Wage Theft by Design: How Opaque Pay Algorithms Shortchange Gig Workers

Ask a delivery rider or a cab driver in any Indian city what they earned last week, and you will get an honest shrug. Not because they are careless with money — they are the most careful people I know — but because gig worker pay in India has become genuinely unknowable. The number that lands in their bank account is the output of an algorithm they cannot see, running rules they were never told, on inputs they cannot check. They accepted a ride, they finished it, and a figure appeared. Whether it was the right figure is a question the system is designed to never let them answer.

I have spent a long time studying how power hides inside things that look neutral — apps, dashboards, pricing screens. And when I look at how gig platforms in India actually set pay, I see one of the cleanest examples I know of a familiar pattern: take something a person used to be able to see and verify, wrap it in software, and quietly move the machinery where they can no longer reach it. It is the same capture story I trace in how technology gets captured — except here the thing being captured is the wage itself.

The wage nobody can see

A generation ago, piece-rate work was crude but legible. You were told the rate per delivery, you counted your deliveries, you multiplied. If the contractor cheated you, you could point to the arithmetic. The fight was ugly, but at least both sides were arguing about the same visible number.

The gig platform quietly removed the visible number. Today a rider is quoted a fare or a payout that already bundles a base rate, a distance component, a time component, some surge or incentive multiplier, and a set of deductions — all computed instantly, none of it broken down in a way the worker can reconstruct. Workers report being shown a payout for a trip without ever seeing the formula that produced it. They cannot tell you what the per-kilometre rate was that morning, because the platform does not tell them, and because — many describe — it does not seem to be the same from one hour to the next.

This is the quiet trick. Pay stopped being a rate you were owed and became a result you were handed. And a result you cannot audit is a result you cannot dispute.

The specific mechanisms

When riders and drivers describe the daily reality, the same handful of mechanisms come up again and again. None of them is, on its own, obviously theft. That is exactly what makes them effective.

The moving target

Incentive schemes are often framed as milestones: complete so many trips, or log so many hours, and unlock a bonus. Workers across Indian cities describe a recurring pattern where the target seems to recede as they approach it — the number of trips required creeps up, the window tightens, the bonus that was within reach at trip fourteen has quietly become trip eighteen by the time they get there. Because the terms live inside an app that updates silently, there is rarely a screenshot war to be had. The goalpost simply is where it is now, and nobody remembers agreeing to move it.

The incentive that evaporates

Closely related is the bonus that dissolves as you close in on it. Riders report grinding through a punishing day chasing a weekly incentive, only to find the qualifying rules interpreted in a way that disqualifies them at the last mile — a cancelled order that was not their fault, a shift that ended minutes short, a rating dip they cannot trace. The reward was real enough to work for and conditional enough to withhold. You cannot prove it was engineered that way. You also cannot prove it wasn’t.

The per-trip rate that drifts down

Many longtime workers describe the same slow erosion: the effective rate per trip today is lower than it was a year or two ago, even as fuel and living costs climbed. There is rarely an announcement. No letter arrives saying “your rate is being cut.” The base simply drifts, incentives that once padded earnings thin out, and the worker runs harder to hold the same weekly total. Because there was never a published rate to begin with, there is nothing concrete to point at and say that is what changed.

Pay stopped being a rate you were owed and became a result you were handed. And a result you cannot audit is a result you cannot dispute.

The deductions and surge nobody explains

Then there are the line items that appear without narration: commissions, service fees, adjustments, taxes, insurance charges. Workers report payouts arriving lighter than expected with a deduction they cannot decode. On the other side sits surge — the multiplier that supposedly rewards driving in high-demand hours. But surge is calculated by the platform, shown to the customer and the worker in different forms, and settled in a way the worker cannot reconcile. Did the surge the customer paid actually reach the driver? He has no way to check. He sees his slice; he never sees the pie.

Why “wage theft by design” fits

I choose those words carefully, because they are strong. “Wage theft” usually means an employer failing to pay a legally owed wage — an overtime hour skipped, a shift shaved. What happens on gig platforms is subtler, and in some ways worse, because it is built into the architecture rather than committed as a discrete act.

No single instance looks like theft. A rider is short by a small amount on one trip — a rounding here, a deduction there, an incentive that didn’t trigger. It is too small to fight, too ambiguous to prove, too cheap to be worth a day off the road arguing about. Individually, every loss is deniable. The platform can always say it was the algorithm, the demand curve, the rules you agreed to.

