Here’s one from the too depressing to read before I’ve had my 5th cup of coffee file. Our friends at Bloomberg Business are reporting that an algorithm did a better job of selecting job candidates than real live human beings. Like you. And me.
The National Bureau of Economic Research (motto: 4% More Boring Than You Think We Are™) compared the tenure of more than 300,000 hires in low-skill service-sector jobs (like data entry and call center work) hired based on the algorithmic recommendations of a job test with individuals that humans hired. (The test asked the applicants a variety of questions and ran their responses through an algorithm, which then ranked the job candidates: green for high potential ones, yellow for moderate potential, and red for the lowest rated.)
- Greens stayed at the job 12 days longer than yellows, who stayed 17 days longer than reds.
- That may not sound like much, however, according to the article, the median duration of employees in these jobs is only about three months to begin with.
- The more managers deviated from the test’s recommendations, the less likely candidates were to stay in their jobs.
An example: when recruiters hired a yellow instead of available greens, who were subsequently hired to fill other open positions, those greens stayed at the jobs about 8% longer.
- The study also suggests that the individuals hired by humans were no more and in some cases, less productive that the algorithm’s recommended hires.
The actual study is available here - for $5.
It would be interesting to see if these results could be replicated for hires in more skilled industries. Until then, there’s only one sensible response to this automated takeover of the HR industry, and it ain’t another cup of coffee.