Early generations of artificial intelligence focused on the tool’s ability to spot patterns, especially in large data sets. Yet AI could only spot the patterns it had been trained to see, and it had a habit of assuming a common pattern would repeat if it had no information to suggest otherwise.
Many of these early, simplistic algorithms are still in use today. Companies that rely on this increasingly outdated software, however, may find that it hinders their efforts toward inclusive, thoughtful hiring more than it helps. Rather than spotting hidden talent, a too-simple algorithm may exclude qualified candidates.
Early ATS Algorithms Baked in Bias, Missed Key Skills
Early attempts to use artificial intelligence to improve hiring decisions were based on a few core assumptions. They assumed, for example, that reviewing the careers of past superstars at a company would reveal which traits these workers had in common. They also assumed that hiring for those traits would automatically result in more superstars.
“Since algorithms are backward-looking and learn from past data, the decisions made by AI programs may reflect and repeat past biases,” write Andrew Gray IV and Melinda Riechert of the law firm of Morgan Lewis.
For example, an early AI program used to examine a company’s past successes may conclude that one of the top qualifications for success is having graduated from one of three specific colleges or universities. Companies trusting that insight may then focus their hiring on these schools, not realizing that the pattern exists because in the past, the company’s founders — each a graduate from one of those three schools — tended to hire candidates who shared the experience of attending that college or university.
Today, the best AI-enabled talent management and hiring systems are constantly monitored and updated to help them avoid falling into bias traps. They also have access to much larger data sets, which can help reduce the impact of localized biases like a company’s tendency to hire graduates of the same colleges as its leadership. Early algorithms, however, lack this support, making them more likely to bake in bias.
To Expand Your Talent Pool, Expand Your Software’s Abilities
Artificial intelligence has learned and grown over the last several years. The data sets AI can access to make deeper, more insightful decisions have also increased exponentially in size. Today, the best talent-management software is built to take full advantage both of advances in AI technology and in access to massive data sets.
A focus on skills is one benefit of current AI tools used for hiring. Many workers have the skills to fill certain in-demand roles, even if their career to date has been on a different track or in a different industry. An AI-enabled system that can identify adjacent skills can help companies identify these workers as promising candidates, write Emily Field and fellow researchers at McKinsey.
Dive deeper into the talent pool by going one step further: Choose software that maps workers’ careers as sets of skills and paths of growth, rather than as discrete roles. These more advanced AI tools can spot, for example, whether a candidate coming from a different industry has developed the same skills required for the role your team needs to fill.
Such software can also help your current staff rethink their own career paths. With the insights provided by up-to-date AI, your staff can imagine and then move themselves into roles they may not have previously considered because they didn’t think of the role as related to their own or did not realize how closely the skills required match the skills they have.
As you reconsider your digital tools, rethink how your hiring processes approach skills and capabilities as well. According to Sapana Agrawal and fellow researchers at McKinsey, companies seeking to hire for skills “should quickly identify crucial value drivers and employee groups,” determine exactly how each role contributes to the business’s overall value creation, and consider how these roles may shift in coming years.
Images by: rawpixel/©123RF.com, nito500/©123RF.com,