An increasing number of companies are turning to artificial intelligence as a way to improve and streamline the hiring process — and weed out common human biases. But can you truly take the human out of human resources? Does AI really know best?
When it comes to humans, hiring biases are inevitable
Human interviewers make instant (and often unconscious) judgements about applicants based on things like the color of their shirt, their hairstyle, what sports they played in high school, and even their smell. These biases go beyond the applicant’s physical and personality traits too. If an interviewer is having a particularly good or bad day, that can significantly impact how they rate candidates. If their set of questions vary from interview to interview, as is often the case, some candidates will have an advantage over others, simply because the field isn’t level. And the interviewer may even have an inaccurate idea of what personality type or skill set it takes to be successful in the role they’re hiring for. When that’s the case, their hiring choices are practically doomed from the start.
AI offers one alternative to widespread hiring biases
While people are prone to judge one another or let their rough morning influence their hiring decisions, AI is immune to many common biases. It can be programmed to sift through thousands of applications and highlight the best job fits. It can record interviews, transcribe the conversation, and recognize answers that imply good (or bad) job fit. And it never gets tired of asking the question, “Will you tell me about yourself?” AI overlooks the factors humans commonly misidentify as red flags, allowing companies to process a greater number of applicants quickly, fill roles efficiently, and help more people find the right jobs.
But is AI really unbiased when it comes to interviewing and hiring?
The reality is, AI is built and trained by humans. Human bias is impossible to completely overcome. When the dataset that teaches AI is biased from the start, the technology only amplifies the problem. And that’s a significant flaw Amazon discovered the hard way.
In 2014, Amazon started building an AI tool that would process applications and assign applicants a rating of one to five stars, five being the best. The system would then recommend the top applicants out of the stack, and Amazon would hire them. But within a year, Amazon realized the system was strongly favoring male applicants. The model was taught to recognize talent potential based on the quality and longevity of the past 10 years of employees. Those employees were largely men. By the time the problem was recognized, Amazon had already filtered out numerous applications containing any variation of the word “women” and downgraded applicants who attended women-only universities.
For many hiring professionals, Amazon’s misstep was a cautionary tale. If left unchecked, AI can be no better, and maybe even worse, than real humans.
Regulators are watching closely, but AI hiring is gaining traction
AI technology has come a long way since 2014. Today, many companies that use this technology now have an ethics board in place to double check the algorithms. While the AI systems of old were described as “black boxes” shrouded in secrecy, today’s algorithms are clearly explained up front to HR buyers. Some AI hiring programs even send the results to the applicants themselves, explaining how they ranked and why they were or were not a good fit for the position.
The risks of using AI in hiring are not ignorable, but in the midst of today’s widespread hiring crisis, the efficiency gains are worth it for a growing number of organizations.
