For almost two decades, the Internet and related technologies have made more information available to information users than they can handle. The decentralization of content creation that is a feature of Web 2.0 has only exacerbated this problem. This state of overload, combined with our tendency toward hypothesis-confirming behavior, can result in biased information selection, and threatens both civil discourse and effective decision-making. In this paper, we describe a study of a technique designed to mitigate filtering by enabling content consumers to see a greater diversity of information. The results of our experiment support the notion that the strength of people's opinions can be changed by reading relevant information, but provide only weak support for the effectiveness of categorizing information content. We discuss how the results will guide our future research and inform theory and practice.