Recommender systems often overfocus on dominant interests, neglecting diversity. Shaped introduces a method to calibrate recommendations using minimum-cost flow optimization, ensuring results reflect the breadth of user preferences. This approach improves balance and relevance, outperforming standard methods.