Amazon

Context

From a history of Decentralized Commerce, by Ori Shimoni:

"The platform era began when companies moved beyond simply connecting buyers and sellers to creating comprehensive ecosystems that controlled every aspect of user interaction. In 2000, Amazon led this transition when it opened its website to third-party sellers. Initially positioned as a retailer selling books directly to consumers, Amazon transformed into a platform where independent merchants could list their products alongside Amazon's own inventory.

Unlike eBay's peer-to-peer approach, Amazon's marketplace created a fundamentally different relationship between the platform and merchants. Sellers operated as businesses rather than individuals, and Amazon maintained strict control over the customer relationship. Merchants couldn't communicate directly with buyers, had to compete with Amazon's products, and faced suspension for violating the platform's complex policies. The marketplace model had evolved from enabling peer-to-peer trade to managing a vast network of dependent businesses.

Unlike eBay, Amazon positioned itself as actively guaranteeing customer satisfaction through its A-to-Z Guarantee. It often sided with buyers in disputes and required sellers to maintain high performance metrics. The introduction of Fulfillment by Amazon (FBA) allowed sellers to ship their inventory to Amazon warehouses, and the platform handled storage, packing, shipping, customer service, and returns. This improved reliability and convenience compared to eBay's seller-managed logistics, but increased seller dependency on Amazon's infrastructure and algorithmic control.

The sharing economy emerged as the next evolution in platform marketplaces. In the wake of the 2008 financial crisis, as smartphones became ubiquitous, platforms like Airbnb and Uber introduced a new vision focused on services and asset utilization rather than goods. These platforms enabled individuals to monetize their homes, cars, and skills through integrated systems for discovery, booking, payment, and reputation. Initially positioning themselves as lightweight peer connectors with narratives of microentrepreneurship and empowerment, the “sharing economy” promised to democratize opportunity while creating more efficient markets. Airbnb’s marketing highlighted how it helped homeowners earn extra income by renting spare rooms, while Uber emphasized how anyone with a car could become a driver in their spare time.

As these platforms scaled, their relationship with service providers evolved significantly. These services came to comprise the “gig economy”—a labor market characterized precarious working conditions, lack of benefits, and unstable income. The model expanded beyond ride and home sharing to food delivery and grocery shopping (DoorDash, Deliveroo, Uber Eats, Instacart), household tasks (TaskRabbit, Thumbtack), professional services (Upwork, Fiverr), and more.

What began as lightweight intermediation evolved into comprehensive control over the transaction lifecycle through algorithmic means, including:

• Pricing control: Dynamic pricing algorithms (like Uber's surge pricing) determined what providers could charge.

• Work allocation: Algorithms determined which providers received opportunities based on various factors.

• Behavioral management: Rating systems and acceptance requirements enforced platform-determined standards.

This control came with higher costs, with commission rates increasing from 10-15% to 20-30%, but it improved safety and reliability. Platforms built comprehensive trust systems—Airbnb implemented host verification, payment protection, and property damage coverage; Uber deployed background checks and emergency response. They prevented users from circumventing fees through technical barriers—filtering messages to block contact information and designing apps to limit direct communication.

Critics termed the result “platform feudalism”—a system where service providers owned their assets (cars, homes, labor) but operated within an economic relationship where platforms exercised near-complete control while extracting significant value. This growing power imbalance sparked backlash between 2016-2018, with worker protests, regulatory challenges, and increasing public skepticism about the sharing economy's promise."

([[1]])

Discussion: Labor Issues

Densification of Labor at Amazon

Trebor Scholz:

“Let’s start by asking for a show of hands: who has ever purchased anything from Amazon.com? It is incredibly convenient but, as I will show, in the shadows of this convenience linger the social costs for Amazon’s workers. Behind the screen: wage theft and total workplace surveillance. This is one of Amazon’s so-called “inactivity reports.” These documents are issued by the company to its warehouse workers. The example here is from Germany; I translated it for you:

- Colleague … was inactive 07:27 am to 07:36 am (9 minutes). [Worker] … and [worker] … were seen standing in between shelves 05-06 and 05-07. Already on [date], 2014 [worker] was seen inactive from 8:15 am-8:17 am (2 minutes). Also on [date], 2014 [worker] was inactive from 07:13 to 07:14 (1 minute).

In Leipzig, Germany, one logistics worker in an Amazon “fulfillment center” was accused of having been inactive on two occasions. Consequently, he was fired five minutes after his second “digression.”

Such densification of labor, to use labor scholar Ursula Huws’s term, is possible because workers are carrying scanners that can be tracked and supervisors constantly monitor warehouse workmen. It’s Taylor’s scientific management on steroids.

Also part of the Amazon story, just a few weeks ago, the U.S. Supreme Court issued a ruling stating that workers in these warehouses have no right to be paid for the time that they are waiting for mandatory security screenings, when they are leaving the warehouse.” (http://www.publicseminar.org/2015/04/think-outside-the-boss/)

The online labor brokerage: Amazon Mechanical Turk

Trebor Scholz:

“Since 2005, Amazon also operates an online labor brokerage: Amazon Mechanical Turk (AMT). Workers log on to the website and pick tasks from long listings. Similar to traditional piecemeal work in the garment industry, Mechanical Turk allows for a project to be broken down into thousands of bits, which are then assigned as individual tasks to so-called crowd workers.

