Gender Bias in AI: The 2024 Wake-Up Call

Ahead of International Women’s Day 2024, the global community received a sobering reminder of the technological challenges facing gender equality. A comprehensive UNESCO study has revealed deeply worrying tendencies within Large Language Models (LLMs) to produce not just gender bias, but also homophobia and racial stereotyping. The findings underscore a systemic failure in the current trajectory of artificial intelligence development, where women are described as working in domestic roles four times more often than men. Female names are frequently associated with domestic concepts like “home,” “family,” and “children,” while male names remain linked to “business,” “executive,” “salary,” and “career.”

This alarming evidence is further supported by an extensive analysis conducted by the Berkeley Haas Center for Equity, Gender and Leadership. After examining 133 AI systems across various industries, the center found that 44% of them displayed clear gender bias, with 25% exhibiting a intersectional combination of both gender and racial bias. These biases manifest in visual representations as well; for instance, the simple prompt “lawyer” disproportionately generates images of older Caucasian men, while the prompt “nurse” almost exclusively results in images appearing female. This “digital normalization” of stereotypes risks cementing outdated social roles into the very infrastructure of our future.

Beyond the internal logic of the machines, a 2024 survey by the Federal Reserve Bank of New York has revealed a concerning behavioral trend: women are increasingly avoiding generative AI technology. While 50% of men reported using generative AI in the past twelve months, only 33% of women did the same. Across most major studies, the share of women adopting AI tools is consistently 10 to 40 percent smaller than that of men. Many women are raising critical ethical questions about whether it is responsible to use these tools at all, given their current flaws. However, this avoidance creates a double-edged sword, as shying away from AI could lead to widening pay gaps, reduced job opportunities, and a significant disadvantage in a tech-driven global economy.

The mechanism by which AI reinforces these stereotypes is rooted in the “training data bias” that mirrors existing societal prejudices. As AI researcher Beyza Doğuç explains, artificial intelligence acts as a mirror to the biases present in our society, which then manifest in the data used to train these systems. Furthermore, since AIs featuring female characteristics are predominantly developed by men, they often reflect male-centric ideas and fantasies about women. This representation crisis is echoed by students like Natacha Sangwa, who observed during international coding camps that AI is mostly developed by men and trained on datasets that primarily represent male perspectives and experiences.

To counter this trajectory, the movement for “Feminist Data Practices” advocates for a radical shift in how we handle information. This involves moveing past the gender binary, valuing multiple forms of knowledge, and prioritizing local and Indigenous wisdom to challenge unequal power structures. There are signs of corporate movement as well; in February 2024, eight global tech giants, including Microsoft, endorsed the UNESCO Recommendation on the Ethics of AI. This commitment calls for ensuring gender equality in tool design, ring-fencing funds for gender-parity schemes, and providing financial incentives for women’s entrepreneurship in the tech sector.

Systemic change also requires robust policy interventions, starting with a significant increase in women’s participation in AI development. This must be achieved through quotas, incentives, and the creation of truly inclusive work environments. Simultaneously, existing systems must undergo regular audits for bias with transparent reporting mechanisms and mandatory corrective measures. We must diversify datasets to ensure representative training data that includes diverse voices and avoids perpetuating historical injustices. The urgency cannot be overstated; as AI technology rapidly integrates into every facet of life, failing to address these biases will roll back decades of progress on gender equality.

Ultimately, the future of artificial intelligence must be a “Feminist AI”—one that is inclusive, equitable, and representative of the entire human species. Everyone has a role to play in this transformation, from educating ourselves about algorithmic bias to advocating for systemic change in our workplaces. By supporting organizations working on ethical AI and participating directly in tech development, we can ensure that diverse voices are not just heard, but are foundational to the code itself. We must ensure that AI does not become another “lie of equality,” but instead serves as a powerful tool for advancing justice for all.