Gender gap and AI
Despite the transformative potential of AI, women are still underrepresented in its development, leading to prejudice and reinforcing inequality. Addressing this gap is essential to building fair, ethical and inclusive AI systems.
Currently, only 22% of tech jobs in the EU are held by women - a figure that drops even further when you look at the path from education to employment. You can view the relevant statistics here. In Germany, the proportion of female AI professionals is around 20%, indicating a significant gender imbalance in this field.
The impact of AI on gender equality
AI can open up new opportunities. The use of AI goes beyond traditional STEM (Science, Technology, Engineering, Mathematics) roles and creates opportunities for women from diverse backgrounds. You can read the full report here. However, AI is more likely to change women’s jobs than men’s. According to a UN report, about 1 in 10 jobs usually done by women could be changed by AI – for men, it’s only about 1 in 30. This is because AI is taking over more and more office work, like secretarial and administrative tasks, which are often done by women.
Gender bias in AI
- In AI systems, we see that there is a gender bias against women. For example, Amazon 's system taught itself to prefer male applicants. It discriminated against CVs that contained the word "women", such as "captain of the women's chess club". It also downgraded graduates from two women-only colleges, according to people familiar with the matter.
UNESCO's report on gender bias in AI states:
- "AI models associate male names with career, business and leadership (e.g. 'executive', 'salary').
- Female names are associated with domestic roles (e.g. "home", "family", "children").
- Women were more often associated with occupations such as nurse, teacher or domestic worker, while men were associated with doctor, engineer or CEO.
- Dr. Joy Buolamwini's research on bias in AI systems has been instrumental in advancing gender equality in engineering and technology. Her research found that AI-powered facial recognition systems had an error rate of 34.7% in identifying dark-skinned women, compared to just 0.8% for light-skinned men.
Gender bias is not limited to these examples. For example, if you ask an AI to generate a picture of a CEO, most models will create male representations. AI voice assistants also use female voices by default (e.g. Siri, Alexa).
Translate bias
Even in languages with gender-specific nouns, Google Translate or DeepL often reinforce traditional gender roles. For example, "The doctor is intelligent. The nurse is caring." is translated as "El doctor es inteligente. La enfermera es cariñosa." is translated as "The doctor is intelligent. "Doctor" is rendered as masculine and "nurse" as feminine. The same can also be observed in German.
What can be done? - Promoting gender equality in AI
We at Pforzheim University are expressly committed to gender equality in the field of AI. We are convinced that while gender bias in AI is a major challenge, measures can be taken to create fairer and more inclusive systems.
Five steps to more inclusive AI systems
Villar, who works with UN Women, suggests the world take five steps towards a better AI sphere:
- Utilize diverse and representative data sets to train AI systems
- Improving the transparency of algorithms in AI systems
- Ensure that AI development and research teams are diverse and inclusive to avoid blind spots
- Introducing strong ethical frameworks for AI systems
- Integrating gender-equitable guidelines into the development of AI systems
Below you will find events and courses specifically for women in AI
