Tech and Democracy
The Algorithm That Decides What a Citizen Knows
The most powerful editorial decision in the history of media is made billions of times a day by a system that has no idea what democracy is.
The Idea
For most of the 20th century, the gatekeepers of public information were at least theoretically accountable — editors could be fired, broadcasters could lose their licences, newspapers could be sued. The architecture of accountability was imperfect, but it existed. What replaced it is something structurally different: a set of engagement-optimisation algorithms that decide, at scale, what information most people encounter about the world. The critical thing to understand about these systems is that they were not designed to inform. They were designed to retain attention. Those two goals can overlap — a genuinely fascinating, true story holds attention — but they diverge sharply and predictably in the direction of outrage, novelty, and tribal confirmation. A system rewarding engagement will, over time, surface content that provokes strong emotional responses, because strong emotional responses reliably produce clicks, shares, and time-on-site. This creates a structural pressure on political information specifically. Policy is complex and slow. Scandal is simple and fast. Nuance requires effort to engage with. Moral clarity — even false moral clarity — is immediately satisfying. The algorithm does not choose misinformation deliberately; it simply has no mechanism for preferring truth over falsity when falsehood performs better in the first six hours. What's genuinely underappreciated here is that this isn't a bug that better engineering will fix. It is the product working as intended — just not intended for a healthy democratic information environment.
In the World
In 2021, a team of researchers at New York University's Center for Social Media and Politics ran an experiment that became one of the most cited studies in this space. They paid a group of participants to deactivate their Facebook accounts for four weeks before the 2018 US midterm elections, then compared them to a control group who stayed on. The deactivated group knew less about current events — the algorithm had been their primary news source — but they also held more moderate political views by the end of the study. The platform hadn't just informed them; it had moved them. Facebook's own internal researchers reached a similar conclusion even earlier. In 2018, a memo leaked to the Wall Street Journal showed that the company's data scientists had found that their systems were driving users toward 'more and more extreme content' and that 64% of people who had joined extremist groups on the platform did so because the algorithm had recommended it. The researchers proposed changes. Most were not implemented, in part because they were projected to reduce engagement. This is the democratic problem in concentrated form: the platform knew, the platform chose not to change, and the information environment of hundreds of millions of voters was shaped accordingly. No election commission approved this. No parliament voted on it. It simply happened, at the speed of software deployment.
Why It Matters
Most conversations about social media and democracy get stuck on content moderation — which posts get removed, which accounts get banned. That's a real debate, but it's downstream of a more fundamental question: who controls the information architecture that shapes what citizens believe is happening in their world? If you think of democracy as a system that requires a roughly shared factual substrate — a citizenry that may disagree fiercely on values but is at least arguing about the same reality — then the algorithmic fragmentation of information is a structural challenge, not just a cultural one. You can't vote well on climate policy if your feed has spent three years feeding you content selected for its ability to provoke, not its accuracy. Having sat with this idea, you might find yourself noticing your own information diet differently. Not just what you believe, but how you came to believe it — and whether the path there was designed by anyone who had your democratic participation in mind. The question isn't whether to use these platforms. It's whether to use them as if they were neutral.
A Question to Ponder
If the platforms that shape most people's political understanding were required to optimise for informed participation rather than engagement, what would they look like — and who would get to define what 'informed' means?
Get a new one of these every morning.
Start learning with Thinkable