Maximizing Expected Utilities in the Prisoner's Dilemma