3 Ways Machine Learning is Improving the Hiring Process
Technology’s advance into all industries and jobs tends to send ripples of worry with each evolution. It started with computers and continues with artificial intelligence, machine learning, IoT, big data and automation. There are conflicting views on how new technology will impact the future of jobs. But it’s becoming clear that humans will need to work with technology to be successful — especially as it relates to the hiring process.
There’s a great example of this explained by Luke Beseda and Cat Surane, talent partners for Lightspeed Ventures. On a recent Talk Talent To Me podcast episode, they spoke with the talent team at Hired, where I work, about why it’s critical to understand why a candidate is pursuing a given job. They concluded that machines can’t properly manage the qualitative aspect of hiring. For example, machines can’t tell if a candidate is seeking higher compensation or leveraging a job offer to negotiate new terms with their current employer.Humans can.
However, machines are better at making processes more efficient. For example, machine learning brings value by processing job applications faster than humans — which can reduce the amount of time it takes to recruit and hire a new employee. With that in mind, here are three ways machine learning is improving the hiring process:
Recommendations To The Rescue
Most HR professionals today use recruitment platforms to find potential employees through a search-based system where they can narrow down a list of candidates based on factors like skill, industry, experience and location. But with machine learning capabilities, hiring managers don’t have to manually dig through applications from hundreds of candidates to find the best fit. Instead, they can rely on networking and job sites to leverage machine learning and offer intelligent recommendations on the candidates who can fill a given role. This enables a more efficient hiring process for both job seekers and recruiters.
The Elimination Of Bias
Machine learning can help level the playing field in hiring. It can be employed to provide equal exposure to opportunities, regardless of a candidate’s pedigree or background. Algorithms should focus on skill-based data, not on the universities where a candidate has studied, the companies where they have worked, or their ethnicity or gender.
One factor to consider is that candidates simply don’t know their value and what compensation they should ask for. This is another area where machine learning can help. It can expose salary data for a candidate’s specific role and geography, thereby making them better informed. On the employer side, it can also analyze and source salary data. This gives companies a clearer picture of a suitable salary offer, which is based on a candidate’s skills and experience instead of their previous salary.
Humans are inherently biased. But in most cases, we suffer from unconscious bias. Machine learning can help humans overcome these biases with data. Leveraging data, we can create awareness around preferences of hiring managers and recruiters. We have a natural propensity to surround ourselves with people who look like us. This is where data can help. By surfacing statistics around diversity, employers can be better informed and not risk losing out on diverse candidates.
There are several studies that show that diverse teams result in better financial outcomes. However, these studies are based on surveys. Machine learning can quantify these outcomes for their organizations, which will then incentivize better human behavior when it comes to hiring.
Organizations that rely on machine learning to strengthen their hiring processes will probably find that one of the biggest challenges is building or using a platform that’s free of biased hiring and wage gaps. And while machine learning can certainly improve the hiring process, the responsibility to ensure the integrity of the algorithm lies with the human engineers building the platform. Hence, it’s incredibly important to do this right — otherwise, the algorithms will perpetuate human biases.
First, training data sets for all algorithms is very important. If the training data set has the same inequality that we see in our world, then the algorithm will not help solve these inequalities. To this end, it’s essential to normalize the training data. If, for example, there are 1,000 candidates in a data set (800 men and 200 women) and the data is only used to determine which candidate will be a successful hire, the algorithm will likely select men as the most successful candidates based on sheer numbers.
But there’s a way to normalize the candidate pool. By analyzing 200 applications from women and 200 from men instead of the full data set, the algorithm will have a lower chance of bias. This level of quality assurance is especially critical for the industries with a higher percentage of male candidates. With more male candidates, then the volume of inherent data will be increasingly biased, and over time, women might not even make the first cut on the platform.
Secondly, it is important for algorithms to make recommendations based on skills. Focusing on pedigree is only going to promote a smaller select pool of candidates.
Lastly, algorithms and humans can be equally biased. So it’s important that both are held to the same benchmarks and that these benchmarks are measured. Humans, working in concert with algorithms, can make recruiting efficient and fair.
Machine learning will bring the most success to the organizations that use its capabilities to increase productivity for their employees. This is especially true with hiring. Where machine learning can help narrow down and suggest job candidates, hiring managers can handle interviewing, negotiating and understanding the human on the other side of the table.
This piece originally appeared on Forbes.