Working for the algorithm Machines will help employers overcome bias - The Economist
Who is best placed to judge a firm’s workers? In 2018 employees everywhere will increasingly feel the effects of the rise of “talent analytics”, also known as “people analytics”, as they go about their daily work. Having been relatively slow compared with other corporate departments in making use of big data, in 2018 human-resources (HR) folk will become its most enthusiastic proponents—with significant implications for who gets hired, what they are paid and whether they are promoted. Employees will have to get used to being (often unwitting) guinea pigs in frequent HR experiments. And wise ones will think ever more carefully about how they express themselves in e-mails and on digital collaborative-working platforms such as Slack.
One reason is the pressure HR executives will face to make workplaces better for women and minority groups. The limitations of established approaches, such as training and awareness programmes, had caused “diversity fatigue” to set in. But it has become a corporate priority again after shocking headlines in 2017 about sexual discrimination and harassment in Silicon Valley, Hollywood, professional sports and big media firms, which reminded the world that bad corporate culture is a serious business risk.
So companies will be crunching the numbers with ever-greater analytical sophistication to see if their recruitment and promotion policies are fair. New analytical tools from firms such as LinkedIn will help. A key test of this approach will be the fate of lawsuits brought against Google by America’s Labour Department and some female ex-employees based on allegations of sexual discrimination that Google says its talent analytics (long presumed to be cutting-edge) show are false. The search giant’s decision this summer to fire an employee who circulated a memo criticising its diversity efforts has made it a bellwether for the prospects of talent analytics.
“Unconscious bias” has become a hot topic in HR, so employers will deploy software designed to overcome it. A tool called Blendoor hides “data that are not relevant” (such as photos and names of applicants) and highlights data that are (skills, experience). Indexio scans job ads for unconsciously biased language. BetterWorks, a Silicon Valley startup, is testing software that scans conversations on Slack, using natural-language processing to detect sexist or other biased talk.
Cracking the code
Algorithms are only as good as those who code them. Some early talent-analytic software, based on studying the behaviour of recruitment firms, simply embedded existing unconscious biases, such as a preference for those who attended fancy colleges. But their quality is improving fast, and will continue to do so in 2018. Just as “Moneyball” data analytics undermined long-held beliefs in baseball about which players to recruit, so talent analytics will challenge traditional approaches to hiring and promotion in every business. Annual performance reviews will be transformed from uncomfortable exchanges of anecdotal flimflam into evidence-based conversations using real-time data. If, that is, annual reviews (and associated rankings of employees) survive at all. More organisations will begin to question their worth, although so far the decision by firms such as Microsoft and ge to replace reviews with continuous assessment has not been widely imitated. The preferred solution may turn out to be continuous assessment plus annual reviews.
Talent analytics will encourage a greater focus on how best to incentivise and reward employees. This will be one area in which HR people become active data scientists, coming up with experiments to see what works. When Google looked at its best workers, for example, solving its famous brain-teasers for applicants was not a good indicator of later success. Talent analytics may also encourage firms to give employees the right to take as much holiday as they wish, as some Silicon Valley firms already do. The data show that workers generally love this perk, though very few take more leave than would have been allowed under the old system.
Sceptics may worry about what misery might be wrought by a flawed boss using talent analytics. Clearly there are risks that these tools will give a boost to micro-managers, and that they will be used to squeeze every last ounce of effort out of relatively powerless workers in the gig economy. More likely, talent analytics will shed much needed empirical light on who is a good manager, and who is not. Overall, their greater use should make the world of work fairer, more inclusive and more meritrocratic.