CharmMe : applying machine learning to facilitate meaningful interactions at the MIT Media Lab

Abstract

CharmMe is a social discovery application to help people connect with others of similar interests at a company, organization, or conference. Unlike traditional social networking or matching algorithms, CharmMe discovers connections automatically without the need for new profiles or tagging. By using natural language processing, we create a model of an organization by "reading" existing information related to the people being matched, such as their publications or social media accounts. Additionally, the application takes data provided by users Checking-in to conference talks or Liking projects, which are actions made popular by the social networking sites Facebook and Foursquare. To facilitate the actual introduction process, the application makes available the location of all recommended people using RFID technology. In addition, possible opening topics of conversation are suggested based on similar interests shared by users. In this paper, we investigate how effective CharmMe is at predicting new connections that are desirable and describe its deployment during a conference event at the MIT Media Lab. Additionally, we evaluate the effectiveness of the recommendations provided by the system and whether results improve with incorporating user feedback. Ultimately, we think this application will help people create better relationships by encouraging purposeful interactions, eliminating certain social inefficiencies, as well as decrease the opportunity for a missed but potentially meaningful connection.