This article is more than 2 years old Just nine out of 116 AI professionals in key films are women, study finds This article is more than 2 years old Report says pattern seen in films such as Ex Machina risks contributing to lack of women in tech

When data ignores women Metropolis, Her, Ex Machina: Women have been appearing as Artificial Intelligence in films for a long time. However, AI developers in films are almost always male. Just 9 out of 116 of them are female. And even when they are women, they are often in the service of men. In reality, the situation is only marginally better: Only 22 percent of AI experts are women − and the most powerful positions are reserved for men. AI systems are often given a female voice because this makes them sound more helpful and friendly. On the other hand, female voices are more frequently not understood by AI than male voices. Why is that? The answer is hidden in the data used to train the systems. Today, on International Women’s Day, we explain how data discriminates against women and show how we can do better. More gaps in data − more disadvantages for everyone Nowadays, data is used in almost all areas of life, from urban planning to medicine. Many medical studies are based on data from male test subjects, which means that symptoms specific to women, for example before a heart attack, are often not recognized correctly. Data is therefore usually collected from a male perspective, which ignores the reality of women. When it comes to snow clearing services, urban planning is often orientated towards the commuting routes of average male employees. In Sweden, this has led to women suffering more accidents than men in winter, as they are more likely to use footpaths and cycle paths that are only cleared after the main roads. Women tend to travel on less direct routes, partly because they spend 44.3 percent more time each day on unpaid care work, such as voluntary work and childcare. This everyday life of women is hardly taken into account in urban planning. So, how snow is cleared away on public paths can be sexist if women are disadvantaged and exposed to dangers as a result. What does all this have to do with AI? Without data, there is no Artificial Intelligence. If this data omits certain areas or contains discriminatory features, the respective AI system “internalizes” such patterns. AI can only be as good as its training data. Data gaps, whether in the analog world of the previous examples or automated decision-making systems, can have serious consequences for previously insufficiently considered population groups. This was certainly the case with “Försäkringskassan”, the Swedish Social Insurance Agency. An investigation by Lighthouse Reports in 2024 found that an opaque AI system disproportionately estimated the risk of certain groups − including women, people with a migrant background, and low-income earners − as higher for committing benefit fraud. The authority had been using the system since 2013, which has since led to unjustified fraud investigations. Many of those affected had to wait a long time for their social security benefits. This often jeopardized their livelihood. Discriminatory machines? Not with us! AI is increasingly making important decisions − but this often leads to women and marginalized groups being disadvantaged. With your AlgorithmWatch membership, you support important research and advocacy work on the topic of AI. Regulate AI – protect human rights! We must actively fight against discrimination! Discriminatory, wrong decisions by AI systems do not disappear by themselves, as the systems are often not transparent and access to them is blocked. The Swedish Social Insurance Agency, for example, repeatedly rejected requests for access under the Freedom of Information Act. As a result, discrimination against many people in Sweden remained in the dark for a long time. So, what can we do to make data sets better? One answer is data feminism. With gender-specific data, discriminatory patterns can be recognized and removed. AI made by ALL: Why we need more women in AI development Last but not least, something needs to be done about the low representation of women in AI development. More diverse perspectives in AI development would better meet the needs of more diverse populations. In Europe, the gender imbalance in working life has been eliminated by 75 percent − with Sweden and Germany leading the way. In the AI sector, however, the picture is very different: At around 20 percent, the two countries have one of the lowest numbers of female AI professionals in the EU. And what is AlgorithmWatch doing? We shed light on data gaps in AI systems and research cases of discrimination caused by automated decisions. We are committed to ensuring that there is no algorithmic discrimination in the labor market, in education, in administration, or elsewhere. When AI systems influence decisions about essential areas of our daily lives, there must be full transparency about their use. If you would like to find out more about our work on discrimination and AI, you can read all our articles and positions on our website here.”