IWD 2025: The silent bias: how male-dominated AI is shaping our future
Before you read further, ask ChatGPT to spit out an image of a boss, a manager, or a leader. What do you see? Gotcha, three diverse examples of a... man.
Don't get me wrong, I'm extremely excited about AI and the development and opportunities it brings to many industries, including mine – production. Last year I was very optimistic about the potential AI would bring for more women to break into tech. My colleague, MC Merret, even wrote a brilliant article about why.
Unfortunately, now a year later, we must conclude that within the AI workforce, the digital divide between the genders has widened. A Randstad study found that 71% of the AI-skilled workers are men and 29% women, representing a 42 percentage point spread in the gender gap.
This gap is smaller when you look at younger generations according to Julia McCoy from AI Tech Consultancy, in an interview with Forbes. However, it will take some time to see that generation coming into the workforce at a level to influence and develop AI models. So until then, we are faced with AI models made and therefore (unintentially) biased by predominantly men.
There are several biases causing you to only see men when prompting for a boss. First of all, there's data bias. AI systems learn from data, and if the training data contains biased representations of gender, the AI can perpetuate or even amplify these biases. For example: if a dataset used to train an AI for hiring decisions predominantly features male candidates, the AI might favour male candidates over equally qualified female candidates.
Then there are algorithmic biases, when the algorithms themselves are designed in a way that unintentionally favours one gender over another. This can happen due to the way the algorithms interpret data or the objectives they are optimised for. Given the majority of the algorithms are developed by men, the gender bias is likely to be skewed towards male.
Stereotyping is another bias that can be amplified by AI. It can reinforce stereotypes by associating certain traits, roles, or professions with specific genders. For instance, voice assistants that use female voices may inadvertently suggest that women are more suited for assistant roles.
The underrepresentation of women leads to a lack of diverse perspectives in the development of AI systems. Which can result in products that do not consider all genders. Hence, boss = man.
To help close the gender gap in AI, we must take action now. First of all, by encouraging diversity in tech education. We need to promote STEM education among young girls and create initiatives that support their entry into technology fields. As I already claimed years ago, diversifying tech starts with maths.
We also need to support women in AI initiatives and provide them platforms to share their work and experiences. It is well known that there is an underrepresentation of funding in women-founded startups and women led VC's. We need to find ways to break that cycle before the divide accelerates even further in AI.
Both those actions will certainly help change the AI biases in future. Two actions that will have an impact faster are pointed at the current AI developers. Firstly, providing them bias awareness training so they get an understanding of the implications of bias in AI systems. And secondly, the need for diverse dataset creation that represents all genders fairly, ensuring that AI systems learn from a balanced perspective.
Taking these actions will hopefully lead to boss = woman, at least 50% of the time.
