This archive presents the canonical formulation of curiosity‑driven attention within a human‑aligned decision architecture. The work defines attention as a bounded function of internal curiosity and external sensory salience, providing a stable mechanism for processing information under uncertainty. Rather than modifying model internals, the framework is designed as a coordination layer that supports interpretability, reduces instability, and improves collaboration across human‑AI and multi‑agent systems. This release is shared as an early research artifact to support open discussion, interdisciplinary collaboration, and future empirical validation.