Responsible, ethical and feminist AI for development: Challenges and solutions

Members of East Africa’s artificial intelligence (AI) community say there is an urgent need for ethical considerations, feminist approaches and inclusive African policies and strategies in the development and implementation of AI on the continent. Those are just some of the recommendations that came out of a gender café on "Responsible, Ethical and Feminist AI for Development," co-hosted by Global Affairs Canada and IDRC in Nairobi in March 2024. At the event, a group of Nairobi-based experts, practitioners, policymakers and researchers, many of whom are supported by IDRC, delved into how AI technologies hold the promise of advancement and innovation but also pose significant challenges and risks, especially regarding gender disparities and inclusion.

Impact of AI on women’s employment 

The International Labour Organization estimates that AI could displace 3.7% of all female employment globally, or 48 million jobs. That’s compared to a loss of 1.4% for males. IDRC-supported research by Strathmore University’s Centre for Intellectual Property and Information Technology Law (CIPIT) on Kenya’s banking industry found that AI would first take over jobs in such areas as customer service and human resources, which are predominantly held by women. The research project also studied people’s attitudes to the possible changes. “There was a lot of fear from the respondents on AI deployment and what it meant for their jobs. But on the side of the employer, it [the reaction] was a positive thing because [they thought] I'm going to use AI and everything is going to be great,” a participant involved in the research noted.

Café participants also discussed how the data feeding AI systems can reinforce gender biases. For instance, when people reinforce negative stereotypes (e.g., "all women love money") in social media, that digital information is integrated into AI's learning process, affecting how AI systems understand gender, which then go on to perpetuate and even normalize stereotypes and biases.

Inclusion in AI development: How are we addressing issues of power relations?  

The lack of representation of women and marginalized groups in AI's development and decision-making processes was identified as a critical issue. AI and tech industries are predominantly led by middle-class males. This hierarchy leaves marginalized groups, especially African women, with limited influence and access to AI development.

Another challenge raised was the dominance of 'colonial models' of AI, which are not designed to understand or incorporate indigenous African models or perspectives. A socio-technical systems and AI ethics expert pointed out that true inclusion goes beyond mere numerical representation. “Ideally, we should be asking ourselves, how are we addressing issues of power relations? Who has the money? Who has expertise? Who is designing what? Who is shaping those perspectives and whose political and corporate interests are we driving?” she asked.

“You do not have to be an expert to talk about AI, because if that's the case, then the global majority will not be able to have a seat at the table because by design, the right seat or the seat for technologies is [San] Francisco, is China or Dublin,” a participant noted. She said that the democratization of AI discussions ensures that a wider range of perspectives, especially those from underrepresented groups, can influence the development and application of AI, making it more equitable and reflective of global diversity.

AI policies and strategies in Africa: Putting Africans first  

Countries like Kenya, Egypt, Mauritius, Rwanda and Senegal are developing AI policies in growing recognition of its importance across the continent. The African Union is also developing an AI continental policy roadmap, with support from IDRC, among other development partners. Participants welcomed strategy development, but underlined that it should be driven by a clear understanding of the specific objectives, challenges and opportunities AI presents in the African context.

One of the speakers warned about the potential of the ‘Brussels effect’ where EU regulations influence global norms due to the requirements of doing business with the EU. She noted that African countries are often under pressure to align with global regulations, such as the General Data Protection Regulation, that do not necessarily fit Africa’s communal considerations nor recognize the nuances of marginalized communities. This underscores the importance of tailoring AI policies to local realities.

She added that technology deployment, including AI, is inherently political, with the potential to reflect and reinforce the political will of those in power. Therefore, shaping AI policies and strategies in Africa requires careful consideration of the political landscape to ensure these technologies benefit society as a whole.

In Kenya for example, existing laws scarcely address AI, with little to no mention of gender considerations. The rejection of a draft robotics and AI bill by AI experts due to the lack of women's representation points to a significant gap in ensuring gender-inclusive AI policies. Additionally, the absence of AI in educational curricula and the lack of a comprehensive AI policy signal areas needing urgent attention. 

Speakers also raised the importance of inclusive global conversations in developing AI strategies, stressing the need for voices from the Global South, particularly Africa, to be heard.  

Here are some recommendations that emanated from the discussion: 

• foster inclusive policymaking by ensuring AI policymaking processes involve diverse stakeholders from various economic, social and cultural backgrounds. For global policies, special attention should be given to including voices from the Global South and marginalized communities to ensure AI strategies are equitable and reflective of diverse needs. 
• address biases in AI through platforms and forums to specifically address and mitigate social biases in AI development and usage. This includes gender biases, ensuring that AI technologies do not perpetuate existing inequalities but instead contribute to gender equality. 
• revise and update legal frameworks, especially in countries like Kenya, where existing laws do not reflect the current technological landscape.  
• enhance AI education and awareness by integrating AI into educational curricula at all levels to build a foundational understanding of AI among students.  
• promote local AI research and development initiatives as the best way to ensure that AI solutions are tailored to meet the specific needs of the population.  
• actively work towards gender inclusivity in AI development teams, policymaking forums and research initiatives. 

This article is based on a gender café co-hosted by IDRC's Eastern and Southern Africa Regional Office and the Canadian High Commission in Nairobi on February 15, 2024, with contributions from panelists Angella Ndaka (socio-technical systems and AI ethics expert), Nanjira Sambuli (technology, public policy and global governance expert), Melissa Omino (intellectual property expert), Irene Mwendwa (human rights and tech policy expert) and other participants.

Contributors: Loise Ochanda, program officer, IDRC, with Elizabeth Muriithi, regional advisor, IDRC, and Betsy Muriithi, Professional Development Award recipient, IDRC, and Kagure Wakaba (gender specialist).