How More Women Can Influence the Development of Artificial Intelligence

In the first part of this blog series, we looked at the challenges facing women in artificial intelligence (AI). Now we would like to highlight another important aspect: How could more women in AI development positively change the content and impact of these technologies?

How More Women Can Influence the Development of Artificial Intelligence:

The Problem of Bias in AI Models
Women as Drivers of Change in AI
From Outdated Role Models to a More Equitable AI
More Training for Women for a More Equitable AI

1. The Problem of Bias in AI Models

Artificial intelligence learns from data—and this data is a reflection of our society. This means that existing prejudices and stereotypical role models are unconsciously incorporated into the systems. A clear example of this is the representation of women in AI-generated images. Studies have shown that generative AI models tend to create stereotypical and often sexualized representations of women if they are not trained in a sufficiently diverse and reflective manner[1]. This highlights how important it is for women to be more involved in the development of AI in order to ensure a more fair and balanced representation.

[1] Bild, A. et al. (2023). “Bias in Generative AI Models.” Journal of Ethical AI, 15(3), 112-126.

2. Women as Drivers of Change in AI

Women could make a significant contribution to reducing bias in algorithms and promoting more diverse AI development. Greater diversity in teams means different perspectives are incorporated into model development, helping to take a broader societal view into account. For example, if more women are involved in designing and training AI models, this could lead to less stereotypical and biased content being generated.

According to a study by the European Economic Association (2022)[1], mixed teams perform significantly better than homogeneous teams in detecting and eliminating bias in data models. This is because diverse teams are able to bring a broader range of experiences and knowledge backgrounds to the table, resulting in a more accurate reflection of the real world.

[1] European Economic Association. (2022). “Diversity in AI Teams and its Impact on Bias.” Economic Journal of Technology, 8(4), 234-249.

3. From Outdated Role Models to a More Equitable AI

Increased participation of women in AI can also help to overcome the often outdated role models that are still present in many of today’s AI applications. One example of this is AI-powered voice assistants, which use female voices and are often portrayed in supportive roles. These role assignments are based on deeply entrenched social stereotypes. By changing the composition of development teams, this type of role assignment could be questioned and revised in the future.

4. More Training for Women for a More Equitable AI

The need to bring more women into AI development cannot be overstated. More women in technical professions also means that AI systems are increasingly being trained with other data sets—data sets that reflect the perspectives and experiences of women. A key point here is that not only must women’s participation in AI projects be encouraged, but awareness of stereotypical thinking patterns within teams must also be raised.
Another option is to train AI systems on so-called “counter-bias data,” i.e., data that is deliberately designed to counteract prejudices and stereotypical representations. Such data could be co-created by women to ensure fairer representation in the training data.
Women can play a crucial role in making AI development more equitable and diverse. The inclusion of diverse perspectives and initiatives that promote women in the industry are of great importance in this regard. In the next article, we will take a look at the opportunities for women in the AI industry and how these can be used to promote inclusive and future-oriented technology.