As a business buzzword, “machine learning” doesn’t measure up to some of its more popular cousins – big data and transformation being two of the standouts – but this may not be the case for much longer.
What was once something only used by people working in fields like software engineering, statistical modeling, and artificial intelligence, is now being used in many more recognizable and consumable forms.
Online services such as Google, Netflix, Amazon, and Spotify all use machine learning to improve the customer experience. For example, if you’ve been streaming Barry White all week, Spotify’s “Discover Weekly” playlist will begin suggesting other similar artists. Listeners may start seeing Luther Vandross pop up in their playlist, for example.
As Spotify learns your musical tastes, it begins suggesting – or, more accurately, predicting – relevant artists or genres based on other listeners showing their approval for similar suggestions. In the same way, insurance companies use machine learning to help combine credit card information, driving history, and biographical data to predict what sort of driver an applicant may be based on the accident history of drivers with similar information and behavior, and then set premiums accordingly.
Machine Learning and Recruiting
So using historical data to predict what a human being will do or like isn’t that new but it is only now that the world’s HR departments are realizing how valuable it can be. Combining employee and candidate data in the right way can help companies get more out of their most important assets: human capital.
Machine learning is particularly helpful when it comes to recruiting. Removing human biases and improving selection within the hiring cycle is something that every head of HR asks their recruitment team to do. The advent and widespread adoption of psychometric assessments was a significant stride forward, and now algorithmic assessments – a form of machine learning – seems to be the next step in the process.
Algorithmic assessments apply statistical models to candidate information (e.g., application form data) and, as the name suggests, use algorithms to predict the likelihood of whether a candidate will be a good fit.
This replicates what recruiters and hiring managers try to do when they look at a candidate’s background except the algorithms draw on far more information for an informed decision, can compute results quickly and within a specific degree of statistical confidence, and are not limited in the number of features they can consider at any given time. All of this tends to make the analytical process produce far more robust decisions.
Less Incremental Effort
Another stark difference between traditional screening methods and algorithmic assessments is that there is far less incremental effort required to consider a higher number of candidates. Since algorithms are created by looking for patterns of success or failure (in this, case whether a new hire was successful in their new role) in historical candidate information, all that is required to come up with a prediction of success or failure for prospective candidates is a comparable piece of prospective candidate information.
Once a candidate submits an application, the algorithm instantly analyzes it and returns a predictive score to the recruiter. This score can be used in multiple ways: to either push the applicant to the next stage of the hiring process or not, or as a complementary score along with a psychometric or skills-based assessment. Indeed, some of the most forward-thinking companies have already begun to make much more use of data to hire and manage their employees.
Future posts in this series will look more in-depth at algorithmic assessments to understand how they differ from conventional, psychometric assessments, how they can be applied in many different business scenarios, and how HR teams can implement the technology.