Conversations in the Eye of the Storm

Abstract

This work propels social media research beyond the single post as the unit of analysis toward fuller treatment of interaction by making the construct of the conversation analytically available. We offer a method for constructing @reply conversations in Twitter to apprehend social media conversational features at scale. We apply this method to the high-tempo, high-stakes environment of 2012's Hurricane Sandy, with its high volume of online talk by affected locals and distinct disaster-stage phasing by which to consider interactional difference. We investigate the temporality of conversations; the relationality of who speaks to whom; the number and kind of conversationalists; and how content affects temporal features. The analysis reveals that, during the height of the emergency, people expand conversations both in number and kind of conversational partners-just as their information search intensifies. This expansion contributes to longer, slower-paced conversations in the high-emergency period, suggesting reliance on online relationships during times of greatest uncertainty.