Instead of doing the English EW due yesterday, many of us have spent countless afternoons browsing through Youtube, Instagram, or TikTok. We don't notice until the bell for absence rings and 4 more vocab tests for tomorrow appear, but we have just succumbed to the evil tentacles of social media algorithms that the hegemonies of the technology industry employ. Amazingly, no matter how much you scroll, the Instagram reels that pop up mostly seem to be what you enjoy watching, be that Harry Maguire fails, chess opening traps, or Khaby reactions. Specifically, the more you view a certain type of video, the more it is likely to appear on future recommendations. With that in mind, how do social media algorithms work and what are some of the problems associated with them?
On a surface level, social media algorithms use data collected in order to promote more relevant recommendations. At its infancy, Facebook used reaction buttons as the main input; now, it has evolved to take into account multiple other factors: these include the number of times the video was watched, whether people scrolled past it without finishing, the number of times it was shared, and many more. Different social media platforms are tailored to different purposes.
There are a few reasons why social media algorithms are so important for their platform. First, the sheer magnitude and range of content (an estimated 95 million new posts occur on Instagram per day) means that some form of fine-tuned selection is necessary. Second, social media platforms profit off of screen-time. Their main source of income is through advertisements (now you know why there is a sponsored post every once in a while), and the more time you spend on the site, the more ads are likely to catch your attention and the more money platforms will make.
Let’s look at the specific example of Instagram. It has all a social media platform could ask for: posts from friends, posts from celebrities and news agencies, DMs, reels, shops, and more. There are 5 major factors that Instagram’s algorithm prioritises for determining whether a specific post should be in your feed.
These are few among many factors which are fit through a machine learning model that recommends certain posts and reels. The result is an extremely nuanced yet accurate collection of content that is largely relevant. Like any ML model, social media algorithms are constantly training and perfecting, whilst adapting to changing circumstances. Hence, even the most insignificant action of scrolling past a magic trick video because it is boring can add that marginal effect to your future recommendations.
There are several benefits to such algorithms. First, deterred by public backlash, platforms have dedicated extensive energy to filtering through inappropriate content. The same algorithms which promote progressive and relevant content also censors those with the potential to harm. Second, the impact of greater time spent on social media has generated a tidal wave of job opportunities. Content creators have become successful through their ability to monetise their passions and share them with the world. In fact, 50 million identify as social media influencers around the world. Furthermore, this widespread access allows for progressive content, in terms of discrimination and equality, to spread far easier. Third, it is undoubted that these algorithms do create a more personalised and enjoyable experience. Given that people are likely going to procrastinate regardless, it is better that they at least get happiness out of it.
There are also numerous important impacts resulting from the existence of these algorithms. At the individual level, it incentivises content creators to do a wide variety of tricks to allow the algorithm to prioritise their content. Obviously, they aim to create reasonably high-quality content to boost views and likes as well; often more importantly though, they employ certain such tricks. Below are several examples and why they work.
The data scientists among you know that any successful machine learning model requires a huge dataset to operate on. This means that contingent to social media algorithms is the collection of user data. It has been a subject of huge controversy for some time, where industry leaders such as Mark Zuckerberg and Sundar Pichai have all been questioned at congressional hearings related to this matter. People are more concerned about the use of personal data as goods, bought and sold between corporations and big tech companies. One breach can have catastrophic effects. Nevertheless, there are checks and balances which mitigate this. Social media platforms have cybersecurity measures to prevent it, too.
We will continue to be captivated and shackled by the algorithms, and we will continue to have to write English essays through the night to make up for it; whether it is for good or for ill, we cannot doubt the growing influence that the pinnacle of technology has over us.