AI Alignment via Incentives and Correction