AI gender bias stems from training data and social stereotypes

Artificial intelligence systems are increasingly being identified as tools that reproduce human biases and gender stereotypes. Experts note that these biases are not intentional but are a byproduct of the data used to train large language models. Because these models learn from existing materials, they can perpetuate social stereotypes, such as associating specific professions with a particular gender.

A United Nations study of 133 AI systems found that 44 percent demonstrated gender bias. This can lead to practical harms, such as financial tools failing to account for women’s longer life spans or chatbots providing information that ignores women’s unique healthcare needs.

To combat these issues, advocates emphasize the need for greater female participation in the design and development of technology. In Cyprus, Gender Equality Commissioner Josie Christodoulou highlighted the importance of encouraging girls to engage in science from an early age. Meanwhile, entrepreneurs like Shubhi Rao, founder of Uplevyl, are developing specialized AI models specifically designed to address women’s needs in the workplace, finance, and reproductive health.