Learning to Access Computation: Accessibility Plasticity as a Principle of Adaptive Intelligence

This paper proposes Accessibility Plasticity as a distinct axis of adaptation in neural systems. Instead of changing only parameters, capabilities, or modules, a system can reorganize which existing computations are allowed to interact and participate. The authors formalize this idea through a relationship-based operational realization and propose a reuse-first adaptation hierarchy: modify accessibility before making more costly capability or structural changes. A proof-of-concept evaluation on sequential learning tasks reportedly reduced capability modification while maintaining comparable task performance, although the abstract provides no quantitative metrics.