Recent research has found how toxic and sexism filled are algorithms used by social media in identifying women and men, inculcating stereotypes that reduce women to their looks. The Artificial Intelligence used in these platforms is found to have categorised pictures of men under labels like “official” whereas women under labels such as, “smile” and “chin".
New study shows strong gender bias in image-recognition algorithms from Google, Amazon & Microsoft.
— MIT CSAIL (@MIT_CSAIL) November 20, 2020
Photo of man: "official," "speaker"
Photo of woman: "hairstyle," "smile"
Paper: https://t.co/6xdpXd0okL
More: https://t.co/OUiQGj2rjB (v/@tsimonite @WIRED) https://t.co/3l11lNydJI pic.twitter.com/w9fzr6ks71
The US and European researchers submitted pictures of congress members to Google’s cloud image recognition service.
Carsten Schwemmer who is a postdoctoral researcher at GESIS Leibniz Institute for the Social Sciences in Köln, Germany, said that this perpetuates the idea that women only need to look pretty while men are recognised for their work and as business leaders.
This also means that algorithms used by AI services do not see the world with mathematical detachment only but instead get influenced by historical cultural biases against women.
Added to that, the majority of the codes are written by men and many of them tend to propagate this gender bias.
AI services were always on point in identifying the gender of white men, but very inaccurate at identifying the gender of black women, as per earlier study named Gender Shades.