Magazine article HRMagazine

The Promise and Peril of 'Big Data'

Magazine article HRMagazine

The Promise and Peril of 'Big Data'

Article excerpt

"Big data" is everywhere-including the workplace-and can now be routinely accessed from devices that can be carried in a pocket. Types of data can include hours worked, inventory tracking, measures of customer interactions, social media usage, and performance and productivity assessments.

Once translated into discrete data sets, the information harvested from myriad sources can offer useful and measurable insights into your workforce. It can shape nearly all aspects of managing current and prospective employees, including recruiting, hiring, scheduling, benchmarking, succession planning and determining security risks.

But while analytics can help HR professionals and business leaders make better decisions and cut costs, using them is not without legal risk.

"Employers are deriving considerable benefits from analytics," says Kate Bischoff, SHRM-SCP, an attorney and HR consultant at tHRive Law and Consulting LLC in Minneapolis.

"As attorneys, if we were to tell employers that they can't use new technology because of the potential for risk, our clients would fire us," she says. "So I promise all my clients that I won't say 'no' to new technology that has a legitimate return on investment. We just need to install a metaphorical seatbelt or airbag on it."

In other words, you have an attorney's blessing to mine big data. But you need to understand the risks and how to mitigate them.

Discrimination Danger

Among the most pressing concerns inherent in relying on big data is that improperly used HR analytics can result in employment discrimination.

Consider this scenario: An employer wants to beef up its sales force. To tailor its recruitment efforts, HR professionals and hiring managers seek to identify the qualities in the company's current sales team that correspond to high sales returns. So they plug in all the information they have about the activities each individual representative engages in: the number of customer calls made versus e-mails sent, the frequency of interaction with the sales manager and perhaps even how often the customer lead database is accessed. They also input demographic information: birth date, race, gender. The data reveal a surprising finding: Being white, male and 32 to 37 years old correlates to strong sales numbers.

But correlation is not causation. Perhaps the vast majority of the sales team is white-making it unlikely to detect a meaningful relationship between sales results and other races- or maybe performance is related to a third "confounding" factor that also correlates to a certain age range or ethnicity. If HR professionals and hiring managers were to ignore these possibilities and take the data at face value, they would risk making unwise hiring decisions based on erroneous-and biased-assumptions.

More important, they would also be acting unlawfully by using age, gender and race as hiring criteria.

"If you're trying to define what makes the best employee based on historical data you already have, you potentially re-create the homogeneity, which is not what you want," Bischoff says. Because a poorly conceived algorithm can produce discriminatory outcomes, it's important to make sure you validate all algorithms before acting on them. Consider whether data inputs fairly correspond to desired traits or whether the use of certain data sets skews the analysis.

"We need to figure out how to use the valuable data we have and then work against the bias that we unconsciously have," she notes. "We have to ensure we don't use patterns that can create potential disparate impact"-an adverse effect based on race, color, sex, national origin, religion, age or disability. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.