Afro-feminist AI governance; challenges and lessons

Event report

The existence of subsisting and emerging technologies and their accompanying challenges to human rights and inclusion require in-depth, critical analyses. Understanding these risks and their consequences for people has spurred efforts to develop and build frameworks shaping the creation of AI globally. Nonetheless, these efforts have been led and shaped by values and stakeholders from the global north, with the exclusion of the Global South. 

Adequately regulating the use of emerging technologies in Africa requires the development of Afro-centric frameworks involving African stakeholders with an understanding of the relevant issues. A significant part of this includes developing an Afro-feminist AI governance framework, taking advantage of past and growing experience and research. The framework will address technologists, members of civil society, and policymakers, highlighting some of the trends in the growing development of AI in Africa, the gaps in other frameworks, and novel insights to note in AI development. An early iteration of the Afro-feminist AI governance framework has been presented at the DataFest in July 2022. 

Various frameworks address trustworthy AI, the OECD principles, for instance, or the African Union Data Policy Framework. Both mention a need to create human-centred AI based on human-centred values and fairness. But what does human-centred AI mean? Do these frameworks consider different views and perspectives of under-represented groups such as African women? Research shows that the answer is, in most cases, ‘No’, and the harm inflicted on women and people with disabilities is disproportionally high. 

The reasons for these challenges vary and evolve around:

• Lack of Data: Many countries in Africa face the problem of data collection challenges due the lack of digitalised data and limitations to internet access.

• Access to digital devices: Vulnerable groups such as women often do not have the means to access the internet or make their voices heard. Families often share one smartphone or digital device, mainly used by men. 

• Digital illiteracy: Policymakers in designing AI frameworks and regulations often lack the necessary knowledge of the function of AI. On the other hand, users of digital services often lack knowledge of their digital rights.

• Inequities in the data economy: Many policy documents across countries have fashioned data as an asset. This places different sets of actors in different fashions, some in unfavourable positions. Currently, the way in which data exists and is shared is largely controlled by data processes and data controllers based on non-negotiable, one-sided contracts.
• Enforcement Problem: Countries face a lack of capacity in enforcing data protection rules or AI framework policies. 

To address these problems, actors have started initiatives such as the GIZ Handbook for Implementing a Capacity Building Programme for Policy Makers, the Smart Africa project, and the ICTworks works platform.

Significant recommendations were: 

• Fund data-training and digital literacy initiatives: Funders, partners, and technology companies should focus their efforts and funds on supporting and developing data training initiatives on the continent. 

• Fund Afro-Feminist AI research: Feminist movements should be funded to conduct research from an inclusive, feminist lens. 

• Develop independent and intersectional data collaboratives: A decentralised approach to data collection could address the challenges. Governments and other stakeholders should consider setting up a number of independent data centres under different models of data stewardship that are accountable to citizens.

• Address women’s safety online: Social media platforms should emphasise protecting women and should engage with content moderators who understand the nuances and contexts of local cultures and languages. 

Strategic Foresight: Use strategic foresight when exploring the future and new technological developments. This allows policy makers to prepare for future scenarios and include different views from all relevant stakeholders.

The session in keywords