Gig Workers Aren't Burned Out From Hours. They're Burned Out From Never Clocking Out.
A driver drops off her last passenger of the night, closes the app, and starts making dinner. At 10:40 the phone buzzes. One star. "Rude, took the long way." She wasn't rude and she didn't take the long way — the app routed her — but none of that matters now, because she's not in her kitchen anymore. She's back in the car, replaying the ride from the pickup, looking for the moment she supposedly did something wrong. Dinner is still on the stove. She is not there for it.
The standard explanation for gig worker burnout is hours: too many shifts, too little pay per ride, no benefits cushioning the volatility. All true, all measurable, all beside the point for what actually breaks people in this kind of work. Hours are visible and bounded — you can count them, cap them, recover from them with a day off. What isn't bounded is the rating. It can land any time after the interaction that generated it, with no warning, and it does something specific to the nervous system that a bad hourly wage doesn't do: it stops you from switching off. A 2025 study in the Journal of Organizational Behavior found the exact mechanism, and it's more precise than "bad feedback feels bad." Negative customer feedback measurably impairs psychological detachment — the mental process of disengaging from work once the workday is over. That's the thesis worth sitting with. The problem isn't the workload. It's that the system has removed the off switch.
Why a One-Star Rating Outlives the Ride
The study is by Liang and colleagues, published in the Journal of Organizational Behavior in 2025 as "I Remember It All Too Well: Gig Workers' Psychological Detachment After Receiving Negative Customer Feedback and the Roles of Job Security and Handling Time," and it's a genuinely useful piece of research because it doesn't stop at "ratings cause stress." Everyone already knew that. What it isolates is the specific psychological function that negative feedback disrupts, which is detachment — the brain's normal process of setting work aside once the shift ends so the body can actually rest. That process isn't automatic. It's a cognitive shift that has to happen for recovery to occur, and the study found that a bad rating interrupts it directly, not as a side effect of general unpleasantness but as its own distinct mechanism.
Here's why the timing matters more than the rating itself. A rude customer during the shift is bad, but it's bounded — you finish the interaction, move to the next one, and the workday absorbs it the way workdays absorb friction. A rating that arrives after the shift has already ended does something structurally different. It reopens an interaction the worker had already filed away as finished. The driver in the kitchen wasn't dwelling on a bad night. She'd moved past it. The notification is what dragged her back, hours later, into a moment she no longer has any ability to fix or explain. That's not stress accumulating. That's stress being reinstated on a schedule the worker doesn't control.
What Psychological Detachment Actually Protects
Detachment isn't a soft concept. It's the specific recovery mechanism that lets your body come down from the physiological activation of working — lower cortisol, slower heart rate, a mind that stops rehearsing the day. Skip it consistently and you don't just feel more tired. You accumulate the kind of chronic activation that eventually shows up as the clinical stuff: exhaustion, cynicism, the whole burnout profile. Detachment is the thing standing between a hard day and a hard month.
What the Liang study makes clear is that algorithmic rating systems don't just add a stressor to the pile. They target the recovery mechanism itself. A worker can tolerate a genuinely difficult shift and still recover overnight, if the shift actually ends when the shift ends. What the rating system does is keep the shift open in every way that matters, for as long as it takes a rating to arrive, get processed, and get resolved in the worker's own head. For a lot of gig workers, that window is the entire evening. Some nights it's the next morning, when the app's weekly summary lands and reopens every low score from the past seven days at once.
This is worth separating from ordinary workplace anxiety, because the two get conflated constantly. Anticipatory dread about an upcoming task is forward-facing — your brain rehearsing something that hasn't happened yet. What a late-arriving rating produces is backward-facing and, worse, unresolvable. There's no future action that closes the loop, because the interaction that generated the score is already over and the worker usually never learns anything specific enough to fix. The nervous system stays alert to a threat that has already concluded, with no mechanism for declaring it finished.
Job Security and Handling Time Change How Hard It Hits
The study doesn't stop at showing the effect exists. It identifies two things that determine how badly a given worker gets hit, and both point at design, not personality.
The first is job security. Workers who feel more precarious, closer to the deactivation threshold and more dependent on the platform's good graces to keep working at all, show worse detachment impairment from the same negative rating than workers who feel secure. That's not a surprising finding in the abstract, but it's important in the specific: it means the platforms with the harshest deactivation policies are, by mechanism, producing the worst recovery outcomes in their own workforce. The threat isn't hypothetical stress. It's a felt, real risk to income, and the nervous system treats it accordingly.
The second is handling time — how much opportunity the worker actually had to respond to or address the negative feedback in the moment it happened. Workers who got some chance to course-correct, apologize, explain, or otherwise close the loop while the interaction was still live showed less detachment impairment than workers who found out about the problem only after the fact, with zero chance to do anything about it. This is the part that indicts the design most directly. The system that could give workers real-time signal instead saves it up and delivers it later, stripped of any opportunity to resolve it, which is precisely the condition that keeps the loop open the longest.
Put those two together and you get a clean design critique instead of a worker-blame narrative. It's not that some drivers are more resilient than others. It's that the workers with the least security and the least real-time recourse, usually the same workers, since precarity and platform opacity travel together, are structurally exposed to the worst version of a mechanism the platform itself controls.
So Actually — the Algorithm Doesn't Keep Office Hours
The instinct is to file this under gig work specifically, since Uber and DoorDash are the obvious examples. That undersells what the study actually points at. Rating-based algorithmic management has migrated well past gig platforms. Support agents get scored on CSAT after every ticket, often with the score visible days later. Creators watch engagement metrics update in real time on content they published hours or weeks ago. Salaried employees at logistics and retail companies increasingly work under systems that generate a live performance score from customer surveys, call monitoring, or productivity trackers, with the number following them home on the same phone that has their family group chat on it.
The mechanism Liang's team documented doesn't care whether the person receiving the score is a 1099 contractor or a full-time employee with benefits. It cares whether feedback about a completed interaction can arrive after the interaction is over, outside the worker's control, without a real chance to respond to it. Anywhere that's true, the same detachment failure is plausible. The gig economy didn't invent algorithmic non-detachment. It just built the version transparent enough to study first, because the ratings are the most visible, most explicit version of a mechanism that's quietly running underneath a lot more jobs than the word "gig" covers.
That reframe changes what the fix looks like, and it's not a wellness app. Telling a driver to "practice better boundaries" with a notification she can't turn off because turning it off risks deactivation is asking her to solve a design problem with willpower. The actual levers are the ones the study points at: batch feedback instead of real-time drip delivery, give workers a genuine window to respond before a score is finalized, and stop tying single data points to livelihood-level consequences. None of that requires new technology. It requires treating detachment as an engineering constraint on the system, not a personal skill the worker is failing to master.
The driver in the kitchen didn't need more resilience. She needed the rating to have waited until morning, or to have come with five minutes to explain herself, or to not carry the weight of her ability to keep working at all. The phone kept the shift open. Nothing in her was broken. The system just never told her it was over.