What people are saying
Nobody has said anything yet.
Brand New Thoughts/Tech and Labor
AI Is Learning the Lowest Wage You'll Accept
Algorithmic pay can turn worker behavior into data and use opaque systems to vary what different people are offered for similar work.
By ElleJanelle|July 31, 2026|4 min read
What If the Price Being Personalized Is Your Paycheck?
Consumers are already familiar with prices that move.
Airfare changes. Hotel rates change. Ride prices surge. Online stores test offers. Delivery fees rise and fall.
But during Sophisticated Unshitification, Brand New Thoughts ran into a more unsettling version of the same logic: what happens when the variable being optimized is not what a company charges you, but what a company pays you?
ElleJanelle was working through questions about capitalism, bargaining power and exploitation when the conversation reached scholar Veena Dubal's term algorithmic wage discrimination. The possibility immediately sharpened the earlier debate. If a platform can observe worker behavior and vary offers, the important question is no longer only "What is this job worth?"
It can become:
What is the least this particular worker might accept?
When the algorithm knows more about your willingness to work than you know about everybody else's pay, bargaining stops being symmetrical.
Brand New Thoughts
What Algorithmic Wage Discrimination Means
In her Columbia Law Review article, Veena Dubal defines algorithmic wage discrimination as a system in which platform workers performing similar work can receive different pay based on opaque, changing formulas and granular data.
That is different from an ordinary posted wage.
If a restaurant advertises $22 an hour for a position, applicants can compare the number with another restaurant's offer. Workers can talk to one another about whether the rate is fair. The wage is visible enough to become a shared fact.
Personalized algorithmic pay can fragment that shared fact.
Worker A receives one offer.
Worker B receives another.
Worker C sees something else.
Each person may know what they were offered without knowing the distribution of offers around them.
The important word is: can
The provocative headline of this article should not be mistaken for a claim that every gig-work platform literally possesses a secret number labeled YOUR MINIMUM WAGE beside each worker's name.
That is not what the evidence establishes.
The deeper concern is that modern platforms can collect large amounts of behavioral data, vary pay and work allocation, and learn from how workers respond.
That creates the conditions for a company to optimize offers around individual behavior in ways that traditional wage setting could not.
The uncertainty is part of the problem. Workers cannot meaningfully evaluate a system they cannot inspect.
Information Is Bargaining Power
Consider two negotiations.
In the first, both sides know roughly what comparable workers earn, what the job usually pays and what alternatives are available.
In the second, the company has data on when you work, which offers you reject, which offers you accept, where you tend to drive, how quickly you respond and perhaps how your behavior changes over time. You receive an offer, but you cannot see the formula that produced it or what other workers were offered for comparable work.
Those negotiations may both technically contain a "choice."
They do not contain equal information.
That distinction is central to the argument ElleJanelle was trying to unshittify on the live. Saying "the worker accepted the offer" tells us something important. It does not tell us everything important about the structure that produced the offer.
What worker data can reveal
Algorithmic management can potentially infer patterns such as:
- which kinds of jobs a worker regularly accepts;
- which offers a worker declines;
- when a worker is usually available;
- how long a worker waits before accepting work;
- where demand is high or low;
- how workers respond when compensation changes.
None of those data points is magical on its own.
Combined at scale, however, they can give a platform a much stronger picture of worker behavior than any individual worker has of the platform's behavior.
That is why the fight over algorithmic wages is also a fight over transparency.
An old incentive with new precision
Technology did not invent the desire to minimize labor costs.
Employers have always had incentives to understand how little they can pay while still attracting enough workers.
What changes with algorithmic management is the precision, speed and opacity with which that incentive can operate.
A manager might once have set one wage for a shift.
A platform can potentially make enormous numbers of individualized decisions continuously.
That does not automatically make every decision exploitative. It does make older assumptions about a single "market wage" harder to apply.
Five questions workers could ask
A useful policy debate begins with questions more specific than "AI bad" or "markets good."
Workers could reasonably ask:
- Can I know the basis on which my offer was calculated?
- Can two workers compare what they were offered for comparable work?
- Is there a predictable wage floor after expenses?
- Can a worker challenge an automated decision?
- Can researchers or regulators audit whether the system creates systematic disparities?
Those questions do not require eliminating flexible work.
They require deciding whether flexibility should also mean unreviewable, individualized opacity.
The Algorithm Does Not Need to Read Your Mind
It only needs data.
That may be the most important lesson from the Brand New Thoughts discussion.
A system does not have to know that you are desperate, tired, behind on rent or trying to reach a weekly earnings goal in the human sense of "knowing."
It only has to observe patterns that correlate with what you accept.
Once pay becomes a continuously optimized prediction, the old question "What is this labor worth?" risks being replaced by a more profitable one:
What can we get this person to say yes to?
Watch the episode
Should two workers see the same pay?
Add to the DiscussionIf an algorithm offers different people different amounts for comparable work, what information should workers have a right to see?