Have you noticed how there’s always new content to watch on the internet that really speaks to you? How it almost feels tailor-made for your consumption? What about news articles which convince you that the political party of your choice is doing fantastic work while the others are train-wrecks? This phenomenon of hyper-personalization of entertainment and politics is driven to a very large extent by a class of Artificial Intelligence algorithms known as recommendation systems.
Like the name suggests, recommendation systems have a single objective — to make a few specific and personalized recommendations out of a large number of choices to keep individuals glued to their devices. The algorithms simultaneously find items that are similar to each other, and individuals who are similar to each other. This information is then used to recommend new items to each individual. So for example, you might get a recommendation because it has similar characteristics to something you already like — or because someone else, who largely enjoys the same things you do, liked this new thing.
A concrete example from everyday life: a streaming platform (like Netflix) contains viewing data of different users on different films. Consider two users on the platform, 'Reshma' and 'Shruti'. The data shows that they tend to watch a lot of the same films, particularly films with a strong female lead. The data also shows that Reshma has watched a film in this category which Shruti hasn’t, “Raazi”. A data-driven recommendation for Shruti is ready: watch “Raazi”.
The algorithm is essentially coded common sense, it behaves like a virtual friend who says to Shruti, “Reshma and you are so similar, she loved this movie and I’m sure you will too”. What makes recommendation systems particularly effective is that instead of a single common friend, they use...