Read a sample: 1. The Suggestions Around You
You finish watching a video about repairing an old bicycle. Before you have decided what to do next, the screen offers several possibilities: a guide to adjusting brakes, a tour of a bicycle workshop, and a review of a tool kit. One of the videos looks useful, so you select it.
A little later, you open a music app. Its home screen offers a playlist described as a good match for you. Then you visit an online store to look at a lamp, and the page displays bulbs, side tables, and other lamps you might want to consider.
These moments feel ordinary because suggestions are woven into so many digital experiences. They appear when you are relaxing, shopping, learning, or catching up with people. Some are helpful. Others seem strange, repetitive, or completely unrelated to anything you had in mind.
Behind many of these suggestions is a recommendation system: a way of selecting or ordering options that a service presents to someone. The options might be videos, songs, products, articles, accounts, or posts. A system may use information about the options, the person, the situation, or patterns across many people to decide what to show.
That description is deliberately broad. Recommendation systems do not all work alike, and a suggestion on a screen does not reveal exactly how it was chosen. The goal of this book is to make those systems easier to think about without pretending that an outsider can see every decision a particular service makes.
We will begin with a skill that is useful throughout the book: noticing what is actually on the screen, then separating it from a story about why it appeared.
Imagine walking into a small bookshop and asking the bookseller for something to read. The bookseller might ask what you have enjoyed before, point toward a popular new novel, or suggest a title similar to the one you are carrying. You may appreciate the help even if the bookseller knows very little about you.
A digital recommendation can serve a similar purpose. It offers a possible next choice when there are more options than you would reasonably inspect one by one. A streaming service might have a large catalog, but its home page has room for only a small portion of it. A shopping site might carry many lamps, while a product page displays just a few alternatives. A social feed must decide which available posts to place where you are likely to see them.
The analogy has limits. A bookseller can explain a suggestion in conversation, and you can ask follow-up questions. A digital service may offer a short label such as “Because you watched” or “Popular now,” but it usually does not show every consideration behind a list. The suggestion may also be generated for a particular moment rather than by someone who has formed an understanding of you.
Most importantly, a recommendation is not a verdict. A suggested documentary does not prove that you love documentaries. A suggested kitchen gadget does not prove that you intend to buy one. The service is presenting an option, not reading your mind.
Entertainment services provide some of the easiest examples to spot. A video platform may display clips on its home page or beside a video you are watching. A music app may assemble a playlist, suggest an artist, or continue playing related tracks after an album ends. A streaming service may arrange films into rows with labels based on genre, popularity, or something you recently watched.
In each case, notice that the recommendation can involve more than the choice of an individual item. A service might also choose which row appears first, which image represents a film, or which song begins a playlist. The position and presentation can affect what catches your eye, even when several options are available.
Shopping suggestions take many forms too. A page for a coffee maker might show filters or mugs that could be used with it. It might also show competing coffee makers at different prices. After you browse, the store’s home page might display a collection that appears related to your visit.
Those placements do not necessarily have the same purpose. A mug shown alongside a coffee maker may be offered as a companion item; another coffee maker may be offered as an alternative. A prominent product might also be an advertisement. A single page can contain recommendations, ads, ordinary catalog links, and results from a search you performed. The labels and layout help you describe what you see, though they may not disclose the complete process behind each placement.
Social feeds can be less obvious because recommendations are mixed with material from people or accounts you have chosen to follow. A feed might include a friend’s photo, a post from an unfamiliar creator, a suggested group, and an advertisement. Even when all the posts come from accounts you follow, their order may reflect a selection or ranking process rather than a simple sequence of publication times.
In a ranked feed, the first post you see is not necessarily the newest one, and the tenth post is not necessarily the tenth newest. The service may have arranged available material using criteria that are not visible to you. Different services, and different parts of the same service, may arrange content differently.
Suggestions also appear outside these familiar settings. A news app might highlight articles, a library…