Learning Pareto-Efficient Decisions with Confidence