Quality of Hire Blog

arrowBack to Quality of Hire Blog

Pitfalls and possibilities: How data science is shaping selection science

by Jen Wason

Nationally recognized labor and employment attorney Mark Girouard and Eric Sydell, Shaker’s vice president of Research and Innovation, provided insights and advice during a December 13 web seminar about tackling some of the thorny issues around applying big data and machine learning to talent acquisition.

 

The increasing availability—and tantalizing promises—of big data solutions to find and hire candidates means understanding these tools and their risks and potential is essential to ensuring you’re making the right choices for your hiring methods.

Carefully considered application of big data techniques can offer insights into candidate populations, optimize sourcing, and potentially streamline hiring processes. But they also pose legal and other business risks employers need to anticipate and avoid.

What are big data and machine learning anyway?

Big data describes extremely large data sets that are analyzed using advanced processing techniques to reveal patterns and relationships. Machine learning is the programming that allows computers to draw inferences, offer insights, and predict outcomes beyond what human-scale analysis has ever been capable of achieving. Using algorithms that permit rapid trial and error adjustments, machine learning extracts patterns and predictions from data sets after an initial “teaching routine”.

Big data influences almost every part of modern living, Sydell explained. The technological breakthroughs that led to the development of the latest generation of artificial intelligence, including IBM’s Watson and Google’s sophisticated search engine, are part of deep learning technology at the heart of everything from self-driving cars to supercomputer-enabled medical diagnostics and refrigerators that will reorder butter just before you run out.

Industrial and organizational psychology uses many of the same analytic methods to offer talent acquisition professionals algorithms and predictive power to improve their hiring outcomes. Read more about how mature talent analytics can enable you to combine the best of computer-based predictive modeling and human decision making.

What is the future of hiring in a big data world?

Talent acquisition already involves more data than can possibly be processed by humans, Sydell reminded. An average of 250 applications are received for every corporate job. The average recruiter spends just six seconds evaluating each resume. Candidate data is high-volume, high-velocity, and vulnerable to exaggeration and misrepresentation. No wonder employers and hiring managers are looking for better ways to make better hiring choices.

Popular new big data hiring tools include social media mining, application and resume search tools, interview response analysis, and gamified and other interactive assessments. Data gathered and analyzed from these sources may highlight possible candidate's strengths, but on-the-job performance data is required to understand how a candidate will actually perform in a role. Knowing how to design the right methods and conduct relevant analysis are what differentiates selection science from data science.

“There is a lot of information out there, and not all of it is good information,” Sydell cautioned, “Just because something correlates doesn’t mean it means something.” The influence of data science on selection science must be tempered by an insistence on correlations that are not just statistically meaningful, but practically meaningful. When evaluating conclusions drawn from data science, ensure you probe what’s behind the claim, know the size and context of the data set. And, as with any data used in influencing hiring outcomes, be sure you can verify the job relevance of the outcome.

What are the risks?

Big data tools used in conjunction with hiring are considered selection processes and subject to the Uniform Guidelines on Employee Selection Procedures. Employers must be sure they can demonstrate job relevance and validity and that these practices are fair and unbiased, Girouard explained.

The EEOC is monitoring recent trends in the use of big data solutions to solve hiring problems and held a public meeting in October 2016 to explore current practices and attitudes regarding the issue. While formal guidance from the agency is pending, the convened panel highlighted a range of risks associated with big data hiring solutions, including disparate treatment (intentional discrimination against protected groups) and disparate impact (neutral selection tools that impact a protected group differently than other groups). Also, because some vendors do not disclose data sources, analysis may reveal or rely on prohibited considerations, such as disability status, protected leaves of absence, and other medical information.

Other risks include relying on conclusions drawn from statistical correlations unearthed in big data that may not, in fact, provide employers with meaningful information about an applicant's future job performance. The EEOC’s Uniform Guidelines require that employers demonstrate the importance of selection criteria to the job.

What are the rewards?

Selection science can help limit the human bias of individual decision makers in the hiring process. For example, some hiring managers may inadvertently practice like me hiring—where they choose applicants whose experiences, values, or personalities they most identify with. But kinship doesn’t predict performance. While care must be taken not to build such bias into an algorithm on the front end, big data can eliminate such bias by removing human judgment from the equation, Girouard explained.

Big data and machine learning tools conscientiously applied to selection science may offer greater predictive power than any existing sourcing strategies, optimize the hiring process, and improve the recruiter experience.

What's next?

In the race to make sense of the tidal wave of big data, you must first understand and confront the challenges of plowing through what’s bogus data to get to what’s valuable. When it comes to talent analytics, being attentive to objectivity, job relevance, and fairness are crucial.

Want to know more about capturing the right data to improve your quality of hire? Check out our industry-leading Virtual Job Tryout technology.

Save

Save

Save

Save

Save

Save