Concept · part of Groups, splits and scale

Picture two reasonable people. Each trusts their own side a little more, and each sensibly gives less weight to evidence from a source they trust less. Nobody is lying and nobody is stupid. Over time they drift apart anyway, and the drift makes each trust the other even less next round.

O’Connor and Weatherall formalized this loop. Start with a mild difference between two communities, add the habit of weighing evidence by how much you trust its source, and the communities diverge. It needs no irrationality and no misinformation. The trap runs on good faith.

The consequence for the book is uncomfortable. Building a channel between two distrustful groups can speed up the split, so “just connect them” fails. The loop also eats referees: once a community discounts a curation institution, that institution is one more outside voice.

The book ties the loop to what happens inside one head, the curse of expertise, and argues that one cure serves both: deep specialists with trained flexibility. It also asks for institutions built to work on partial trust. The opposite warning is the Zollman effect, where too much connection hurts accuracy.

Read the full argument

How it connects


← Echo chambers vs. bubbles (Nguyen) · Groups, splits and scale (4 of 7) · The Zollman effect →

Explore all the ideas · Read the book in order · Open the interactive map