People increasingly keep up with the "newest news" through information streams (such as Twitter). To alleviate information overload and better direct user attention, we explored dimensions for designing a recommender system that selects promising subsets of content for consideration, models user topic interest, and leverages social interaction processes. The best performing algorithm -- implemented as a prototype web-based tool -- improved the percentage of interesting content to 72% (from a baseline of 33%). The competencies and results of this work can be generalized to other enterprise and consumer information streams.