In a world increasingly driven by artificial intelligence (AI), experts are calling for a more nuanced, culturally-sensitive approach to AI development and deployment in Africa. Speaking at the Second Regional Conference on Countering Disinformation and Hate Speech in West Africa, Philip Ayazika from Uganda highlighted the need for an Afro-feminist perspective in AI development.
The Explainer reports, the conference organized by the Centre for Democracy and Development (CDD-West Africa) in collaboration with the National Democratic Institute (NDI) and funded by Global Affairs Canada, is taking place from July 9th to 11th, 2024, in Lagos, Nigeria.
Ayazika emphasized the inherently exploitative nature of the data economy driving AI, pointing out the entrenched power asymmetries controlled by powerful entities.
He stated, “The data economy driving AI is inherently exploitative, structured around entrenched power asymmetries controlled by powerful entities,”
“Individuals’ digital lives are transformed into computable data for capital gain, systematically depriving users of control in a multidimensional and intersectional manner.”
He highlighted how big tech companies from the Global North often hinder African countries’ attempts to develop and deploy their own AI technologies. As an example, he cited South Africa, where “Google and Facebook’s capture of advertising revenue has severely impaired traditional media companies’ ability to compete.”
A critical aspect of Afro-feminist AI, according to Ayazika, is making visible the often-hidden workforce behind AI systems.
“The idea of perfect, effortless automation hides the reality that many ‘invisible’ workers, often from the Global South, do short-term, informal jobs like labeling data,” he explained.
Ayazika also stressed the importance of maintaining individual and community autonomy in the face of information capitalism.
“To ensure that individuals and communities maintain their autonomy, we need to see the data used in AI as an extension of their ability to make decisions and control their own lives,” he said.
Grounding AI in Afro-feminism
Ayazika outlined four key principles for grounding AI in Afro-feminism:
1. Balance Data Rights: Respecting both individual and collective data rights.
2. Address Exploitation: Redistributing power, control, and benefits over data equitably.
3. Make Workers Visible: Protecting the rights of microworkers and ensuring fair compensation.
4. Support Autonomy: Treating data as an extension of individual and community decision-making power.
The speaker presented a detailed framework for regulating AI, emphasizing the need to consider factors such as agency, equality, dignity, non-discrimination, safety, security, human impact, accountability, oversight, redress, privacy, and data protection.
Ayazika illustrated these concepts using a hypothetical AI tool called NewsAI, demonstrating how regulatory thresholds could be applied to real-world scenarios. He stressed the importance of human oversight, particularly in high-stakes applications of AI.
Ayazika emphasized the importance of participation and inclusivity in AI procurement and development. He called for gender impact assessments and highlighted the need for transparency in AI capabilities, from marketing to licensing.
“Even the most useful models are susceptible to bad actors,” Ayazika warned, using the example of a hypothetical non-profit program that could potentially misuse refugee data. He urged stakeholders to question how the full extent of philanthropic interventions can be understood and how potential harm could be mitigated.
As Africa continues to embrace AI technologies, Ayazika’s call for an Afro-feminist approach offers a roadmap for ethical, inclusive, and culturally-sensitive AI development.
By addressing issues of exploitation, visibility, and autonomy, this approach could help ensure that AI benefits all members of African societies, rather than exacerbating existing inequalities