Gig Workers Can Fight the Algorithm. Only the Ones Who Can Afford To.

Based on the research ofJiang, Wiener, Benlian, Adam and Mähring, "How Platform Workers Contest Algorithmic Management: Theorizing the Dynamics of Algoactivistic Practices," Information Systems Research, 2026

Two Uber drivers can be throttled by the exact same dispatch algorithm, the exact same rating threshold, the exact same deactivation risk, and end up in completely different places, because resisting an algorithm turns out to cost something, and only one of them can pay it.

That is the claim at the center of a new paper in Information Systems Research by Jennifer Jiang, Martin Wiener, Alexander Benlian, Martin Adam, and Magnus Mähring. Using a computer-assisted grounded theory approach, topic modeling paired with qualitative coding across multiple sources of Uber driver data, the authors build a process theory of what they call algoactivism: worker resistance to algorithmic management. Their finding complicates two assumptions that most coverage of gig work, including a lot of academic work, has taken for granted. The first is that resisting the algorithm is something any affected worker can simply choose to do. The second is that resistance is one big undifferentiated pushback against "the algorithm" as a whole. The paper's alternative, drawn from labor process theory's idea of a contested terrain, is that algoactivism emerges through situated, reflective, and resource-dependent processes, and that workers pursue three different kinds of resourcing, described in the paper as algorithm, market, and voice resourcing, to mount practices the authors group as self-optimizing, distancing, or confronting. Whether a driver can do any of that at all is not a function of how badly the algorithm is squeezing them. It is a function of what they had walking in the door.

Who Gets to Game the Algorithm

The clearest real-world illustration of resource-rich algoactivism predates the paper by nearly a decade, but it maps onto the same logic. In a study covered by PBS NewsHour in 2017, researchers Mareike Möhlmann and Ola Henfridsson of Warwick Business School and Lior Zalmanson of New York University analyzed driver posts on the forum Uberpeople.net and documented drivers organizing coordinated mass "switch-offs," logging out of the app together in a given area so Uber's surge algorithm read a supply shortage and raised prices. One forum post the researchers quoted read simply: "Guys stay logged off until surge." The same drivers described gaming UberPOOL by accepting the first passenger, then ignoring or refusing detour requests, so they collected the higher UberPOOL commission without doing the unprofitable extra pickups the algorithm was routing to them. Uber's own response at the time was that the behavior was "neither widespread nor permissible."

What that tactic actually required is the interesting part. A driver needed to already be active in an online community, understand enough about how the dispatch algorithm interpreted supply signals to know that a synchronized log-off would move it, and be able to absorb the lost fares during the window when the group was intentionally offline. That is time, technical literacy, and a network, three resources at once, deployed toward a narrow, specific target: the surge-pricing mechanism, not "Uber" as an abstraction. It is exactly the kind of situated, targeted, resource-dependent practice the ISR paper theorizes, not blanket defiance but a precise intervention aimed at one lever of the contested terrain.

Deactivated and Out of Options

Now put a driver on the other side of that resource line. In February 2023, the Asian Law Caucus and Rideshare Drivers United published survey findings from 810 California Uber and Lyft drivers in a report titled "Fired by an App." Two-thirds of all surveyed drivers had experienced permanent or temporary deactivation. The deactivation was not evenly distributed: 69 percent of drivers of color reported some form of deactivation compared with 57 percent of white drivers, and 86 percent of drivers who did not speak English reported deactivation compared with 61 percent of drivers fluent in English. Thirty percent of deactivated drivers said they were given no explanation at all for why they had been cut off. Only 3 percent of drivers who filed a complaint said Uber or Lyft actually investigated and resolved it.

Nicole Moore, president of Rideshare Drivers United, described what contesting a deactivation actually takes in practice: relentlessly calling, emailing, and showing up at a hub office and hoping to get lucky, a process she said amounts to "wearing people down until they give up." That is a description of voice resourcing collapsing in real time. Appealing a deactivation requires English fluency, the free hours to sit on hold or drive to a hub, and enough of a financial cushion to survive the income gap while the appeal drags on. Eighty-one percent of the surveyed drivers said the apps were their primary income, and 18 percent of deactivated drivers had lost their car as a result. A driver deactivated by the same opaque rating threshold as someone in the Uberpeople.net forums, but without the language, the spare hours, or the community, is not choosing to accept it quietly. They are out of resourcing.

A Policy Win the Platform Re-Engineers Around

The lockouts are the platform doing to a whole workforce what the individual resource-poor driver cannot do to the platform: quietly rewriting the terrain after losing a round on it. A collective, resourced win, secured through years of political organizing that most individual drivers could never replicate alone, still had to be defended a second time against an algorithmic countermove. That is the paper's contested-terrain frame playing out in public, in real dollars, over a real regulation.

The Solidarity the System Is Built to Break

Resistance is not a right every worker holds by default; it is a resource some workers have and others were never given.

Amazon's delivery and warehouse operations show what happens when the algorithm is designed to choke off the community resourcing that made the Uber forum tactics possible in the first place. Amazon has installed Netradyne "Driveri" camera systems in delivery vans, with cameras facing both outward and at the driver, running continuously and feeding a productivity and safety score. Warehouse workers are tracked against a "time off task" metric, where time spent interacting with coworkers away from the assigned task counts against them; more than 30 minutes logged as off-task in a day can trigger a written warning. A peer-reviewed study of the 2021 Bessemer, Alabama union campaign, published in 2025 and titled "Weaponizing the Workplace," found that Amazon's algorithmic management shaped its anti-union campaign there, including at least one instance of the time-off-task tracking being quietly relaxed to try to win workers over during the vote.

A system that penalizes the minutes workers spend talking to each other is not incidentally suppressing solidarity. It is removing, by design, the exact resource, community, that the Uberpeople.net drivers needed to coordinate a switch-off. Workers under that kind of monitoring cannot easily build the informal networks that turn isolated frustration into a shared, targeted tactic. The algorithm is not just setting the pay and the pace; it is also rationing the raw material that collective algoactivism runs on.

The Uneven Terrain Any Platform Should Expect

Put these four cases together and the pattern is not that some workers resist and others do not. It is that resistance requires inputs, algorithm literacy, market alternatives, and a voice channel that actually gets heard, and those inputs are distributed exactly as unevenly as everything else in the gig economy. The Uberpeople.net drivers had community and technical fluency. The California drivers surveyed by the Asian Law Caucus mostly did not have English fluency, spare hours, or a functioning appeal channel, and it was disproportionately drivers of color and non-English speakers who paid for the gap. The New York drivers who won a minimum pay standard had years of organized political capital behind them, and even that had to be defended again against a scheduling countermove. Amazon's monitoring shows a platform actively managing the supply of the community resource that makes any of this possible at all.

The operating implication is not comforting. A platform that assumes worker feedback, forum chatter, or appeal volume represents the true scale of a problem is only hearing from the workers who could afford to speak. The workers most exposed to a bad algorithmic call, the ones with the least time, the fewest alternatives, the weakest community ties, are structurally the least likely to generate a signal you can see. If you are counting complaints as your ground truth, you are measuring resourcing, not harm.