PODIUM: Procuring opinions from diverse users in a multi-dimensional world

The procurement of opinions is an important task in many contexts. When selecting members of a certain population to ask for their opinions, diversity inside the selected subset is a central consideration. People with diverse profiles are assumed to provide a wider range of opinions and thus to better represent the opinions of the entire population. However, in platforms with a large user base such as crowdsourcing applications and social networks, defining and realizing notions of diversity are both nontrivial. The profiles of users typically contain information that is high-dimensional and semantically rich. We present PODIUM, a tool for opinion procurement that accounts for complex user profiles and enables customizable user selection. Beyond selecting a subset of users with diverse profiles, PODIUM produces explanation for the choice of each user and visual AIDS to compare the selected subset to the entire population on different dimensions. We demonstrate the use of PODIUM on the TripAdvisor user base, which further enables us to examine the ability of our system to predict diverse opinions in user reviews.