An AI tutoring system is designed to teach students how to solve multi-step algebra problems. The system's main objective is to model and reward correct problem-solving methodology, not just getting the final answer right. The system generates a solution path, and an automated evaluator must score its quality. The evaluator can either: A) check only if the final answer is correct, or B) check the validity of each individual algebraic step in the solution. Analyze the strengths and weaknesses of each evaluation approach (A and B) in achieving the system's primary objective.