Characterizing Regional and Behavioral Device Variations Across the Twitter Timeline: A Longitudinal Study

Abstract

Physical devices, such as smartphones and laptops, provide the key interface through which users engage with the social media world. Yet despite the broad range of devices used on social media platforms, relatively little is known about how usage varies on a device to device level. In this work, we use a 10 sample of Twitter data spanning two consecutive years and encompassing 365.98 million users to perform a longitudinal, measurement-driven analysis of five prevalent device types - Android, iPhone-iOS, BlackBerry, other mobile devices, and nonmobile devices - across the time period. We study the global and regional usage patterns, as well as the regional distribution, of devices; investigate differences and similarities in behavioral patterns across devices with respect to tweet sentiment, daytime usage patterns, and feature usage (such as mentions, URLs, hashtags) over time; and quantify the level of "device homophily" (i.e., assortativity) within the Twitter device network. Our results reveal that key variations exist among these device groups, in addition to notable similarities. To the best of our knowledge, this is the first large-scale longitudinal analysis of various distinct Twitter devices.