Written by Erica Farmer, AI and Future Skills Speaker, Trainer and Author
When conversations about AI bias come up, there is a tendency to place them into the diversity and inclusion box, because we hear things like;.
“It’s a women’s issue.”
But AI bias is an ‘everyone problem’ because the systems increasingly shaping our decisions, opportunities, careers and learning experiences are being trained on data created by humans. And humans, whether we like it or not, are biased.
AI did not invent bias; it inherited it. It was trained in male dominated environments, by male dominated teams. There’s no question of that.
The question is whether we are willing to do something about it.
The Data Problem We Don’t Talk About Enough
Large Language Models (LLMs) learn from vast amounts of human-generated content and countless digital interactions to understand patterns and predict responses. The challenge is that the internet does not represent humanity equally.
Historically, men have held greater visibility in many sectors, particularly technology and science. As a result of this, the data used to train AI systems often reflects a world that is unbalanced.
When AI learns from biased data, it can amplify it. This means it can reinforce stereotypes, overlook perspectives and replicate inequalities at scale. Just ask ChatGPT for an image of a typical boardroom and representatives. It more diverse now than it was 12 months ago but still has a way to go!
This matters because AI is becoming our digital colleague. It is helping us learn, recruit, communicate and make decisions, which is all good practice when done consciously and with care. However, if the underlying intelligence is skewed, the outputs will be too.
And there’s the problem.
The AI Adoption Gap Creates Another Risk
Harvard Business School found that women are adopting AI tools at approximately 25% lower rates than men across multiple studies and countries. Similar research places the gap somewhere between 20% and 30%, depending on the study. And men are far more likely to employ reward seeking behaviours with using new technology, such as seeking praise, which amplifies the visibility of their practice and reduces that of women, who are far less likely to demonstrate those behaviours.
At first glance, that might sound like a women’s issue.
It isn’t and let me tell you why.
If half of the population is underrepresented in both the development and use of AI, then AI itself becomes less representative. This is an issue because the people who use technology shape it, the people who train it influence it and the people who challenge it improve it. So, if fewer women are actively engaging with AI, we lose valuable perspectives that could help identify blind spots, challenge assumptions and create better outcomes for everyone.
Why This Matters to Business
Many organisations are investing heavily in AI capability by rolling out tools, running training programmes and launching adoption campaigns. Yet very few are stopping and asking a critical question:
Who is actually using it?
If adoption rates differ significantly across different groups, then the benefits of AI may not be distributed equally. This means some employees gain confidence, skills and visibility and others fall behind.
This becomes even more of a societal issue when some people get to shape the future, and others have the future shaped for them. This is the point that should be keeping us up at night as we can’t repeat the gender focussed mistakes of the past.
We All Need to Do Better
The solution is not to ask women, ‘just to use AI more’ and it’s certainly not to ask women to fix AI bias. It is for all of us to take responsibility. But what does that look like? From the top to the front line, I see these are 4 steps to take right now;
1) Leaders in organisations need to actively monitor who is engaging with AI and who is not and communicate this openly.
2) Educators need to ensure examples, case studies and learning experiences reflect diverse perspectives.
3) Tech companies need to improve datasets, test outputs and identify bias.
4) And individuals need to challenge what AI produces.
If AI is key to shaping the future of work, then that future needs to reflect all of us because AI bias is not a women’s problem, it is a humanity problem.
