Week 1: The algorithm has a boss too

What "algorithmic management" means as a term of art, and how the next eleven weeks are shaped around one industry's version of it

Open the slides

What “algorithmic management” means#

Lee, Kusbit, Metsky and Dabbish’s 2015 study, “Working with Machines: The Impact of Algorithmic and Data-Driven Management on Human Workers” (presented at CHI, the premier human-computer interaction venue), gave this course its working definition: algorithmic management is what happens when software takes over decisions a human manager used to make — who does which job, how their performance is rated, when they get warned or let go. None of that requires a robot boss standing over anyone. It only requires that the deciding moved from a person into a rule nobody in the room can point to.

That definition sat mostly inside academic HCI for a few years before it became a mainstream policy term. Kellogg, Valentine and Christin’s 2020 review, “Algorithms at Work: The New Contested Terrain of Control” (Academy of Management Annals), is the piece that best maps how far the idea travelled — from a description of how gig platforms behaved into a general theory of workplace control that now gets cited in labour economics, organisational sociology and the regulatory determinations you’ll read in the second half of this course. Knowing that lineage matters for your essays: when you cite “algorithmic management,” you’re invoking a specific, traceable body of work, not a piece of jargon anyone coined to sound serious.

Food delivery is the cleanest place on the planet to watch this happen, because almost the entire job — which order you get, how long you’re told it should take, whether you’re rated well enough to keep getting more work — is decided by software before a human supervisor ever enters the picture. A warehouse picker or a call-centre agent doing algorithmically managed work still usually has a shift supervisor to appeal to when a metric looks wrong. A delivery rider frequently does not: the app is the only interface between the rider and the decision, start to finish. That’s why this course keeps returning to delivery work as its object, whichever platform or country a given week is looking at.

How the semester is shaped#

Two halves. Weeks 2–6 open up the machinery itself — dispatch and ETA prediction, rating and lateness thresholds, the resistance riders build back, and the economics that make labour the one line item a platform treats as flexible — anchored in a single deep case: Uber, read mainly through Rosenblat and Stark’s (2016) study of Uber drivers. Week 3 sets that alongside a second, closely-documented case from a different country: China’s food-delivery sector, via Renwu magazine’s 2020 feature “Delivery Riders, Trapped in the System,” the report that lent this course its title. Weeks 7–12 turn the same mechanisms over to regulators, comparing how China, Australia and the EU each answer the same question differently: what is an algorithm allowed to decide about a worker’s day.

The two halves are designed to depend on each other. An essay in week 11 that compares regulatory regimes without a concrete mechanism to point to is just describing three press releases; a week 3 close reading of a case study without weeks 7–10’s legal vocabulary can describe what happened but not why the platforms’ response took the specific legal shape it did. You need both halves to do either one well.

The claim we’re testing#

“The algorithm decided” is a sentence you’ll hear from every platform this semester touches, and it can mean two very different things. Sometimes it’s an accurate description of a system genuinely too complex for any one person to have specified the outcome by hand. Sometimes it’s a place where a decision was made by a specific person, on a specific date, and the sentence exists to make the paper trail stop there. Telling the two apart — not assuming either one by default — is the actual skill this course is trying to build, and it’s the standard every essay this semester is held to.

Before week 2#

Open a food-delivery app you don’t already use, add a few items to a cart and step through to checkout — no need to complete the order — paying attention to what the app tells you about time and price along the way, and what it doesn’t. Bring one specific observation to the dispatch-mechanics lab.