On AMT, like in many crowdsourcing environments, inexperienced, novice workers are making between two and three dollars an hour. Just like migrant workers, barristers, or workers in the fast food industry, they are working long hours, are underpaid, are insufficiently protected by labor laws, have few or no benefits, and are often treated poorly by their bosses.

Amazon claims that Mechanical Turk has a labor pool of 500,000 workers. Other crowdsourcing companies, like CrowdFlower, point to an even larger invisible, global workforce that for all practical purposes remains anonymous. Compared to traditional workplaces, these are solo workers who are isolated from each other. Some companies even have clauses in their terms of use that prohibit workers from getting in touch with each other. The labor brokerage oDesk-eLance may have a labor pool of as many as eight million such workers.

On Mechanical Turk, wage theft, while explicitly tolerated by Amazon, is a daily occurrence. Some consigners reject accurately executed work to avoid payment. Rejecting it, does not, however, stop these “black hats” and scammers from still using the work.

AMT’s conditions of use clearly state that consigners own the work immediately upon receipt, which means that they can do whatever they please with it. In a further twist, they don’t even have to explain their rejection of already performed work. Wage theft is a feature, not a bug. Consequently, it is not surprising that the turnaround among Turkers is roughly 70% every six months, according to Turker Rochelle LaPlante.

You might object: wage theft, payments below minimum wage  — how is that even possible? After all, there is the Fair Labor Standards Act and many other legal protections for workers that would immediately power down such operations. A democratic society would not stand for exploitative work environments, right?

For now, such violations are tolerated by a broad coalition of silence. In 2011, the Department of Labor had just 1000 inspectors who were responsible for one 130 workers in seven million enterprises. Such strategic understaffing of the Department of Labor means that employers who violate labor regulations in terms of wage or safety only have a very small chance of getting caught and even if they get caught, all they have to do is give workers what they owe them. It’s like robbing a bank with the only disincentive being that they might have to return the loot when they are caught. Startups cleverly sail around the definition of employment by restructuring work in such a way that the people who are executing the tasks can be categorized as independent contractors instead of employees.

Amazon Mechanical Turk is relatively small; the actual number of active workers might be closer to 10,000. There are no conclusive studies about the size of the workforce in the crowd sourcing sector overall and Amazon treats these numbers as a trade secret. I am using Mechanical Turk as an example because it is an infectious business model that is mirrored on countless other upstarts. Now that the cruel genie is out of the bottle, the business logic of Mechanical Turk has been adapted by companies like 99designs and countless others.

Amazon‘s reputation is, of course, not solely built on micro wages and total workplace surveillance; it is also widely appreciated for its low prices and convenience (and that even before drone delivery). But this consumptive advantage comes at a price. To understand this, we can remind ourselves of last year’s face-off between Amazon and publishers. Jeff Bezos, the CEO of Amazon, told a group of publishers, including Hachette, that “Amazon should approach … publishers the way a cheetah would pursue a sickly gazelle.”

A few days ago, Amazon announced a new venture, called Home Services, which installs them as an online middleman when you are hiring an electrician or plumber. In cooperation with TaskRabbit, Amazon aims to enter this sector to collect rent on your home repairs and services, interestingly also including “academic services.”

So far, based on the evidence from Mechanical Turk and the labor practices in Amazon’s warehouses, there is no reason to believe that the company would understand its relationship to digital laborers any differently than that of a predator pursuing its vulnerable prey.”

let me discuss the invisible workforce of Amazon Mechanical Turk, before I move on to the “sharing economy.”

The original “Mechanical Turk” was designed by the Hungarian nobleman Wolfgang von Kempelen in 1769. A small-bodied chess player who controlled the mechanical hands of the Turk operated this “automaton,” hidden in a wooden case. The spectacle of the seemingly complex, mechanized chess-playing machine, complete with a turban, put small technical details on display as distraction while keeping the actual human labor out of sight. The operator-worker remains hidden in the black box, quite literally. It was a mega-hit in Europe at the time with dignitaries like Catherine the Great, Charles Babbage, and Edgar Allan Poe coming to experience it. In 1940, Walter Benjamin references Mechanical Turk in Theses on the Philosophy of History. Already in 1814, E.T.A. Hoffmann wrote about the Mechanical Turk in a short story entitled “Automata.” The name “Amazon Mechanical Turk” pays homage to this 18th century machine.

Where the historical Turk showed off technology to draw attention away from the human laborer, today, Mechanical Turk’s and CrowdFlower’s and Handy’s and oDesk’s crowd sorcerers work with coolness and the spectacle of innovation to conceal the worker.

As Alex Rivera put it in his film Sleep Dealer: owners are getting “all the work without the worker.”

Crowdsourced micro-tasks are paid between one cent and several dollars each. Tasks include the description or categorization of products, the filling in of surveys, the filtering out of pornography on social media, the removal of stuff that violates the terms of service, the tagging and labeling of images, and the transcription of audio and video recordings or receipts.

About 18% of Turkers are treating AMT as a full-time job, based on their reports that they are “sometimes or always” relying on Mechanical Turk to “make basic ends meet” (Ross et al.).

(http://www.publicseminar.org/2015/04/think-outside-the-boss/)

More information

Books:

• Fulfillment: Winning and Losing in One-Click America

by Alec MacGillis Farrar, Straus, and Giroux, 2021, 400 pages

• Amazon Unbound: Jeff Bezos and the Invention of a Global Empire

by Brad Stone Simon & Schuster, 2021, 488 page