CHS: Small: How recommendation and explanation affect preferences in social networks

Cornell University, Ithaca NY

Investigators

Abstract

This project seeks to advance research on a novel perspective on recommendation called network-centric recommender systems. Sharing information with others is a key driver of social media activity. But, surprisingly, relatively little is known about how people make choices around what to share and to read, or whom to share with and pay attention to. Unlike traditional e-commerce and research contexts that focus on accurate prediction for individuals at a given moment, the network-centric perspective sees recommendation as a dynamic social process. Understanding more about the factors that influence recommendation processes will have practical impact on the design of social media platforms and recommender systems, leading to tools that better serve both individual and societal goals around sharing information. It will also set the stage for an understanding of how information spreads in social networks that is deeper than current diffusion models that fail to account for these factors, allowing scientists to build more accurate and more general models of sharing behavior. A series of laboratory experiments will ask people to make and take recommendations from each other. These experiments will use real social media data such as Facebook Likes and Twitter hashtag usage, but control the algorithms, interfaces, and partners people interact with in the experiment. Doing this will help weigh the importance of factors that influence people's sharing and reading behavior. Among these factors are how users are influenced by goals such as self-expression and helping people, system characteristics such as filtering algorithms and social explanations for shared items, and social forces such as similarity and trust between a sharer and a receiver.