Concept · part of When the map moves the territory

Suppose one doctor tells you a drug cuts your risk of a heart attack in half. A second tells you it lowers your risk from two in a hundred to one in a hundred. Both are right. You only see what each framing was doing to you once you have both.

A single framework fails silently. From inside it, the things it hides look absent, with nothing to mark the gap. The practical defense, across every field that has studied this, is to use more than one and watch where they disagree, because the disagreement shows the shape of each. Forecasters combine models. Careful AI evaluators report results across many prompt formats (prompt-format-sensitivity). Careful institutions keep some measures away from targets and swap them over time (Goodhart’s law). Good reviewers question the categories as well as the conclusions (torque).

The book calls the practice framework rotation, and it matches the book’s larger answer. A person with deep training in one field and the habit of stepping outside it is practicing rotation alone (versatile-expertise). An institution built so that people who start from different assumptions can all check its work is practicing it at scale (survivable-polarization). The aim is modest: a second view, not a view from nowhere.

Read the full argument

How it connects


← QWERTY and the disputed lock-in story · When the map moves the territory (17 of 17)

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