When Algorithms Run the Workplace

Workers at the mercy of algorithmic bosses are frequently perplexed by their choices
article cover

Francis Scialabba

· 3 min read

Stay up to date on tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.

Who should you listen to: the traffic cop or your boss?

Drivers for and Meituan Dianping (China’s versions of DoorDash and UberEats) grapple with that question regularly. Several have said their manager—an app, not a human—directs them to speed or drive against traffic to meet delivery goals.

That’s an extreme outcome of algorithmic management, a practice Uber pioneered to efficiently manage its distributed driver workforce. It’s now a pillar of the global gig economy—and some traditional employers, like retailers, are embracing the tactic, too.

Under the algorithmic hood

What goes in: worker data (consumer reviews; a car’s maintenance history) and broad contextual data (weather; seasonal patterns).

What comes out: algorithmically generated decisions, often in the form of app-based nudges. AI techniques like machine learning help generate these prompts, which:

  • Incentivize (e.g., complete a few more rides to reach a bonus)
  • Assign/delegate (shifts could be changed due to projected spikes in foot traffic)
  • Penalize (worker ratings may be downgraded if they reject assignments)

Black-box bosses

It’s often unclear to algorithmically managed workers why an important performance rating changes, or why a certain timetable is set. That unpredictability can destabilize their schedules and finances.

Pushback: Last week, shoppers for Target-owned delivery service Shipt in the U.S. announced protests against a new “black-box algorithm pay structure,” which they say 1) slashes pay and 2) offers “zero transparency.” Shipt argues the change will make shoppers’ pay more accurate.

  • “Our main goal is a reversion back to the original, clear, and transparent commission-based pay model,” Willy Solis, a Shipt shopper in Dallas and lead organizer at Gig Workers Collective, told us.
  • Under the old model, Shipt shoppers could easily verify the accuracy of their paychecks.

And in early September, drivers for ridesharing company Ola took the company to court in the Netherlands to get transparency into its “fraud probability score,” which can impact driver pay.

Companies argue against algorithmic transparency on the basis that these systems are trade secrets. Data & Society researcher Alexandra Mateescu told us there’s another incentive at play too: maintaining asymmetry. According to her, companies may view transparency as unappealing because typically, “the way algorithmic control works is because workers don’t know how it’s functioning.”

Big picture: Not every employee who interacts with an algorithm views that computational sequence as their boss. But those who do answer to algorithms are often frustrated by a combination of surveillance, opaqueness, and algorithmic indifference.

Stay up to date on tech

Drones, automation, AI, and more. The technologies that will shape the future of business, all in one newsletter.