Podcast brief
Gig Work, Algorithms, and the Ethics of Pay
Host Barry Lam hears from Shipt shoppers Heidi and Willy Solis, data scientist Dan Calacci, organizer Drew Ambrogi, and philosopher Daniel Halliday on whether opaque pay algorithms treat gig workers unjustly.
A Stoa brief of an episode from Hi-Phi Nation
The Problem with Gig Work
Listen to the original episode
The episode belongs to its makers. This page summarizes its arguments in Stoa's words and points to the moments where they are made. Brief updated .
The brief
This conversation argues that gig companies like Shipt commit a distinct injustice when they switch workers to opaque, algorithmically determined pay, even if the company can claim the new system pays more overall. The wrong is not just about the paycheck. It is about what workers are allowed to know and choose.
The story starts with two Shipt shoppers, Heidi and Willy Solis, who saw their pay drop when Shipt replaced a transparent commission formula with an algorithm that supposedly priced each job by estimated effort. Unable to get a straight answer from the company, they worked with organizer Drew Ambrogi and data scientist Dan Calacci to crowdsource pay screenshots from other shoppers. The resulting analysis found that although Shipt paid out more in aggregate under the new system, 41 percent of individual workers earned less, a share that grew to 60 percent over time, while average payouts fell 15 percent. Calacci likens this to Taylorist scientific management: setting pay by a prototypical worker's estimated time rather than by any individual's actual circumstances, whether that is a bad leg or an older, less efficient car.
Political philosopher Daniel Halliday supplies the moral frame. He distinguishes what employees and freelancers each trade away: employees give up freedoms like setting their price and hours for a secure wage, while freelancers keep those freedoms but accept market risk. Gig companies, he argues, impose employee-style controls, vehicle standards, dress codes, deactivation, while still classifying workers as freelancers, so they get the burdens of both categories and the protections of neither. He treats withholding basic information, like a job's pay before it's accepted, as its own wrong, since no party has a legitimate interest in hiding it.
The strongest complication comes from Halliday himself: he considers whether high enough pay could make the loss of freelancer freedoms unobjectionable, which suggests the deeper problem is precarity and low pay, not classification as such. Shipt never changed course, though several jurisdictions have since sued over misclassification. Calacci closes by pointing to worker-owned cooperative platforms as one alternative: pay algorithms designed with workers rather than imposed on them.
Strongest arguments
Shipt's algorithm switch hid pay information from workers
4:01Shipt moved from a transparent commission model, 7.5 percent of order total plus five dollars, to an opaque algorithm supposedly based on effort. Workers could no longer predict what any given job would pay before accepting it, turning every shop into a gamble on whether it would be worth their time and gas money.
Freelancers and employees trade different freedoms for different securities
8:10Daniel Halliday argues that being an employee means giving up freedoms like setting your price, hours, and dress in exchange for a secure paycheck, while freelancers keep those freedoms but lack security and cannot be fired, only lose reputation. Gig companies impose employee-like restrictions such as vehicle age requirements and deactivation policies while still classifying workers as freelancers, denying them employee protections.
Data-driven analysis found a large minority of workers paid less under the new algorithm
24:16Dan Calacci's analysis of crowdsourced pay screenshots found that while Shipt paid out more in aggregate under V2, 41 percent of individual workers earned less than before, and that share rose to 60 percent by the end of the data collection period, with average payouts falling 15 percent over time.
Averaged algorithmic pay resembles old Taylorist scientific management
31:58Calacci compares Shipt's practice of estimating an average shopping time and paying a flat rate accordingly to early twentieth century time studies in factories, where management measured a prototypical worker's task time to set wages as low as possible. He argues that companies now have enough individualized data, such as car mileage and driving habits, to calculate a genuinely fair wage rather than relying on an average that penalizes anyone slower than typical.
Withholding pay information from workers is a distinct moral wrong
38:40Halliday argues that beyond low pay, the deliberate withholding of information, such as not telling a driver the destination or not telling a shopper what a job will pay, is itself an injustice, since no legitimate party has an evident interest in that information being hidden from the worker.
Higher pay could in principle offset lost freelancer freedoms, but the real issue is low pay
41:40Halliday considers whether sufficiently high compensation could justify taking away freelancer freedoms, concluding that if Uber drivers were paid very well, the loss of freedoms might not bother us. This suggests the deeper problem is not the freelancer-employee classification itself but that gig work is low-paid and precarious.
Worker-designed, participatory algorithms as a future alternative
49:02Calacci proposes that algorithms should be co-designed with workers rather than designed unilaterally by company engineers, pointing to a worker-owned cooperative delivery platform in Colorado that incorporates the interests of workers, restaurants, and customers into its algorithm and pay design as a model for the future.
Disagreements
Whether flat average pay or individualized pay is fairer
31:33Shipt frames its new algorithm as fairer because it pays based on estimated effort rather than commission, avoiding the situation where cheap heavy items pay less than expensive light ones. Dan Calacci disagrees, arguing that paying everyone the same average rate regardless of individual circumstances such as a bad leg or an older, less fuel-efficient car is not truly fair and instead resembles Taylorist wage-suppression tactics.
Philosophers and works discussed
- Daniel Halliday
Questions this episode answers
-
What moral difference is there between being an employee and a freelancer?
8:10Being an employee means trading freedoms, such as setting your price, hours, and dress, for the security of a fixed paycheck. Being a freelancer means keeping those freedoms but accepting market precariousness, and a freelancer cannot simply be fired the way an employee can, only lose reputation and future business.
-
Why do gig economy companies like Uber and Shipt treat workers unfairly according to Daniel Halliday?
8:00Halliday argues these companies impose employee-like controls, such as vehicle standards, dress requirements, and deactivation for poor ratings, while still classifying workers as freelancers, denying them the security and benefits of employment. This lets companies take the burdens of freelance status off workers while withholding freelance freedoms like price setting and access to information.
-
Is withholding pay information from gig workers morally wrong even if it is legal?
38:40Daniel Halliday argues that withholding information such as a job's pay or destination from a worker is unjust because no other party, including the customer, has an evident interest that justifies hiding it. He treats this as a distinct wrong from underpayment, one of unjustly depriving freelancers of a freedom they are entitled to.
-
How does paying gig workers by an averaged algorithm resemble Taylorist scientific management?
31:58Dan Calacci compares Shipt's practice of estimating an average shopping time from store square footage and item counts, then paying a flat rate for that estimate, to early twentieth century factory time studies that measured a prototypical worker to set wages as low as possible. He argues companies now hold enough individual data to pay a genuinely fair, individualized wage instead.
-
Can higher pay justify taking away a freelancer's freedoms?
41:40Daniel Halliday suggests that if compensation were high enough, the loss of freelancer freedoms might not seem so troubling, using the example that we would be less bothered by Uber's restrictions if drivers earned enormous amounts. He concludes the deeper problem in the gig economy is not classification itself but that the work is low paid and precarious.
-
What would a fairer, worker-designed gig economy algorithm look like?
49:02Dan Calacci argues that fair algorithms should be co-designed with workers rather than imposed unilaterally by company engineers, pointing to worker-owned cooperative delivery platforms, such as one operating in Colorado, that build pay and platform design around the interests of workers, restaurants, and customers together.
Sources
- Hi-Phi Nation, The Problem with Gig Work Podcast episode, original episode