Question 1

How do platforms experiment on users?

Accepted Answer

Continuously, through **A/B tests** running on live users without individual notice — which is standard practice across the industry and is why your app can differ from a friend's on the same version.

**How it works.** Users are assigned randomly to variants: a different layout, ranking, notification schedule, wording or feature. The platform measures outcomes — time spent, clicks, retention, revenue — and ships whichever performs better on the chosen metric. Large platforms run **hundreds or thousands of experiments simultaneously**, so no two users see quite the same product.

**What gets tested:** interface changes; **recommendation and ranking weights**, which is the most consequential category; notification timing and frequency; onboarding flows; pricing and paywalls; and the wording of prompts and consent dialogues.

**Why it is defended.** Intuition about interfaces is unreliable, and testing replaces the opinion of whoever is most senior with evidence. It catches harmful changes before a full rollout, allows gradual deployment, and is how most measurable product improvement actually happens.

**Where the genuine objections lie:**

**Consent.** Users agree to terms of service, not to specific experiments. Whether blanket consent covers experiments that affect mood or behaviour is a real ethical question, raised sharply by a well-known emotional contagion study that manipulated feed sentiment and caused substantial controversy.

**The metric is the value judgement.** Optimising for time spent produces a different product from optimising for whether people report the time as well spent. **The choice of metric is where the ethics live**, not in the testing.

**Vulnerable users** are not excluded, and cannot easily be identified.

**Academic research would require ethics approval** for the same intervention that a company can run commercially — a genuine asymmetry.

**What is changing:** internal review boards at some companies; regulatory attention to design and dark patterns; researcher access requirements in some jurisdictions; and greater scrutiny of experiments involving minors.

**What you can do:** very little directly. Feature flags are not user-controllable. Noticing that your experience is personalised and provisional is itself useful.

Question 2

How do platforms actually roll out changes?

Accepted Answer

Gradually, to small randomised slices of users, measured against a control group — which is why your feed can differ from a friend's on the same day, and why "they changed the algorithm" is

Question 3

Why do social platforms keep copying each other's features?

Accepted Answer

Because copying is cheap, fast and low-risk relative to inventing — and because the thing being copied has already been validated by someone else's users, which removes the main uncertainty in

Question 4

Why do platforms keep changing or shutting off their APIs?

Accepted Answer

An API (application programming interface) lets external software interact with a platform programmatically. Platforms have progressively restricted, priced or closed theirs, and the reasons are

Question 5

Why do platforms get worse over time?

Accepted Answer

Because the incentives change as a platform matures, in a pattern regular enough to have been named — enshittification, Cory Doctorow's term for a three-stage decay from serving users, to serving

Question 6

What is social commerce, and why are platforms pushing it so hard?

Accepted Answer

Buying and selling inside a social app rather than clicking through to a retailer's website — and platforms are pushing it because it solves their two biggest commercial problems at once. Why

Question 7

How do social platforms actually make money, and how does that shape what you see?

Accepted Answer

Almost entirely by selling attention to advertisers, which means the product being optimised is not your satisfaction but the time and engagement that can be sold — and nearly every design decision

Related Questions