"I like to know what people are saying"

A good share of what you read isn't news in the strict sense. It's reaction to news. That covers the opinion column, the furious thread, the clip of one politician attacking another, and the comments underneath. On Monday you read one piece at lunch that made you angry and another that made you gloomy. You might defend this as hearing all sides. You like to know what people are saying and how the argument is going.

The argument is worth knowing, in moderation. A good column by someone who knows the subject can clarify an issue in ten minutes. What I'd question is the kind of opinion that reaches you, because it's selected even more ruthlessly than the news itself.

Steve Rathje, Jay Van Bavel and Sander van der Linden analysed 2.7 million posts on two large social networks, from news organisations and American politicians. They wanted to know what made a post spread. The strongest predictor by far was whether it was about the other side. Posts about the political opposition were shared about twice as often as posts about the poster's own side. Each additional word referring to the opposing group raised the odds of a post being shared by 67 per cent. That effect was several times stronger than the effect of negative language in general. Posts about the opposition mostly drew angry reactions.

An earlier study by William Brady and colleagues looked at more than half a million messages about three divisive political issues. Each additional word that was both moral and emotional, words such as "shameful" or "evil", increased a message's spread by about 20 per cent. The spread was mostly within one political camp and seldom crossed to the other.

So the commentary that finds its way to you has passed through a filter that favours anger at opponents, expressed in the language of moral outrage, and circulated among people who already agree. It isn't a sample of what people are saying. It's the output of a system that promotes whatever makes people furious. The reasonable majority, who think the issue is complicated and the other side not wholly wicked, produce little that travels.

This affects your picture of your fellow citizens. If the other side is known to you mainly through its most enraging members, as selected and relayed by your own side, you'll conclude that they're stupider and nastier than they are. They're concluding the same about you, from the same kind of evidence. It's the crime survey finding from chapter 4 again, applied to people instead of crime. Your judgement of the people you meet is fairly accurate, while your judgement of the people you only read about has been distorted.

It affects your mood as well, in a way that's different from the jolt. Anger is more energising than fear, and for a few minutes it can feel good. It feels like clarity and like being on the right side. Then it leaves you stirred up with nowhere to put it. You've had an argument in your head with someone who'll never hear it. The lunchtime column that made you angry took twenty minutes to read and cost you the first hour of the afternoon.

The rule in chapter 15 doesn't exclude opinion. If there are two or three commentators whose judgement you've come to trust, including at least one you often disagree with, read them, at length, in your weekly session. What the rule removes is the constant drip of other people's fury, which tells you very little about what your neighbours think.