CONCEPT
The structure of worldviews and how they hold together.
A belief system is not a list of beliefs. It is a structure with load-bearing members and decorative ones, and the difference governs everything about how it behaves under pressure. A small number of core commitments hold up a much larger set of derived positions, and the system will surrender almost any derived position to protect a core one. This is why arguing against someone's stated view usually changes nothing.
Quine's image in "Two Dogmas of Empiricism" is a web. Beliefs sit at varying distances from the edge. Observation strikes the edge. When something conflicts, you have a choice about where to make the repair, and you almost always repair as far from the centre as the conflict allows.
Lakatos described the same shape in scientific research programmes: a hard core held immune by decision, wrapped in a protective belt of auxiliary assumptions that get modified when predictions fail. The belt exists to absorb hits. Astrology has one and so does general relativity. What separates them is whether the modifications generate new predictions or only explain away old failures.
No single belief is ever refuted on its own. Evidence lands against the whole structure, and the structure decides where to take the damage.
Core commitments are not held because they are better supported. They are held because too much rests on them. Someone whose profession, social circle, and self-concept all sit on one proposition is not being irrational when they resist a clean argument against it. They are pricing in the cost of the rebuild, which is real, and which the person arguing is not paying.
This predicts what actually moves people, and it is not debate. It is a slow accumulation of anomalies with nowhere left to be filed, usually alongside a social environment in which the rebuild is survivable. The argument that lands is nearly always one the person was already halfway to making themselves.
It also predicts the arguer's failure mode. Attack the periphery and you get concession without change. Attack the core head-on and you trigger the defence you were trying to get past. The move that works is to make the anomalies visible and countable, and then stop talking.
The standard picture treats beliefs like objects. Pick one up, set one down, swap it for a better one. On that picture, changing a mind is a matter of supplying the missing fact.
Nobody operates that way, including the people most confident that they do. Coherence is an expensive constraint. A belief that would force the revision of twenty others costs more than one that forces none, and that cost is paid regardless of which belief is true. What looks like stubbornness is usually accurate accounting.
The practical consequence is a better diagnostic. If you want to know what someone actually believes, do not ask what they would defend. Ask what they would have to give up if they turned out to be wrong. That is the core. Everything they volunteered first is belt.