The paper tackles a question about how symmetry groups and probability measures interact in a mathematically precise setting. When you have a group of permutations acting on a countable set, you can study the collection of all probability measures that are "invariant" under that action, meaning the measure does not change when you apply a permutation. These invariant measures form a geometric object called a simplex, and a classical question is what shape that simplex takes. There are two extreme cases: a "Bauer" simplex, where the extreme points (the irreducible building blocks) are nicely separated and closed, and a "Poulsen" simplex, where the extreme points are dense in the whole simplex, making things much more tangled.
The main result is a clean dichotomy: under a certain rigidity condition on the permutation group called relative Property (T), the simplex of invariant measures is always one of these two extreme types, and you can tell which one based on whether the full group satisfies a stronger rigidity condition called Property (T). Roughly speaking, Property (T) is a way of saying a group is "stiff" or hard to deform, and the paper shows this stiffness is exactly what determines the geometric shape of the associated space of measures. This resolves open questions and examples that had been raised in earlier work by Austin.
To prove this, the authors develop new machinery in an area called affine logic, which is a version of mathematical logic adapted to study structures involving probability and averaging rather than just discrete yes-or-no relations. They introduce and study "existentially closed" models in this setting, which are models that are as rich as possible in a specific technical sense. This model-theoretic framework is presented as a contribution of independent value, likely to be useful for other problems sitting at the intersection of logic, ergodic theory, and the study of topological groups.