Wage-Setting Algorithms Are an Abomination
On the basic right to know what you will be paid.
It’s a big day for Uber. The company is about to be added to the S&P 500 index, and the stock price is rising as a result. It is a mark of validation for a company that has long been dogged by accusations that it is a big scheme to benefit earlier investors by ripping off later, dumber investors, and destroying public transportation systems in the process. Now, even many of the more dumb mid-lifespan institutional investors in this company will have a better shot at recovering their money! That’s nice. The truth is, though, that the central elements of Uber’s business model have already established themselves deeply in the corporate world. Let’s take a closer look at one of those elements—the one that has, I think, the greatest potential for making the lives of working people worse, permanently.
If you are reading this, I am going to assume that you are already familiar with the ravages of the con job that is the “gig economy.” (In brief, the “gig economy” is a simple mechanism of eradicating full time jobs, erasing to the greatest extent possible the existence of the employer-employee relationship and the rights and protections and benefits it confers on workers, and getting the same amount of work at a cheaper cost, and funneling the money that would have once gone to workers to investors instead.) No need to REHASH every last awful thing Uber has ever done here. From a consumer’s point of view, one of the things Uber has done is to train us all on the idea that the price of ride is not fixed, but instead fluctuates from moment to moment, based on a secret formula. Prices of goods and services have always fluctuated based on supply and demand—if the orange crop freezes, the price of OJ goes up, and if Old Navy orders too many hoodies, they go on sale—but the algorithmic pricing of Uber rides is much more sophisticated than this crude measure. The usual explanation for their system is “surge pricing,” which implies that when demand rises, the app raises prices, which attracts more drivers, thereby meeting demand, in a system that works for everyone. But the truth is that the pricing algorithm is a black box. The company can and will set the maximum possible price based on any factors it wants to, as we were all reminded when a report earlier this year said that Uber charges more when your phone’s battery is lower. Supply and demand are factors, yes, but they are more of a cover story than a full explanation. What is happening should be understood as a technically advanced method of maximizing profits.
The other side of this coin, and one that consumers think less about, is Uber’s practice of using an algorithm to determine wages as well. As independent contractors, Uber’s drivers cannot unionize, and they cannot negotiate a contract standardizing their pay across different times and places. Their powerlessness is at the heart of Uber’s business model. In exactly the same way that the company builds an algorithm designed to determine the highest amount that you would be willing to pay for a ride at the moment you request one, it also has an equal and opposite algorithm designed to determine the lowest price that they can pay a driver for giving you that ride. The academic name for this practice is “algorithmic wage discrimination.” It is the practice of using modern technology to weaponize the desperation of working people against them in the most effective possible way.
The leading scholarly voice on algorithmic wage discrimination is UC-Irvine law professor Veena Dubal, who has been sounding the alarm for years about the disturbing implications of the practice. Dubal’s latest work is an article in the Columbia Law Review that lays out the problem as one rooted in “information asymmetry,” where “companies can calculate the exact wage rates necessary to incentivize desired behaviors, while workers can only guess how firms determine their wages.” This is not a story of advancing efficiency, the benefits of which accrue to all of us; it is one of an advance in the tools of exploitation. As Dubal writes:
Today, facilitated by independent contractor status, algorithmic wage discrimination turns the basic logic of scientific management on its head. Instead of using data and automation technologies to increase productivity by enabling workers to work more efficiently in a shorter period (to decrease labor overhead), on-demand companies like Uber and Amazon use data extracted from labor, along with insights from behavioral science, to engineer systems in which workers are less productive (they perform the same amount of work over longer hours) and receive lower wages, thereby maintaining a large labor supply while simultaneously keeping labor overhead low.
Notice that a side benefit of this system for the companies is that they can point to the algorithm itself to justify their decision to exploit you. Gee, we would love to pay you more and all, but this highly advanced black box has determined that we must pay you less. It is a sly aid to the (already highly advanced) corporate project of avoiding responsibility for every negative consequence of their business practices. Interviews with drivers reveal that the sheer unpredictably of this wage system transforms work into something more akin to gambling. Like slot machine players always wondering if the next spin will be the lucky one, workers are put in a position of being incentivized to constantly stay available, in the event that the fluctuating wage level happens to rise at any given moment.
[Algorithmic] scientific management deployed by on-demand firms is opaque—and purposefully so. Because of this opacity, workers cannot trust the firm’s or their own market forecasts, nor can they rely on the firm-created incentive structures (or wage manipulators). The time that they must labor to meet their income targets—the primary way in which workers in my research structured their work—is ever changing. Through this process, hard work and long hours become disconnected from any certainty of economic security. Thus, algorithmic wage discrimination, by keeping workers in a state of deep uncertainty, creates profoundly precarious working conditions and wages that belie long-held norms of a moral economy of work.
