Explained: The algorithms that run Facebook

Explained: The algorithms that run Facebook


One of the biggest revelations from the Facebook Papers, a series of news reports based on internal company documents, was about a test account set up by a Facebook researcher in India in 2019. Within three weeks of signing up, the platform’s algorithms had filled the fictitious user’s news feed
with the worst that humanity has to offer, including fake news, gore, and celebrations of death.

While algorithms have long had a bad rap, they are simply lines of code. The real problem is how social media companies use them to keep us hooked to their platforms. Here’s a quick explainer on algorithms and how Facebook’s new feed has used them over the years.

What is an algorithm?

An algorithm is simply a set of instructions for solving a problem or accomplishing a task. A recipe for butter chicken is an algorithm; so is the method for doing long division. Every computerised device uses them. But these simple algorithms are nothing like the ones Facebook, Google and others use to keep their platforms running.

What are those like?

As we mentioned before, there’s no such thing as “the Facebook algorithm”. Instead, what you see in your news feed, and in what order, is the result of several complex algorithms working in concert, or in Facebook’s own words “multiple layers of machine learning models and rankings”. These algorithms are built to predict which posts will be “most valuable and meaningful to an individual over the long term”.

They evaluate every post, give it a score, and then serve them up to you in descending order. To give you an idea of the complexity involved, this process happens every single time one of Facebook’s 2.9 billion users refreshes their news feed.

STARTUP ROCKSTARS IN 2021

Sign-in to see our list of the most promising startups of 2021



If these complex algorithms didn’t exist, your news feed would simply serve up every available Facebook post to you in the order they were posted. In fact, those of a certain age will remember that this is exactly what Facebook’s news feed, or its equivalent, was at the time.

How do Facebook’s algorithms rank posts?

While nobody but Facebook knows all the details of how its algorithms decide which posts to amplify and which ones to bury, the trove of internal company documents that
whistleblower Frances Haugen copied before she left the company offer some revealing glimpses.

The leaked papers show how algorithms designed for one purpose — maximising time spent on the platform — can end up recommending hateful, inciteful and gory content in India, or make Instagram a toxic platform for teenagers in the US.

What we also know is that in general, Facebook’s algorithms, like those of all ads-based social media companies, are designed to maximise the time that users spend scrolling through their feeds, so that they see more ads.

It’s important to note that algorithms, while automatic, are not autonomous. They operate strictly according to their program and must be told — by a human — which signals to pay attention to, and which outcomes to aim for.

How have Facebook’s algorithms evolved over the years?

As we mentioned earlier, Facebook’s algorithms have been growing in complexity since 2009, when it decided to rank and sort the items in users’ news feeds rather than serve them as in the order they were posted. Here are the milestones in the history of “the Facebook algorithm”.

  • 2006: Facebook newsfeed debuts.
  • 2007: Like button is introduced.
  • 2009: Facebook begins a simple method of sorting posts; those with the most Likes get bumped to the top of the feed.
  • 2015: Pages with too much promotional content (ads masquerading as posts start getting downranked. Facebook also lets users influence the algorithm directly by introducing the “See First” feature, which lets users tell Facebook to prioritise a certain Page’s posts in their feed.
  • 2016: Facebook starts measuring a post’s value based on the amount of time users spent on it, even if they don’t like or share it.
  • 2017: Facebook starts prioritising emoji-based emotional reactions over Likes.
  • 2018: Responding to widespread criticism, CEO Mark Zuckerberg says in January that Facebook’s algorithms will now prioritise “posts that spark conversations and meaningful interactions”.
  • 2019: Facebook starts bumping up posts from “close friends” — those that people engage with the most. This comes after it received widespread criticism that its 2018 algorithm changes only ended up increasing outrage and divisiveness and promoting more misinformation and hateful content.
  • 2020: Facebook announces that its algorithm will now evaluate the credibility and quality of news articles to promote substantiated news rather than misinformation.



Source link

Add a Comment

Your email address will not be published.