Source: in-cyprus.philenews.com
The under-representation of women in technology risks reinforcing gender bias in artificial intelligence, Gender Equality Commissioner Josie Christodoulou told the Cyprus Forum on Tuesday, warning that AI systems inevitably reflect the people, data and social assumptions behind them.
Christodoulou pointed to the predominance of men working in technology and innovation, arguing that human biases can consequently find their way into the AI systems being developed.
“I am glad that we are not starting the discussion by asking whether AI is actually biased. We are starting from the recognition that AI is indeed biased,” she said.
Speaking during a panel on AI and equal participation, Christodoulou argued that addressing the problem requires intervention well before AI systems are developed, beginning in schools and with the choices girls are encouraged to make about education and their future careers.
She said Cyprus was already working with the Education Ministry to train career guidance counsellors on equality and conscious and unconscious bias, with the aim of ensuring that girls who want to pursue STEM subjects and related professions are not discouraged by gender stereotypes.
The problem, however, does not end with getting more women into technology. Christodoulou said women also tend to leave such professions along the way, particularly when they reach the stage in their lives when they begin considering starting a family.
Measures including childcare provision and policies that make it easier to combine family and professional life could help, she said, but the deeper challenge remained cultural.
“The biggest problem we have to address is culture,” Christodoulou said.
She pointed to the different expectations society continues to place on boys and girls, saying boys are commonly asked what they want to become professionally, while conversations with girls frequently emphasise motherhood.
Christodoulou stressed that the objective was not to diminish motherhood but to ensure women and men have genuine freedom to decide what they want to do with their lives.
Those attitudes, she argued, are established extremely early and ultimately feed into the assumptions and information from which AI systems learn.
“Stereotypes begin from the moment we learn the sex of our child, before the child is even born,” she said.
Christodoulou referred to research involving schoolchildren in which children asked to imagine an astronaut, doctor or firefighter generally pictured a man, while professions such as nursery teaching were associated with women.
The panel was chaired by Andreas Demetriou of the AI Integrity Institute, who illustrated how the same mechanism can be reproduced by artificial intelligence.
He gave the example of an AI system trained on material in which doctors are consistently portrayed as men and nurses as women. Faced with such data, he said, the system learns to associate being a doctor with being male.
The discussion also addressed the difficulties countries such as Cyprus face when AI models are developed and trained internationally, potentially importing cultural and other biases from outside the country.
Christodoulou said governments, the European Union and private technology companies all had a role to play in tackling the problem, with equality and human rights incorporated into the development of AI policies and systems.
Read more: