16 October 2018

Press Release: Amazon's AI hiring 'bias' against women a warning for all employers

Reports that Amazon scrapped an artificial intelligence (AI) recruiting tool after discovering it had a inherent bias against hiring women is a warning to all employers, according to Bartier Perry Workplace & Culture Partner Darren Gardner.

Amazon began using AI to recruit software developers and other technical staff in 2014 but only a year later noticed it had developed a bias against women.

According to Reuters this was because Amazon’s computer models were trained to vet applicants by observing patterns in resumes submitted over the last decade, most of which had come from men.

“With an increasing number of Australian employers looking to automate their recruitment, the fact Amazon’s AI was reportedly excluding CV’s that even included the word “women” should be a flashing red light, “ said Mr Gardner.

“The problem with AI recruitment is that algorithms are built mainly to learn from data from the past. Data from past hires will be inherently discriminatory. Discrimination law has long recognised that present disadvantage, especially for women, indigenous people and people with disabilities, is the product of past discriminatory practices that resulted in conscious or unconscious bias.

“The assumption then that an algorithm or machine will automatically provide some sort of non-discriminatory workplace nirvana needs to be challenged,” said Mr Gardner.

Along with relying on past bias AI cannot also discern future human potential.

“Really good recruiters have a natural instinct for detecting untapped potential and talent. They do not hire in their own image or because of past practice. They don’t let superficial assumptions of gender, physical appearance and past disadvantage infect a sound business decision.

Most successful people will talk about someone who gave them a ‘lucky break’ in their career. Rather than luck though someone actually made a really smart recruitment decision, based on human intuition, that there was plenty of potential to nurture and grow a future rising star,” he said.

With over 50 percent of United States employers reportedly looking to automate or AI their recruitment process, employers here could face substantial legal risks if those processes are later found to be discriminatory.

“Amazon reportedly noticed the flaws in its processes and scrapped the program which is admirable but if an Australian court finds an employer liable for its automated discriminatory decision-marking, the potential financial consequences and reputational damage could be significant,” said Mr Gardner.