How Do Social Media Algorithms Work? | Explained in Under 10 Minutes

Lewis confesses he used the word "algorithm" three times in clinic last week and isn't entirely sure he was using it correctly. Emma's answer disarms immediately: at its most basic, an algorithm is just a set of instructions. A recipe is an algorithm. Your morning routine is an algorithm. The term comes from al-Khwarizmi, a ninth-century Persian mathematician whose Latinised name became the word itself. What's changed is the speed and scale at which machines can execute instructions.Emma explains the pivot from the simple definition to the social media reality: a recommendation algorithm considers thousands of factors simultaneously — your past behaviour, the time of day, trending content, how long you watched similar videos, whether you scrolled past or lingered. Modern algorithms learn: if showing you cat videos at 11pm keeps you scrolling, the system updates its own rules to show more. The engineers don't program every decision — they design systems that figure out the decisions themselves. Lewis draws the health line he sees clinically: teenagers who can't sleep after three hours of doom-scrolling, patients whose anxiety is fed by content specifically selected to provoke emotional responses. The algorithm doesn't care about cortisol levels. Neev cites the 2019 YouTube research: the recommendation algorithm was responsible for 70% of total watch time; researchers documented how starting with mainstream content, each recommendation would edge incrementally toward more extreme material — not because anyone programmed radicalisation, but because extreme content maximised the metric being measured.Lewis closes on the hopeful note Emma insists on: these are human-designed systems. They can be human-redesigned. Someone chose to optimise for engagement. Someone could choose differently — for understanding, for calm, for genuine connection. The EU's Digital Services Act came into full force in 2024, requiring large platforms to offer users chronological feeds and submit to independent audits. The technology is neutral. The objective function is not. Every scroll is a vote for whatever comes next.If you enjoyed this episode, please subscribe — and send your topic suggestions to [email protected]. We read every one.