What is "The Algorithm"?
"The Algorithm" is what colloquially called the computer program responsible for picking posts that may interest an user in a social media website. This is also known as a recommendation algorithm. Note that "the algorithm" doesn't mean the same thing an algorithm. All computer programs are algorithms, but only a certain way to pick posts is called "the algorithm" on the Internet.
What does The Algorithm do?
Essentially, there's a list of posts (also called a "feed") somewhere (or a list of videos in Youtube's case), and a computer program needs to decide which posts should go into this list. The simplest algorithm to make this decision is the chronological feed, where the computer just takes the newest posts and sorts them from newest to oldest. Despite all computer programs being algorithms, this chronological feed is not known as "the algorithm." Instead, "the algorithm" refers to the alternative, also called an "algorithmic" feed, where a computer program takes in consideration what kind of post an user is interested into an tries to match that with all of the other posts in the social media website, often mixing it with whatever is popular or recent (fresh) at the moment. In some social media that feature both kinds of feeds, such as Twitter and Tumblr, the algorithm feed is labelled "For You" and is accessible through a tab. The screen where the feed is displayed is also known as the "for you page," and sometimes found as the hashtag #fyp.
There are many ways to implement a recommendation algorithm, from simple methods that simply match categories, to complex methods that use machine learning (ML, also called "AI" sometimes). With ML, the program displays a list of posts to you and it knows some properties of the posts, such as which category they belong. If you click on a post, that the program "learns" from that input, and becomes more likely to recommend that type of post to you. The more you click on the same category of posts, the more it reinforces the learning. With ML, it's hard to say why exactly a post is being recommended. All we know is that the program has a reason to recommend a post, and if you give positive feedback (click on it), it validates the reason, so whatever reason that may be, the machine will consider it more often in the future. For example, the machine may randomly try to recommend users who watch videos about hydraulic presses videos about puppies. If users start clicking on videos about puppies after watching videos about hydraulic presses, it would validate the machine's reasoning and the machine would start recommending them even more, even though for a human being that recommendation seems to make no sense at all.
Leave a Reply