Automate HR, Proliferate Mediocrity

Machine learning algorithms are no smarter than the humans they learn from, and in most cases not nearly as smart.

I am no luddite. I am aware of the power of big data when used with good analytics and artificial intelligence in automating functions like logistics, supply chain management, manufacturing and market intelligence. However, business leaders ought to treat automating human resources with particular skepticism and caution.

I once attended an exclusive panel discussion on big data with experts from Google, IBM, SAP and Japan’s Fair Trade Commission. One panelist talked about applications used to automate screening of candidate CVs that use machine learning to mimic the decision-making of human resources managers. An application like this may help human resources screen lots of CVs fast, but is unlikely to help do it better.

Human resources managers are frequently the least effective people in an organization to screen CVs for all but the most menial of roles—and even then, I am not so sure. In fact, the CEO of a client company of mine ended human resources screening of employee candidates after learning that a uniquely ideal candidate had been screened out by his human resources, and subsequently accepted a job with a competitor. The candidate was exceptional, and did not fit any standard mould. Human resources managers and machine learning algorithms alike don’t deal well with exceptions, and that can cost a business dearly in some cases.

Another panelist talked about an experimental application that predicted an employee’s propensity to be promoted with ninety percent accuracy. But who is to say that the managers are promoting the right people in the first place? It makes me shudder to think how this application might be used. Who might be fast tracked for management, and given all sorts of development opportunities? Who might be sidelined? Sideline a star, and he or she will jump to a competitor. Promote a dunce, and watch the best of his or her staff flee your company in search of greener pastures.

Are we approaching the science fiction world of Minority Report, where people are arrested for crimes prior to committing them because a big data algorithm predicted they would? In that world, the technology is viewed as too perfect to make a mistake. In our world, our technology is not quite so perfect. I still get irrelevant Google ads and bizarre product suggestions on Amazon because of what are obviously faulty assumptions about what my surfing behavior means. I can live with that. Human resources applications, on the other hand, are betting the business on questionable technology and playing with people’s lives.

Be sure you exercise yours.