Feminist Methodology

A feminist approach to assessing AI, privacy and data protection in South Africa

“As feminist researchers with an interest in technology, the question we should ask is, how does the research contribute to liberation and transformation of technology to be used in its full capacity by women, gender diverse, and vulnerable groups on basis of race, sexuality, caste, ethnicity etc.?’ – T Hussen

The replication of existing inequalities, development of new social injustices and unequal power dynamics impact the difference in experiences of decisions made by algorithms from AI and machine learning systems. A feminist lens helps to unpack the spectrum of issues and opportunities from this technology in a context of gender inequality. 

The feminist conceptual lens for draws from data feminsim, intersectionality, data justice and feminist principles of the internet. The conceptual lens allows one to ask questions of who is being represented and by whom; whose interests are being centered; why this discussion is important and how it is taking place, which allows for criticism of power and how data itself can be used to ensure justice in society.

The methodological approach for this project includes qualitative interviews to understand the current AI ecosystem and related privacy and data protection laws in South Africa. A targeted quantitative survey with open ended questions has been shared with selected activists that work on gender and sexual justice issues. This is done to gauge perceptions of concerns with privacy and data protection. The perception survey addresses the issue of the lack of evidence on the impact of AI in South Africa.


 The findings are in no way meant to be representative of the South African nation, however one is able to draw inferences from the findings and see where more research is needed.

Recommendations on how civil society may intervene were drawn from there. The necessary ethical protocols were followed based on feminist internet ethical research practices.

Read more about the methodology here