This paper analyzes the dynamic interaction between a fully rational, privately informed sender and a boundedly rational, uninformed receiver with memory constraints. The receiver designs an information-processing and decision-making protocol, modeled as a finite-state machine, that governs how information is interpreted, how internal memory states evolve, and when and what decisions are made. We characterize optimal protocols that balance learning with robustness to strategic manipulation and quantify the extent of both learning and manipulation through payoff bounds for each party. A simple class of protocols disciplines the sender and gives rise to behavioral patterns such as opinion polarization and decision avoidance. The model offers an expressive framework for strategic learning and decision-making under asymmetric rationality, with applications to regulatory review and media distrust.