Designing Intelligent Interfaces to Facilitate Communication in Large-scale Live Streaming

Abstract

Live streaming is a type of medium that integrates video, text chat, and even more modalities to support real-time communication. It has become increasingly popular and accessible in recent years. Large-scale live streams face challenges in both real-time text chat and friendship development in the community. My dissertation explores the design space of intelligent interfaces to facilitate communication in large-scale livestreams.I conducted a qualitative interview study that investigated the motivation, practices, and challenges people face when engaging with the streamer and peer viewers on social network sites. Based on the findings, I applied the similarity-attraction effect from social psychology and proposed a subgrouping method that generates subgroups for viewers based on their similarities. Then I designed and implemented Chatbuddies, an intelligent interface that generates subgroups and interactively visualizes viewers’ information. In a controlled within-subject lab experiment and an unmoderated online experiment, participants reported that chats in subgroups were easier to follow and more enjoyable than chats in a channel without subgroups. They perceived peer viewers as more attractive in a similarity-based subgroup than in random groups. This dissertation contributes to the understanding of challenges in large-scale communication and provides design implications for future AI-embedded CMC.