Recommendation Algorithm

A recommendation algorithm is an automatic system that guesses in advance what you'll like and puts it on your screen. It's behind the endless queue of videos that pops up when you open a video app, and the 'you might also like this' list that shows up on shopping sites. It works a bit like a waiter at a restaurant you've been going to for years, who guesses your favorite dish and suggests it before you even order.

This system constantly watches what you click on, how long you watch something, and what you skip past, looking for patterns like "this person tends to like this kind of thing." It then picks what to show you next based on that pattern. The problem is that this calculation happens entirely inside the company and isn't visible from the outside. It's hard for users to know why a particular piece of content was recommended to them.

Because of this, there's growing pressure for companies to be more transparent about the criteria these systems use to operate. This is especially true for services with very large numbers of users, since a flawed recommendation can escalate into problems like failing to protect minors or spreading misinformation — which is why regulators are moving to look more closely at how these systems actually work.

The article notes that after the EU designated ChatGPT as a 'Very Large Online Search Engine,' a new obligation emerged requiring it to "disclose more transparently how its recommendation algorithm works." This means ChatGPT uses a system to select and surface content, similar to video apps or shopping sites, and the new regulation essentially demands that this mechanism not be kept hidden.