n this talk I will show that many naturally arising problems in social domains wouldallow for applications of dynamic programming algorithms (e.g., knapsack, etc.) if allrelevant parameters were known. Unfortunately, this is hardly the case in real-world sce-narios. However, the core of structured learning is often a decoding machinery that ex-ploits principles of dynamic programming. I argue that in case labeled data is available,the unknown variables can be learned by casting the problem as a structured predictionproblem. The talk revisits structured learning and dynamic programming to motivatenovel application scenarios and experimental results.