0 Comments
An Australian study has warned that job interviews orchestrated by AI have a higher chance of discrimination if they’re non-American or disabled. Videos of applicants sharing their experiences have also been making the rounds on TikTok.
If finding a new job wasn’t already hard enough, a new study in Australia has found that recruiters using AI for interviews could possibly be discriminating against certain candidates.
Those without an American accent or living with a disability are at a greater risk of not being hired or put forward for further stages of the application process, the study suggests. The findings were published in a paper by Dr Natalie Sheard from the University of Melbourne.
Her research estimates that 30% of employers in Australia are currently using AI recruitment tools, with estimations predicting further growth over the next five years.
According to The Guardian, AI recruitment company HireVue surveyed 4,000 employers globally, and found that 72% were engaging with automated technology to make hiring decisions.
Dr Sheard warns that hiring systems are screening and shortlisting candidates based on limited datasets. Most AI models currently being used for administrative processes like these have not been fed extensive information, meaning that choices are potentially being made unfairly.
23 human resource professionals were interviewed as part of the report. 13 said that they had used AI recruitment systems, most frequently for analysis and video interviewing. Most datasets incorporate and favour American data over other nations.
The study also states that error rates for transcriptions of non-native English speakers with international accents were higher than that of American applicants.
More errors means a less coherent record of an applicant, which in turn could lead to a higher likelihood of being rejected by the recruitment algorithm.
Crucially, Dr Sheard argues that there is very limited transparency surrounding the implementation of AI within the hiring process for all parties involved. This makes it much harder to acquire feedback or understand why certain decisions have been made.
