Sustainability. We mean it.


AI, other emerging technologies, and gender bias

EOS Insight
6 May 2020 |
In the first of our two-part Q&A on AI and gender bias, Katie Frame, an engager at EOS at Federated Hermes, talks to Saniye Gülser Corat, director for gender equality at UNESCO, about why we should be concerned by the lack of diversity in the tech sector.

In multiple sectors, studies have shown that there is a positive correlation between higher levels of employee diversity and stronger financial performance, and the technology sector is no different. At the same time, there is evidence to suggest that the underrepresentation of women in the technical and developer workforce reinforces gender inequality, as biases are unintentionally built into AI algorithms. UNESCO’s Division for Gender Equality highlighted this in its publication “I’d Blush if I Could: closing gender divides in digital skills through education”, which explored the digital gender gap, the gender equality paradox and the issues with gendered AI.

Katie Frame, EOS: What is the digital gender gap and why does closing it matter?

Saniye Gülser Corat, UNESCO: The digital gender gap refers to the differences between men and women, or girls and boys, in their ability to access and use digital technologies and participate fully in the online world.

In recent years, due to the rapidly declining price of connectivity and hardware, skills deficits have eclipsed barriers of access as the primary contributor to the digital gender divide.

It is important to close these gaps because increasingly digital skills are essential to manage our lives, generate income, participate in society – even to stay safe and healthy – so the gap puts women at risk of being left behind in all realms of economic, political and social life. Additionally, technologies generated by male-dominated teams and companies often unintentionally reflect troubling gender biases. Women make up over half the world’s population, so this is not only worrying for women themselves but has dire consequences for economic growth and societal well-being as a whole.

Katie Frame, EOS: You raise a worrying point about gender biases being reinforced in technology, and in UNESCO’s research you talk about AI-powered digital voice assistants. Are AI-powered digital voice assistants sexist?

Saniye Gülser Corat, UNESCOAI-powered digital voice assistants, such as Siri, reflect our society, but they also help shape what society looks like and the values we hold. Gender biases are reflected in the names of AI voice assistant technology. Microsoft’s Cortana, for example, is based on the name of a synthetic intelligence in the video game Halo that projects itself as a sensuous unclothed woman. Digital assistants’ voices are also biased, being exclusively female, or female by default. Finally, gender biases are found in their behaviour.

Most leading voice assistants were originally designed to greet verbal abuse with flirtation. Siri, for example, would answer “I’d blush if I could” after being called a slut. Technology companies’ decision to gender digital voice assistants is almost certainly intentional, and while they often justify their choices by claiming that people find the female voice more trustworthy, there is no clear-cut evidence to back this up.

This worries me because the gendering of AI technologies has the potential to spread and reinforce harmful gender stereotypes that already exist in our societies by sending messages about how women and girls should respond to sexual innuendoes and abuse, how they should think about their bodies and how young people should interact. Ultimately, it may further stigmatise and marginalise women and may even offset the considerable progress that societies have made towards gender equality in offline environments.

Therefore, technology does not just replicate gender inequalities, it can also widen them.

Katie Frame, EOS: How did we get here and what caused the gendering of AI-voice assistants in your view?

Saniye Gülser Corat, UNESCOOne of the key causes of gendered AI assistants is that women are often missing from the technology field in the first place: in the STEM and digital skills pipeline at all levels of education; in technology workplaces as developers and programmers; and in the C-suite and the boards of technology companies.

Additionally, the data sets used to train the algorithms that power AI devices may be biased, and “bias in” will lead to “bias out”. This is especially difficult to tackle when biases in data sets reflect historical human biases entrenched in our societies.

Katie Frame, EOS: How can we address the ‘gendering’ of AI and other technologies?

Saniye Gülser Corat, UNESCOTo address gender biases in AI we need to:

  • Acknowledge and address historical human biases
  • Make incomplete or unrepresentative training data sets more diverse and inclusive
  • For each dataset, create a datasheet that documents its motivation, composition, collection process, recommended uses, and so on to increase transparency
  • Make algorithms and AI systems transparent and explainable so that they can be easily audited for unconscious bias
  • End the practice of making digital assistants female by default
  • Explore the feasibility of developing a machine gender for voice assistants that is neither obviously male nor female
  • Hire more women in the technology sector

Katie Frame, EOS: And are technology companies addressing gender bias in AI and other emerging technologies?

Saniye Gülser Corat, UNESCO: Microsoft has added a male Cortana voice and Google is using “adversarial testing” in an attempt to remove bias from its products before launch. Some technology companies have also updated their voice assistants’ responses to verbal sexual harassment.

Beyond voice assistants, Google Translate now offers gender-specific translations for gender-neutral languages and Google’s Vision API no longer uses ‘gender labels’ in images of people.

Katie Frame, EOS: Thank you Gülser for sharing your insights. In part two let’s discuss some recommendations for the private sector to address the gender digital divide.

UNESCO is the United Nations Educational, Scientific and Cultural Organization. It seeks to build peace through international cooperation in education, the sciences, culture, communication and information. UNESCO believes that all forms of discrimination based on gender are violations of human rights, as well as a significant barrier to the achievement of the 2030 Agenda for Sustainable Development and its 17 Sustainable Development Goals. UNESCO has therefore identified gender equality as a global priority for the Organization.

Read the second part of our Q&A here.

Case studies are shown to demonstrate engagement, EOS does not make any investment recommendations and the information is not an offer to buy or sell securities.

Related insights

Lightbulb icon

Get the latest insights straight to your inbox