Artificial intelligence, developing-country science and bilateral co-operation

P.M. Addo
Agence Française de Développement (AFD)

COVID-19 sparked a range of uses of artificial intelligence (AI) in the search for solutions and underscored the importance of data for policy making. This essay points to the discrepancies in AI capabilities between rich and poor countries. It then considers how bilateral and multilateral development co-operation could assist, specifically in connection with AI in science.

The use of AI for research and development (R&D) is still out of reach for most researchers in developing countries. Europe, North America, and East and Central Asia are the world’s dominant sources of AI conference publications. In 2020, East Asia and the Pacific accounted for 27% of all conference publications, North America 22%, and Europe and Central Asia 19%. By contrast, sub-Saharan Africa accounted for just 0.03% (Zhang et al., 2021). Furthermore, researchers from developing countries often play little if any role in key international conversations on AI, especially those held in the United States, Canada and Europe.

In science and more generally, most developing countries are not yet well prepared to harness the opportunities presented by AI technologies. The overall gap in capabilities between developed and developing countries is evident in the findings of the Government AI Readiness Index 2021, which measures the capabilities and enabling factors required for a country to implement AI solutions (Oxford Insights, 2022). Billions of people are still without Internet access; basic technological and data infrastructure is often deficient; and R&D spending is limited. Meanwhile, datasets generated in developed countries are sometimes unsuited to training AI systems to meet local needs. Such deficits could exacerbate inequalities between high- and low-income countries in the productivity of science, in economic performance and in the quality of public services.

COVID-19 highlighted the need for collective global partnerships: development co-operation can help. This section highlights examples of bilateral and multilateral co-operation around AI.

Among other measures, development co-operation can provide fora for dialogue on shared challenges and relevant technological innovations, and help identify synergies between actions regionally and internationally. An example is the Africa Regional Data Cube, an initiative brought about by collaboration among many actors. These include the Committee on Earth Observation Satellites, Strathmore University in Kenya and the Global Partnership for Sustainable Development Data. Directly supporting activities in Ghana, Kenya, Senegal, Sierra Leone and Tanzania, the Data Cube has helped harness the latest Earth observation and satellite technology and data to address issues related to food security, urbanisation, deforestation and more (Global Partnership for Sustainable Development Data, 2018).

Development co-operation can also help countries advance data protection legislation, improve data infrastructures and strengthen overall AI readiness. A good example is the collaboration between The GovLab (an action research centre based at New York University’s Tandon School of Engineering) and the Agence Française de Développement (French Development Agency, or AFD). Together, they launched the recent #Data4COVID19 Africa Challenge. This supported Africa-based organisations to use innovative data sources to respond to the COVID-19 pandemic (Verhulst et al., 2022). Respect for data ethics and data responsibility were a key part of the Africa Challenge. Thus, every initiative proposed complied with the European Union’s General Data Protection Regulation.

Both bilateral and multilateral co-operation can also provide opportunities for knowledge sharing and talent attraction via mobility exchange programmes facilitated by reforms such as easing restrictions on visas.

Bilateral co-operation can also help plan, finance and assist implementation of research and technological development initiatives in an environment favouring multidisciplinary and multi-stakeholder collaboration. For instance, in 2021, France’s Agence Nationale de la Recherche, in partnership with the AFD, launched the IA-Biodiv Challenge, aimed at supporting AI-driven research in biodiversity (AFD, n.d.) This research initiative provides a space for scientists working on AI and biodiversity in France and Africa to mutually learn, share and engage.

Development co-operation can also go beyond sharing data to supporting open science initiatives. For example, most datasets on languages in Africa are not yet openly available, and local AI developers often need to use unrepresentative data from developed countries.1

In addition, grants could support investments in AI R&D in developing countries. This could include the creation and support for centres of research excellence like the African Research Centre on Artificial Intelligence (ARCAI) in the Democratic Republic of Congo (DRC). The ARCAI is the result of a collaboration between the Economic Commission for Africa and the DRC government. ARCAI will assist in AI research, collaborate with universities in Africa, participate in the creation of a network of researchers and contribute to training to help citizens actively participate in the digital transition.

Canada’s International Development Research Centre in collaboration with Sweden’s International Development Cooperation Agency launched the Artificial Intelligence for Development in Africa (AI4D) initiative. This partnership, with an investment of CAD 20 million over four years, aims to support African-led research on using AI to meet local needs. AI4D, in partnership with the Human Sciences Research Council ZA in South Africa, also supports the African Observatory on Responsible AI (AORAI). In addition, AI4D works with the African Union Development Agency to develop a model African AI policy. The Observatory aims to position the African continent in global debates and policy making on responsible AI.

Stakeholders in developing countries could also consider formulating research questions relevant to local priorities and amenable to analysis using AI. The 100 Questions Initiative, launched by the GovLab, could provide inspiration (The 100 Questions, n.d.). This initiative seeks to map the world’s 100 most pressing, high-impact questions that could be addressed if relevant datasets were available.

The selection of such questions could be informed by a dialogue between civil society, the private and public sectors, and academic and research institutions. Knowing priority questions could lead to new forms of data collaboration with the private sector to help advance the necessary science. For example, in the quest for stakeholders in Bangladesh to analyse and respond to climate extremes, a leading telecommunications provider, Grameenphone, shared its anonymous mobile call data records with three partners. Grameenphone, the United Nations University Institute for Environment and Human Security, the International Centre for Climate Change and Development, and the Telenor Group examined population movements before and after cyclone Mahasen struck Bangladesh in May 2013, an extreme climate event that affected 1.3 million people. Private-public collaborations can also stimulate investments in data infrastructures and open data sharing essential for using AI in science.

This essay highlights the low general level of readiness for use of AI in developing countries. It also considers how bilateral co-operation can contribute to improving the productivity of developing-country science, in particular through greater use of AI. Both bilateral and multilateral development co-operation could strengthen science in the developing world, broaden global research agendas, orient uses of AI-enabled science towards problems of particular concern to poor countries and ultimately assist global efforts to achieve the Sustainable Development Goals.


AFD (n.d.), “IA-Biodiv Challenge: Research in Artificial Intelligence in the Field of Diversity”, webpage, (accessed 6 January 2023).

Addo, P.M. et al. (2021), “Emerging uses of technology for development: A new intelligence paradigm”, AFD Policy Papers, No. 6, March, Agence Française de Développement, Paris,

Global Partnership for Sustainable Development Data (2018), Africa Regional Data Cube Initiative website (accessed 6 January 2023).

Hao, K. (2021), “The race to understand the exhilarating, dangerous world of language AI”, 20 May, MIT Technology Review,

Oxford Insights (2022), Government AI Readiness Index 2021, Oxford Insights, Malvern,

The 100 Questions (n.d.), The 100 Questions website, (accessed 6 January 2023).

Verhulst, S. et al. (2022), “Building data infrastructure in development contexts: Lessons from the #Data4COVID19 Africa Challenge”, A Question of Development, No. 56, March, Agence Française de Développement, Paris,

Zhang, D. et al. (2021), The AI Index 2021 Annual Report, AI Index Steering Committee, Human-Centred AI Institute, Stanford University, Stanford,


← 1. Currently, more than 500 researchers around the world are working together, under the BigScience project led by Huggingface, to learn more about the capabilities and limitations of large multilingual language models (Hao, 2021).

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2023

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at