The machine-learning specialists of retail giant Amazon.com unleashed a huge problem that their newly developed recruiting engine did not prefer recruiting women. A few people who were familiar with this told Reuters- “The team of specialists has been developing computer programs since 2014 to review job applicant’s resumes efficiently and aiming at looking out for top talents in a mechanical way. Automating tasks has been a key goal in Amazon’s dominance in the e-commerce region, which includes automation inside warehouses, leading the pricing decisions and much more.
The ML specialist team developed an experimental hiring tool which used artificial intelligence to provide scores to candidates who had applied for jobs. This score ranged from one to five stars, similar to shoppers rating products and their experiences on Amazon website.
One of them said- “Everyone wanted this holy grail. They literally wanted it to be an engine that would take in a number of resumes and spit out the top five, and the recruiters would end up hiring those candidates.”
Amazon tried editing these programs to make them gender-neutral on such terms but this wouldn’t guarantee that the machines wouldn’t try other ways to show up and rate up male candidates.
Ultimately, this was disbanded last year and the people working on this project earlier lost hope. Amazon declined to comment on this recruiting engine or the challenges faced by it, but claim to be committed to providing workplace diversity and equality.
Reuters reported that as a result of Amazon’s experiment, they offered a case study regarding the limitations of machine learning, and also serves as a lesson to the growing list of large companies that include Hilton Worldwide Holdings and Goldman Sachs Group which are looking to automate some parts of the recruitment process. About 55 percent of the human resource managers in the US considered that artificial intelligence would be a part of their work routine within the next five years.
Employers have always wished to fulfill their dream of harnessing AI and ML technology to improve the hiring and reduce the emphasis on the subjective opinions of human recruiters. But some computer scientists said that there remains much work to be done to fulfill this dream.
Apart from gender bias, the other issues included models judgment which would downgrade the candidates from all types of jobs within the company. The results were mostly random and resulted in being shut down.
The CEO of HireVue (which is a startup near Salt Lake City), Kevin Parker said- “Automation is helping firms look beyond the same recruiting networks upon which have been long relied upon. You weren’t going back to the same old places; you weren’t going back to just Ivy League schools“. Their customers include Unilever PLC and Hilton.
Goldman Sachs has created its own resume analysis tool which matches candidates with the “best fit”. Microsoft Corp’s LinkedIn has gone further in the hiring approach and offers employers algorithmic rankings of candidates based on their fit for job postings on their site.
Activists voiced their concerns about the transparency in AI. The American Civil Liberties Union has currently posed a challenge to the law which allows criminal prosecution of researchers and journalists who test hiring websites’ algorithms for discrimination.
Rachel Goodman, a staff attorney with the Racial Justice Program at the ACLU said- “We are increasingly focusing on algorithmic fairness as an issue,” said
Amazon has definitely learned a lot from its failed AI development and now uses a much-watered-down version of the recruiting engines. In addition to this, a new team has been formed in Edinburgh to provide solutions for automated employment screening with a focus on diversity this time.