top of page
  • Writer's pictureShamira Ahmed

An African perspective on gender and AI needs African data and research



Big data is a key element of Artificial Intelligence (AI) and Machine Learning (ML) but the effective use of these disruptive technologies simultaneously depends on the scope and quality of the available data used to train and test algorithms. Algorithms learn from real-world data and historically most traditional data sets are exclusionary and have left out women and other marginalised groups. Thus without careful consideration of the data that informs algorithms, AI can potentially perpetuate existing or even create new unconscious social biases.


Research on the gender dimensions of AI and other frontier technologies in Africa is still in its infancy. Compared to other world regions, Africa has little or no gender-disaggregated data on digital access, skills and participation in the digital economy. There is also limited quantitative and qualitative research on related issues such as: barriers to gender digital equality, the impact of social structures on women’s career paths in innovation, science, technology, engineering, arts, and mathematics, and the effect of intersectional demographic characteristics on people’s experience of technology.


Without representative data and research to drive evidence based policies, Africa cannot realise the full benefits of Fourth Industrial Revolution (4IR) technologies as envisaged in the African Union’s Digital Transformation Strategy (DTS). The implications of these data deserts are becoming more dire as African countries are encouraged to swiftly attempt towards transitioning to 4IR technologies, such as “AI for national development” and future post-Pandemic resilience, without consideration of the unintended consequences and associated risks. Given their immense appetite for and dependence on big data for decision-making, AI systems have the potential to exacerbate existing gender inequalities or even create new disparities if the knowledge that informs AI design is not representative of African realities.

Although many private sector companies on the Continent already analyse big data to improve their services and products as well as understand their customers in real time, from a developmental perspective, African countries lack gender disggregated data on two fronts — (i)the availability of official national statistics to provide open source data and enhance public data value creation , and (ii) the availability of quantitative and qualitative primary research on gender issues in relation to AI.


On the research front, western literature has covered a wide range of topics demonstrating the reality and potential of AI for gender justice from historical and contemporary perspectives. This includes research on the potential for AI to enable the transcending of gender identities and mitigate gender discrimination. Very little of this type of research and reflection is emerging from African scholarship. Much of the available literature consists of commentaries on the potential of AI for national development in general. Those that discuss pertinent issues, such as data bias and other risks, invariably draw on studies from the global North — failing to fully capture the manner in which badly designed AI could exacerbate discrimination and exclusion in the context of developing countries with weak digital rights or inefficient institutional remedies .


While some of the research conclusions from Europe, North America and Asia might be applicable in Africa, the configurations of gender and power in Africa are not necessarily the same. The issues may have different dimensions resulting from specific national contexts. Relying solely on non-African gender research to inform local policy could lead to erroneous decisions or misplaced priorities and actions.

Beyond the privacy issues, capacity limitations, and the digital data divide that pose challenges for big data collection in Africa — If new technologies such as AI are not developed and applied in a context-responsive way, they are likely to reproduce and reinforce existing gender stereotypes and discriminatory social norms.

While improving availability of data and increasing research alone are not the answer, they are important components for gender-responsive evidence based policymaking.


To generate African perspectives on gender and AI, public sector decision makers should:


(i) Increase funding for national statistics offices (NSO’s) and other data stewards to develop strong, modern national statistical systems with capacity to deliver, use, and share better data for longer-term prosperity and sustainable development.


(ii) Enhance private-public partnerships to build bridges and catalyze open data ecosystems that leverage new digital data sources and enhance public data value creation into institutional practice for the public good.


(iii) Update capacity of NSOs to include collecting gender-disaggregated data on emerging technological trends, including at a sub-national level. Sensitise other relevant agencies (e.g. Science and Technology, Communications and Education Ministries) on the importance of mainstreaming gender into policies.


(iii) Support local initiatives that supplement supply-side data from mobile service providers with open source demand-side data directly from consumers.

(iv)Incentivise and support rigorous quantitative and qualitative primary research on gender, technology and society (by academics, non-profits, think tanks, etc.).


This summary from a policy brief was first published here in 2020.

30 views0 comments

Commentaires


bottom of page