The Gender Problem in AI

By Sonwabise Mzinyathi, Acting Chair: SA Women in ICT Forum

“Will those who write the algorithms ever realize their negativity bias?” – Elon Musk, July 2020

Identifying criminal behaviour, treating cancer, preparing for natural disaster, tailored education, marketing – these are but a few ways in which Artificial Intelligence is used to advance society. It is a technology that simulates human intelligence but how do we ensure that in the process, it doesn’t simulate human prejudices and isms. Who’s agenda and gender does it seek to simulate.

Element.ai sampled researchers from three AI conferences in machine learning and machine vision in 2017 and found that across 23 countries, 88% of the researchers were male. This is hardly surprising considering that the computer science and overall tech space is dominated by males across Africa and globally. This is not assisted by the fact that studies have shown that men are promoted for their potential, while women are promoted for their achievements in the sector. In the long run, this only means that the gap will be a perpetual challenge for the industry, impacting design of technology and the gender stereotypes it will continue to promote. You see, behind all algorithms and software development are humans, if stats show that these humans are mostly male, then we face potential bias even in the very technology whose intention it is to eliminate biases.

For example, 74% of AI movie characters are male. While female avatars are most commonly used to play virtual assistants and champions, perpetuating the perspective that helping roles are best left to women. Compare AI gender stereotypes in science fiction – male androids being designed to be heroes and popular villains such as The Terminator, Nero while as females androids are either sexualised or overly submissive such as the female androids in Ex-Machina and Metropolis to name a few.

Datasets from which predictive models are developed, the software and the hardware, design and testing in the industry is mostly done by men naturally operating from their own experiences. Both men and women, people of diverse backgrounds need to be the developers of innovations from design, systems, datasets and evaluations – the entire value chain. The principle of using technology to bridge the inequality gap, seeks to eliminate discrimination, minimise data biases and promote shared benefits and being inclusive and bringing diverse people, including women and people living with disabilities, into the AI value chain is part of it. 

The World Economic Forum lists the importance of conscientious work within the AI research community in order to recruit and promote more female talent. I agree. There also definitely lies a need to find technical solutions for fair and accountable AI. In addition, investors need to support more female founders. We also need to create more images of AI and robots free of gender stereotypes in the media, entertainment and creative sectors.