Artificial intelligence and gender equality policy goals

Risks and opportunities

Artificial intelligence and gender equality have a complex relationship. On the one hand, AI can reinforce existing patterns of inequality – for example, when algorithms are trained on biased or incomplete data and produce discriminatory outcomes. On the other hand, AI can be used to highlight patterns, find knowledge gaps and provide better decision-making for a more equal society.

For AI to contribute to positive developments in the area of gender equality, a more purposeful and knowledge-based approach is required than today. In the report, we describe how AI and gender equality are linked. We highlight both key challenges and opportunities. We also provide recommendations for how AI's potential to contribute to increased gender equality can be further strengthened, says Sophia Ivarsson, Vinnova's expert in gender equality issues.

The report points out that AI development is largely reactive: gender equality issues are often addressed after problems have already arisen, which damages trust in the technology. To seize the opportunities, gender equality must instead be an integrated starting point in all uses of AI – from needs analysis and data collection to development, procurement and follow-up.

Three recurring risks

• AI can create inequality– new ways of working, platforms and automated decisions can provide different term and opportunities for women and men.
• AI can recreate inequality – historical data and established norms can be built into models and disseminated on a large scale.
• AI can lead to gender discrimination – directly or indirectly, for example through proxy variables that co-vary with gender.

The report also highlights areas where the risks are particularly clear. Some examples include discriminatory outcomes in credit and recruitment processes, deficiencies in content moderation and transparency on digital platforms, as well as increased forms of online violence and the spread of deepfakes that disproportionately affect women and girls.

How AI can become a tool for gender equality policy goals

Vinnova has carried out a number of activities within the AI for increased gender equality initiative since 2019. Experience from research and innovation projects shows that AI, under the right conditions, can support follow-up and development of gender equality policy goals.

A key issue is the data that underpins AI use. Gender is often underrepresented in datasets, which can create systematic biases in analysis and decision support. At the same time, technology is never value-neutral, so AI needs to be used as a complement to other tools and methods. It is the interaction between humans and technology that determines whether AI can contribute to more gender equality in society, says Sophia Ivarsson.

Some examples of opportunities aimed at gender equality policy goals:

• Equal distribution of power and influence: AI-supported educational efforts can strengthen girls' and women's paths into STEM and technology development.
• Economic gender equality: decision support can reduce the effects of conscious and unconscious bias in, for example, investments, credits and insurance, provided that the models used have undergone thorough gender equality analyses.
• Gender-equal education: personalized feedback and support that can contribute to better academic results and reduced gender gaps.
• Gender-equal health:improved diagnostics and more accurate treatments, especially in areas where women's symptoms have historically been underexplored.
• Men's violence against women must end: AI that strengthens digital security, detects threats and harassment, and supports preventive efforts, with a focus on safety and rights.

Needs ahead

Today, there is a lack of clear need owners and a well-developed market for AI solutions that specifically address gender equality challenges. AI development is largely dominated by industry. Therefore, clear political initiatives, targeted funding and long-term responsibility from both public and private actors are required.

If we only address gender equality as an afterthought, AI will continue to reinforce inequality. With the right innovation policy, we can instead develop, test and adapt AI that strengthens a gender-equal society, says Sophia Ivarsson.

Five proposals for gender-driven AI

In the report, Vinnova has developed five proposals for measures:

1. Nationella AI-piloter who takes on concrete gender equality challenges in different sectors and builds in gender equality expertise from the start.
2. Development of gender-informed AI models, such as language models and analysis tools that are trained and validated against knowledge of gender equality.
3. Cross-sectoral partnerships between the public sector, business, civil society, AI developers and gender equality experts – with gender research integrated throughout the chain.
4. Procurement as a policy instrument, where eligibility requirements on gender equality analyses and measurable effects drive the market towards desired societal outcomes.
5. Strengthened skills development to meet both AI skills needs and the continued underrepresentation of women in STEM.

The report in brief

• AI has a dual function: it can both reinforce inequality and help reduce it.
• The risks include discriminatory decisions, online violence and lack of transparency in platforms and data.
• Regulation like the EU's AI Act is important, but does not always capture indirect gender effects.
• To seize the opportunities, gender equality needs to be an integrated starting point in development, procurement and use.
• The report suggests pilots, models, partnerships, procurement and skills upgrading as key steps.

More about the report

The report is an expert report for the 2024 Gender Equality Inquiry and has previously been published on the website sou.gov.se.
2024 Gender Equality Report (A 2024:06)