Step 1: A rational agent is defined by a performance measure. Rationality means acting to do as well as possible on that measure.
Step 2: For each possible sequence of percepts, a rational agent picks the action that is expected to maximize its performance measure, given the evidence and its built-in knowledge.
Step 3: It is not limited to explicit human commands (A), it does not optimize memory (B), and learning (D) is only one way to improve, not the selection rule itself.
Step 4: So the action chosen is the one that maximizes expected performance.
Correct option: (C).