Dubal’s entire paper is eye-opening and terrifying in equal measure. I think that most people would agree that this practice is scary, creepy, offensive. But why? Given the fact that we so readily accept the constant fluctuation of prices, why does the similar treatment of wages feel so unfair? It is because this kind of payment system completely destroys what most of us take for granted as a bedrock part of work: The idea that you know what you will be paid. Whether you are paid a full-time salary, or you work at a shitty hourly job, or you are self-employed and paid for each thing you do, you know in each case what you stand to make. Your job may suck, and your boss may be awful, and your paycheck may be a travesty in many ways, but at the very least, once you know how much you’ll be working, you know what that paycheck will be. You can budget. You can plan. When companies take that away, they rob workers of something that goes beyond the normal strictures of poverty; they make it impossible to determine the boundaries of your own working life.
This is the type of thing that demands a very sobering look in the mirror for those of us who care about labor power. Black boxes are inimical to worker power. Transparency is an absolute necessity to determining if you are being treated fairly. (This is why when you negotiate a union contract and the boss says “We can’t afford it,” you demand to see the books.) The algorithms that Uber and other companies use are, by their nature, black boxes. They are not available to be opened up and perused. So, as a starting point, we need to recognize that there is no path to worker empowerment and fair treatment that involves merely poking around in the black box, twisting the dials and hoping that the algorithm improves. If we cannot monitor the formula that determines wages, and we cannot control the formula that determine wages, then we must find another way to protect workers.
This is one of the underlying reasons why it is important to remember that the entire “gig economy” itself should be viewed as an unacceptable assault on the social contract. Some elements of organized labor imagine that they can acquiesce to the general structure of gig work, but improve it enough to make it worthwhile. That is a mistake. These companies will never cede enough control to workers to allow us to feel comfortable that we have achieved fairness and stability. Why do I say that? Because the business model of gig companies is dependent on that not happening! There is a difference between being a realist and being a patsy.
Yes, these companies are big and powerful. We cannot wave a magic wand and delete Uber’s existence. But we, meaning “the combined institutions of organized labor,” can operate with the conviction that the ability of workers to know in advance what they will be paid is a non-negotiable principle of fair work, and that to the extent that gig companies do not do this, they are obligated to change. Which is to say: algorithmic wage discrimination should not exist. Seeking to mitigate it rather than end it is a sucker’s game. Some things need to be red lines, and this is one of them. Indeed, Veena Dubal’s own recommendation is for “a statutory or regulatory nonwaivable ban on algorithmic wage discrimination, including, but not limited to, a ban on compensation through digitalized piece pay.”
A big part of forging a minimally fair set of standards for workers in the age of algorithms is being humble. If you think that you can outsmart or outwork an algorithm designed to exploit you, you are wrong. If you think that long-established workplace practices like “set hourly pay” and “a formal employer-employee relationship” can be tossed out and updated to the benefit of everyone, you are wrong. If you think there is a way to come to a mutually beneficial arrangement with a corporation that creates profits by arbitraging away the portion of revenues that formerly existed to provide a livable wage to employees, you are wrong. It’s like playing tag with a tiger. You’re going to find out, eventually, that it was a bad idea. Don’t allow shiny new technologies to make you forget what the labor movement stands for. If Uber was so great, the executives would be the drivers.
More
I have an essay in the new issue of The Progressive magazine about how to scale up labor organizing, and why the labor movement has failed to do so. As far as I can tell the piece is only in print, and you can get it here. Also relevant to today’s discussion: The Life and Death Stakes of Labor Power, and We Don’t Work For You If We Don’t Work For You.
Hachette, the publisher of my upcoming book “The Hammer,” is doing a giveaway of galley copies on Goodreads. The idea is you sign up and then they give away 50 copies and then you leave a review on Goodreads, presumably along the lines of “I love this book, which I received for free.” So check that out. Please keep in mind that the galley copies are not the final edit of the book, which means A) anything in there that you dislike, well uh… that’s not final, and B) you should still buy an attractive hardcover copy when it is published on February 13, in order to read and appreciate the final version. You can preorder the book (or the audio book, read by me) right here, or wherever books are sold.
Let me take a moment here to say “thank you” to everyone who has subscribed to this publication, How Things Work. This site is an experiment, by me, to discover whether it is possible to make a living as an independent writer today, or whether we should all give up hope and accept that journalism going forward will only be produced by AI or by Thomas Friedman. This publication exists purely thanks to the support of paid subscribers. If you would like to enable it to exist in our unforgiving world, please consider becoming a paid subscriber today. This is socialism in action, my friends. We believe in it!
Recently I had a conference and took an Uber from the Las Vegas airport to my hotel. The cost of this fairly short ride was around $62 without tip. I asked the driver how much he was making, and he said more than normal, which was under $10. I just about fell out. I did what I could giving him a 30% tip on my company’s dime but WTF? So, thank you for delineating this because I understand better now
The ILO formally opposes debt peonage as a form of slavery, and it kinda seems like algorithmic wage discrimination is the same thing. When an employer capriciously changes the value it assigns to a worker’s labor, it is irrelevant whether that value is “paid” as a credit against a debt or as a fluctuating and unpredictable “wage.” The value transfer is the same in both cases and, more importantly, uncorrelated to the cost of the work, in terms of an individual’s time, physical effort, creativity, education/skill, risk to health and well-being, etc. And, in both cases, the accumulated power of the employer has become so large that it eats the worker’s freedom. Which is basically slavery. Surely the based Veena Dubal would agree.