Godwin’s Law famously says that the longer an online discussion continues, the greater the probability that someone eventually mentions Hitler or the Nazis.
The literal version of the law doesn’t seem to survive empirical testing particularly well. Researchers tested it against 199 million Reddit posts and found that reality was more complicated.
But the underlying question remains excellent: Why do perfectly normal online conversations become increasingly unpleasant as they grow?
A new study offers a wonderfully human explanation.
Researchers analyzed around 2 billion comments across 2,150 Reddit communities and found a consistent pattern: as discussion threads become longer, comments become more negative. Communities themselves show something similar as they age.
You could blame algorithms. Naturally. We blame algorithms for almost everything now, including presumably the weather. Except the researchers then ran an experiment with 3,685 people.
One group was simply asked to contribute to an existing conversation. Another was encouraged to say something different from what had already been said. Nobody asked them to be angry. Nobody asked them to attack anyone. They were merely asked to be original. And as the discussion continued, the originality group became increasingly negative.
Which suggests a rather entertaining mechanism. Imagine arriving at a thread containing 500 comments. Someone has already written the obvious positive observation. Someone has written the slightly less obvious positive observation. Someone has made the reasonable criticism. Someone has criticized the criticism. Someone has provided nuance. Someone has already provided nuance to the nuance. You still want to contribute something distinctive.
What’s left? Well. This is all complete bullshit.
Congratulations. Original thought achieved.
The researchers’ explanation is essentially that negativity can function as a cheap source of differentiation. When people are motivated to contribute something novel, increasingly negative positions remain an easy way to distinguish themselves from what came before.
This also explains something I’ve noticed in large communities. Early discussion often feels exploratory. Later discussion increasingly becomes competitive. You’re no longer merely adding information. You’re competing for distinctiveness. And disagreement is extraordinarily good at producing it.
There is a slightly uncomfortable implication here for community design. We spend enormous amounts of time worrying about recommendation algorithms amplifying outrage. And they certainly can. But apparently you don’t necessarily need sophisticated engagement optimization to produce negativity.
Sometimes all you need is a large group of humans, enough time, and the instruction: “Tell us something nobody else has said.”
We’ll handle the rest ourselves.