In researching our weekly Most Important Developments in HR post (if you haven’t checked it out, give it a try. We know you’ll like it!) we found a new series of articles, “AI and bias: Building fair and equitable machine learning systems,” which Harvard Business Review has put together. It’s tackling the conundrum of AI and bias, and the articles are worth a read for any HR pro considering using AI in hiring, and even for those who aren’t.
In these early days of AI and its massive implications for HR, it’s difficult to know exactly what’s what. Every day it seems there’s a new “definitive” study that points in the opposite direction of the study that came out the day before.
One day you read that using AI to pluck the best candidates from a sea of resumes is the best way to avoid bias in hiring. The next day, that theory is flipped on its ear by something like Amazon’s AI debacle. Amazon built its own AI recruiting and hiring tool, launched it, and subsequently found out it was biased against women. Eerily enough, it couldn’t be reprogrammed to eliminate its bias even by the best minds in the tech world — it kept finding ways around the new programming to return to its biased ways. The tool had to be scrapped. It sent Amazon back to the drawing board.
So, what are we left to believe about AI in recruiting and hiring? Does it promote or eliminate bias? That depends. It seems that right now, there’s no clear answer. But here’s what we know.
AI comprises two main technologies: machine learning and deep learning.
Machine learning lets machines make decisions and predictions and, in effect, learn from data. It’s about pattern recognition wherein the algorithms learn from the data. Such is the problem with Amazon’s AI. The data included resumes from engineers who had been hired by Amazon in the past decade or so, with the thought that the AI could spot similar resumes in the vast sea of applicants. It did that. Too well. One variable that the designers failed to consider was, over the past decade, most of the engineers hired in the tech world have been men. So the fact that an applicant was a woman was a strike against her, in the “mind” of Amazon’s AI tool.
Deep learning goes a step further and allows the machine to think even more like a brain, giving us things like image, video and speech recognition, and those eerie chatbots that can understand not just language but tone, too.
In the beginning, AI enabled HR to become more efficient with automation. But now, it’s being used for recruitment, performance management, employee engagement and training, among other applications.
And so where does Amazon’s cautionary tale about bias fit into the mix? That’s the gray area right now, but common sense would dictate that we look at some evident truths in the language we use to talk about this technology and who it’s for. Artificial intelligence. Human resources. The bots aren’t replacing us. They can’t. It seems like a combination of the two will take us where we want to go.