Recommender systems in online platforms for social campaigns

Online social change campaigns have recently become important channels of civic participation. However, most online change platforms tend to use simple indicators (such as popularity) to rank social campaigns, hence creating a situation where the most popular campaigns dominate the rankings and attract the most attention. Online petitions which focus on niche issues are often ranked lower below these more popular campaigns. For example, a petition for an environmental issue in a small town may not even be seen because it is just not popular enough with the majority of users on the platform. The simple ranking mechanism used by most online change platforms cannot effectively link petitions with those who are genuinely concerned with them. At most, less than 10% of online petitions get the chance to fulfill their causes. To solve this problem, we present a design of a novel recommender system that leverages social interaction features, psycholinguistic features, and latent topic features to provide a personalized ranking for different users. Hence, users are presented with better petition recommendations fitting their unique interests and concerns. We evaluate our recommender system against matrix factorization collaborative filtering and content-based filtering with the bag of words features as two baseline recommenders and achieved a 4.2% enhanced recommendation performance over the benchmark. The recommendation system has the potential to improve the user experience and participation within online change platforms. We plan to evaluate user adoption of the recommendation engine in a survey experiment and also have plans to apply our hybrid approach of recommending to additional contexts. Overall, our recommender system offers the potential for civic participations and social change by offering personalized recommendations to users based on their individual interests and concerns.