Stability-Weighted Typicality and the Boltzmann Brain Problem

Cosmological models permitting extremely large spacetime volumes or long-lived equilibrium phases face a well-known difficulty: the predicted dominance of Boltzmann observers. Transient fluctuation observers can vastly outnumber observers embedded in extended, lawlike histories, undermining the reliability of empirical inference. This paper argues that the problem reflects a structural deficiency in how typicality is defined, not a pathology of specific cosmological scenarios. A minimal adequacy condition is introduced: a viable cosmology must render cognitively stable observers typical under any defensible probability assignment compatible with its dynamics. On this basis, the paper advances stability-weighted typicality, the requirement that typicality be evaluated over dynamically stable trajectories rather than instantaneous configurations. This proposal is distinct from worldline ordering in general relativity; it operates at the level of measure over histories, not geometric structure, and identifies a necessary condition that any adequate measure must satisfy without introducing a new probability rule or modifying the underlying dynamics. Under this constraint, Boltzmann observer dominance does not arise. Fluctuation configurations lack the dynamical organisation required to sustain the memory-environment correlations on which coherent observation depends. Stability-weighted typicality provides a principled, measure-theoretic resolution of the Boltzmann brain problem grounded in the physical conditions that make empirical reasoning possible.