"I need to rule out the worst"

You knew by five to eleven that twitches are common and very rarely serious. You kept reading because "very rarely" isn't "never", and you wanted to get from one to the other. The plan was to read about the serious causes, find that they didn't fit, and sleep.

That isn't how it usually goes. Ryen White and Eric Horvitz, two researchers at Microsoft, studied how it does go, using a large set of search logs together with a survey of 515 people. They were interested in what they called escalation. That's a search that begins with a common, probably harmless symptom and moves to reading about a rare and serious condition.

They found that search engines have the potential to escalate medical concerns. Whether a given search escalated depended partly on what people saw. It depended on how much medical content they viewed, and on whether the pages used alarming terms. It also depended on the person, since some people were more inclined than others to go for the grave explanation. And the worry didn't end when the search did. It persisted afterwards and interrupted what people were doing, across later sessions and other days.

Why would a search lean towards the serious? Here's my own reasoning, and I've three suggestions. The first is about what gets written. A serious disease has whole websites devoted to it, run by charities, hospitals and patient groups, and each one lists every early sign, including the common ones. A twitch caused by coffee has nothing of the kind. Nobody founds a charity for it. So the volume of text linking "twitch" to the rare disease is far greater than the volume linking it to caffeine, though caffeine is a far more likely cause.

The second is about who writes about their own symptoms. The man on the forum whose illness began with a twitching calf wrote a long post, because it's the most important thing that ever happened to him. The thousands of people whose calves twitched for a week in a stressful month wrote nothing, because nothing happened. You can't read their posts. If you could, there would be too many of them to get through.

The third is about memory. Amos Tversky and Daniel Kahneman showed in the 1970s that people judge how likely something is partly by how easily examples come to mind. A vivid story about one man comes to mind very easily. "Very rarely a sign of anything serious" is a statistic, and it doesn't stay in the memory the same way. After an hour on the forum the rare disease feels likely, because it's now the most vivid thing you know about twitches.

There's also a flaw in the plan itself. You wanted to rule the worst out, and that can't be done by reading. It takes evidence about your own body, such as strength that's normal on examination, a test that comes back clear, or a twitch that has gone within two weeks. A page can only tell you what the disease looks like, and almost any early symptom list looks a bit like almost anyone. So when you read in order to rule something out, you usually find something that seems to fit.