But multiply that small, deniable, unverifiable loss across millions of workers and hundreds of millions of trips, and it stops being noise and becomes a business model. This is the pattern I keep returning to in my own work: the most durable extraction is never the dramatic heist. It is the small loss, made deniable, repeated at scale. Nobody can point to the moment they were robbed, because there was no moment — there was a system, quietly tilted, running millions of times a day. That is what “by design” means. The theft is not an event. It is the shape of the machine.

The asymmetry is the whole point

Strip away the details and what remains is a single, brutal imbalance of information. The platform sees everything. It knows the true demand in every neighbourhood, the exact rate it is paying every worker, the full fare every customer pays, the precise gap between the two, the effect of every incentive tweak on driver behaviour, tested and measured in real time. It runs experiments on its own workforce and reads the results on a dashboard.

The worker sees almost nothing. One payout at a time, no formula, no history he can export, no way to compare his rate to yesterday’s or to the rider next to him. He cannot see the demand map that determines his surge. He cannot see the fare the customer paid. He is asked to make his living inside a game where only the other side can read the board.

That asymmetry is not a bug the platforms haven’t gotten around to fixing. It is the source of their pricing power. As long as the worker cannot verify his pay, the platform can set it wherever the labour market will bear — and adjust it, silently, whenever it wants. This is why I treat pay as a distinct problem from the broader question of algorithmic management — of being bossed around by an app that assigns, ranks, and disciplines you. The management story is about control over your day. The pay story is about control over the number, and the number is where the money quietly moves.

He is asked to make his living inside a game where only the other side can read the board.

What could actually change it

The asymmetry was built. That means it can be dismantled. None of the fixes is exotic; they are just unwelcome to whoever benefits from the fog.

Pay transparency, as a right

The single most powerful reform is also the most obvious: give the worker the formula. A legally enforceable right to an itemised breakdown of every payout — base, distance, time, surge, each incentive, each deduction — plus the fare the customer actually paid, and an exportable history the worker owns. Transparency alone does not raise the rate. But it converts an unverifiable loss into a verifiable one, and everything else — disputes, bargaining, regulation — becomes possible only once the number can be checked. India’s move to bring gig and platform workers into social-security frameworks is a start; pay legibility belongs in the same conversation, because a benefit you cannot calculate is a benefit you cannot claim.

Minimum standards with a floor under the algorithm

Transparency shows you the number; minimum standards stop it from falling below dignity. A guaranteed floor per hour of logged, available time — not just per completed trip — protects workers from bearing the full cost of the platform’s demand miscalculations. Rules against retroactively changing incentive terms mid-cycle would end the moving-target game. The point is not to freeze a rate; it is to force the machine to operate above a line that a human being drew, in public.

Collective organising

No floor gets legislated without pressure, and gig workers across India have been building exactly that — unions, associations, and coordinated log-offs that make the invisible workforce briefly, powerfully visible. The obstacle is structural: these workers are dispersed, classified as “partners” rather than employees, and set into subtle competition with one another by the very rating and incentive systems that pay them. But dispersed people have won before. I keep coming back to how India’s farmers won — a scattered, supposedly unorganisable mass that held together long enough to force a reversal. The lesson isn’t the tactic. It’s that concentrated power built on an opaque system is more fragile than it looks the moment the people it depends on refuse, together.

Owning the algorithm outright

The deepest fix is to change who the machine answers to. If the pricing algorithm is opaque because it serves shareholders whose returns come from the gap between fare and payout, then the way to make it serve workers is to let workers own it. Driver-run and cooperative platforms — where the people doing the work hold the software and set the rules — turn the algorithm from a black box pointed at them into a tool they can open and audit. I’ve written about the promise and the difficulty of this in platform cooperatives. It is hard to build and harder to scale against venture-funded incumbents. But it is the only model where transparency isn’t a concession the worker has to extract — it’s the default, because there’s no one on the other side of the information wall with a reason to keep it up.

The number is not neutral

The thing I most want a rider or a driver to take from this is simple: the confusion you feel about your own pay is not your failure to understand. It is the intended experience. A system that wanted you to verify your wage would show you the formula. This one doesn’t, and that choice was made by someone, for a reason.

Pay was once a number two parties argued over in the open. Software didn’t make it fairer; it made it invisible, and moved it behind a wall only one side can see over. Naming that honestly — wage theft by design, small deniable losses multiplied across millions — is the first step to insisting the wall come down. The worker is not asking for charity. He is asking to see the arithmetic of his own life. That used to be the baseline. Getting it back is the whole fight.