Two-moment Decision Models and Expected Utility Maximization