In the near future more data collection devices will exist that monitor an individual's actions and interests. The data will be important for creating user models that allow applications to personalize their interactions with the user. For instance, in the field of personalized newspapers, articles will be selected based on what may be interesting or important to individual readers. Independent of the application domain (news, television, etc.) a problem of distrust is likely to ensue if users do not have access to their own user model and do not understand how their user model works.
The PeerGlass architecture for personalized news visualizes user models in two ways: totem poles and model planes. Totem poles are graphical widgets that show the user the strongest matches between an article and the user's model. The four sides of the totem pole describe the topic, source, nature, and type of news in the associated article. The totem pole is intended to help the user understand why an article was selected. The second PeerGlass mechanism for visualizing user models is a set of model planes for each section of the newspaper. Each plane shows a distinct angle of the user model. There are different model planes for explicit preferences, implicit preferences, community influences, editorial sources, etc. Both of these mechanisms for visualizing user models should be extensible to other domains. PeerGlass has undergone a heuristic evaluation and a usability study.