Network Science & Complexity
Why Bad Ideas Travel Faster Than Good Ones (And What Networks Have to Do With It)
The most viral piece of misinformation in recorded history spread six times faster than the truth — and the reason has almost nothing to do with human gullibility.
The Idea
When researchers study how things spread through networks — diseases, rumours, innovations, memes — they keep bumping into a counterintuitive finding: the structure of the network matters far more than the quality of what's spreading. A brilliant idea with a slow-burn diffusion pattern will consistently lose to a mediocre one that happens to enter the network at the right node. The key concept here is the 'superspreader' — not a person who is especially persuasive, but a node that is exceptionally well-connected. In network science, this is measured by something called degree centrality: how many direct connections a node has. But there's a subtler measure called betweenness centrality, which captures how often a node sits on the shortest path between two other nodes. A person with high betweenness isn't necessarily the most popular; they're the one information must travel through to get from one cluster of the network to another. They are bridges. What makes ideas spread fast isn't virality in the pop-culture sense — it's whether an idea reaches these bridge nodes early. Once it does, it jumps across otherwise disconnected communities simultaneously. This is why a fringe belief can seem to appear everywhere at once: it didn't grow gradually, it teleported. And because emotional, simple, or outrage-triggering content is more likely to be reshared impulsively — which is how you reach bridge nodes quickly — the architecture of our networks has a structural bias toward a particular kind of content, regardless of what any individual chooses to promote.
In the World
In 2018, a team at MIT published a landmark study in Science — the largest empirical analysis of misinformation ever conducted. Soroush Vosoughi, Deb Roy, and Sinan Aral tracked roughly 126,000 news stories on Twitter over eleven years, tracing how true and false stories each propagated through the network. The finding was stark: false stories spread to 1,500 people roughly six times faster than true ones, and they penetrated deeper into the network. Crucially, the researchers controlled for bot activity — automated accounts were not the culprit. Human accounts were responsible for spreading false information faster and more broadly. But here's the part that reframes the whole picture: the false stories weren't winning because they were more emotionally manipulative in some obvious way. They were winning because they were novel. True stories, by definition, tend to confirm or extend what people already know. False stories — precisely because they're unconstrained by what actually happened — are disproportionately surprising. And surprise triggers sharing. The network, in other words, is an optimization engine. It reliably surfaces content that generates engagement, which is content that is new, unexpected, or emotionally activating. It has no mechanism to weight accuracy. So the architecture doesn't merely passively allow misinformation to travel — it actively selects for the properties that misinformation tends to have. The problem isn't a bug in human psychology or a failure of platform moderation; it's a structural feature of how information flows through highly connected, engagement-rewarded networks.
Why It Matters
Understanding this shifts the diagnosis — and therefore the solution. If you believe misinformation spreads because people are credulous or lazy, you end up investing in fact-checking, media literacy campaigns, and content moderation. All of those have value, but none of them touches the underlying architecture. If you understand it as a network problem, different interventions emerge. Slowing down the bridge nodes — introducing friction before resharing, for instance — changes the dynamics even if the content itself is unchanged. Reducing the reward signal for engagement-above-all alters what the network optimises for. These are structural solutions to a structural problem. On a personal level, this knowledge is quietly liberating. When you notice that some idea has spread explosively, your first question can shift from 'is this popular because it's true?' to 'what properties of this idea made it travel well, and are those properties correlated with accuracy?' Popularity and network fitness are not the same thing. Knowing this doesn't make you immune — nobody is — but it gives you a useful pause between encounter and belief.
A Question to Ponder
If the networks you participate in are structurally biased toward certain kinds of content, which of your current beliefs might have reached you not because they're well-supported, but because they travelled well?
Get a new one of these every morning.
Start learning with Thinkable