on Responsible AI but should approach the proverbial table with their interests, expectations as well as their values and moral principles such as the communitarian principles exemplified in concepts such as ubuntu and ujamma.

In addition to this, Africa offers the global AI design and implementation landscape opportunities of diversification and generalisability in terms of datasets, skillsets and personnel to achieve Responsible AI. Disproportionate amount of data collected from the Global North contribute to the persistent challenges of unfair bias and discrimination in AI. The data, values and humans who build and deploy AI should include representation from backgrounds in Africa. This focus on diversification can ensure better AI outcomes informed by reliable insights from sufficiently representative datasets. A recent report on AI in Sub-Saharan Africa observed that such a representation should be more than a box-ticking exercise and must be seen as a moral imperative for all stakeholders (Ndung’u and Signe 2020) Stakeholders ought to appreciate the intrinsic values of diverse interests, expectations and perspectives inherent in the data that inform AI. The inclusion of new voices, perspectives and datasets provide new opportunities for designing solutions for more people. For instance, facial recognition systems developed by big tech companies in North America, Asia and Europe predominantly misidentify people based on race and gender owing to both the data and algorithms that shape these systems (Buolamwini and Gebru 2018). In contrast to these systems, a Ghanaian tech start-up (led by Ivorian researcher Charlette N’Guessan) has developed a facial recognition system for the local market trained with more diverse and representative datasets that can accurately identify black faces. Since AI is still at its introductory stage in Africa, there are opportunities of creating systematic diversity mechanisms to reduce the discriminatory effects of AI in the society. Building on from the growing body of AI ethics literature and practice, Africa can build a template for a sustainable consideration of diversity and inclusion in AI design and implementation.

Conclusion

Shaping the future of Responsible AI in Africa is a pertinent concept particularly when we consider the importance of developing Africa’s own contribution to the discourse of AI. For a long time, there have been discussions around AI including those around benefits but also around the ethical challenges associated with the technology. These have mainly been led by the Global North with little contribution from the Global South. This chapter has sought to make a contribution to the discourse of AI from an African perspective by focussing on what AI means and how it may look like for the future when we consider both the benefits and challenges of AI. The chapter has shown that there are clear benefits of AI by looking at a plethora of AI application areas in Africa. In particular, AI is being applied in healthcare, in finance, security, education, transport and logistics, telecommunication, public service delivery, agriculture as well as being used in politics. This is an interesting array of application areas and showcases the fact that although there is a limited discourse of AI on a global level, Africa has taken to AI and continues to do so in abundance. This indicates that there is a need to understand how AI is being applied in an African context and subsequently a need to understand what value-systems are being applied or can be applied as AI becomes mainstream in the African context. By cultivating this understanding, we can then begin to explore the possibilities that lie ahead in future of AI in Africa and as a consequence what can be learnt and shared by Africa in the global discourse of AI. Currently, this chapter notes that AI applicability is synonymous with big tech companies from the Global North, therefore raising concerns around dependency in terms of technology know-how, capability, capacity as well as the inculcation of value-systems from the Global North to the Global South. This raises further questions around the potential and possibility of digital/neo-colonialism which can leave Africa grappling with the technology and not being able to understand fully or find solutions for challenges that result from AI as it is applied in different domains. Simply put, the needs of the Global North are different from those of the Global South, as such, it goes without saying that the application of AI on the African continent may be different in terms of the problems it intends to solve and subsequent benefits the technology will have. Similarly, the ethical and social challenges that may result will differ in a number of ways when compared to the Global North. As such, this calls for Responsible AI particularly in as far as understanding the value-systems and human values that may be applicable when it comes to AI on the African continent. Africa’s challenges are vast and include but are not limited to hunger, poverty, education, health, climate, gender disparities and various inequalities, climate among others. The application and use of AI will go a long way in mitigating some of these challenges, however, this can only be done by incorporating the continent’s values in the technology and not being overly reliant and dependent on those from the Global North which have been embedded in AI due to technology’s origins. Despite AI’s origins, as its use spreads across the globe, there is room to tailor it to the locale of its adoption and use for it to be effective and truly meaningful. It is for this reason that in this chapter we allude to and recognise the importance of embedding African value-systems and principles through philosophies like Ubuntu, Ujamma and others in our quest of framing a truly Responsible AI for Africa. In this case, Responsible AI means to think about, anticipate, design, implement, adopt, adapt and use AI that connects, is communal, respects and works in solidarity with different stakeholders for the common good and meeting head-on the challenges that Africa faces.