I was recently asked to create a minimalistic briefing note on “big data and the workplace” for a group of experienced employment lawyers. Here is what I wrote.
Employers are using data analytics to derive insights about their employees. They are then using the insights to make decisions about individual employees and potential employees. The objective is to make better human resources decisions.
The first and major big data application for employers was hiring. Hiring analytics involves merging historical data about candidates and employees into a database and using software to analyze the data to identify measurable candidate attributes that correlate with successful employment.
Today, employers use data analytics for a range of other applications – those supporting performance management, health and safety and security, for example. All these applications involve a similar process, similar technology and similar techniques to those involved with hiring analytics.
The use of workplace data analytics is popular and legitimate. It is naïve to suggest that the use of data analytics is wrong-headed, though there are legal risks.
The greatest risk is the risk of liability under anti-discrimination statutes. One can hardly blame employers for attempting to determine which candidates are most likely to be successful. Some argue that the use a good predictive model can actually reduce discriminatory bias!
This optimistic view of workplace data analytics is theoretically sound, but problematic in practice. The discrimination risk exists, in part, because the predictive models are typically developed by third-parties can be poorly understood by the employer-enduser – i.e., predictive models exist in a “black box.” This may make defending the use of even the most sound model very costly and risky. And when a model produces a result that disadvantages those with certain protected personal characteristics, a human rights tribunal will certainly question “Why?” “Is there a systemic discrimination problem that underscores the result?” When diversity is now valued by business, workplace data analytics, if used mechanically, can lead organizations to swim against the flow.
Privacy is also an issue, though there is a disconnect between potential employee perceptions and actual privacy impact. Strictly speaking, analyzing data to derive insights about a population is not a use of personal information at all. Service providers and employers are wary of privacy concerns, and usually do not publish insights about small populations (where the risk of identification is high). Employers should also communicate with employees about the nature of their analysis with a view to putting employees at ease and reducing the risk of complaints that arise out of a misunderstanding.
Ensuring the analysis is “true statistical analysis” will address part of the privacy concern associated with the the use of workplace data analytics, though there is still a significant data handling issue that will remain. Workplace data analytics involves compiling existing data (and sometimes augmenting it) to create a large data source. Even if the data source may only be used to understand a populations or groups within a population, to support sound statistical analysis it must include data that is linked to individuals. The data source could therefore be compromised and cause harm to individual privacy. Data security – particularly given the data source will almost always be handed by a third party) – is of paramount importance.