Study finds gender bias present in artificial intelligence algorithms
A research report on promoting gender equality in artificial intelligence algorithms has been released in Shanghai with officials and NGO experts calling for joint efforts from governments, businesses and the local community.
Gender discrimination has been observed in a wide range of applications and AI algorithms, according to the report.
For example, bias exists on AI facial recognition platforms and in the job market. The recruitment algorithms used by some companies give lower scores on resumes with key words indicating that an applicant is female, lowering their job application acceptance rate.
The report was released on Monday by the Mana Data Foundation, a public welfare foundation focusing on data science and technology, with the support of UN Women.
Gender inequality in AI algorithms originates from human factors, said Kuang Kun, leader of the project team behind the report.
A seminar with participants sharing insight into promoting gender equality in AI algorithms
Algorithm engineers do not include gender equality when they develop algorithms, which can magnify discrimination, said Kuang.
Some 58 percent of staff in the AI algorithm industry were not aware of gender discrimination in algorithms, and 73 percent were unaware of malicious algorithms targeting women in the field, according to the report.
The newly adopted China Civil Code promotes gender equality and protects women's rights and interests, and a program to promote gender equality of AI algorithms is being implemented in China, filling a vacuum in the area, said Ma Leijun, senior project officer of UN Women China.
"The discussion of gender equality in AI algorithms is still in the early stages, but it has a big impact on what society will look like in the future," said Guo Yundan, deputy director of the Women's Federation of the Pudong New Area.
"The digital process poses a bigger challenge in promoting gender equality," she said.
The report called for measures such as enhancing gender equality awareness of data producers and conducting algorithms screened by third parties.
