How populations use social media in disasters is essential for evaluating the representation of subpopulations while analyzing social media data for emergency response and disaster research. Existing machine learning models can extract, characterize and make sense of digital trace data from social media, but are unable to account for diversity in population groups and use of social media. Consequently, the reliability of their decision-making ability remains questionable. This paper presents an exploratory analysis of empirical household survey data on the information seeking, sharing activity, and perceptions of information reliability on social media platforms across different population groups during three major hurricane storm events in the United States between 2017 and 2018. The results of this analysis suggest significant associations between social media use and socioeconomic factors: (1) Socioeconomic factors along with geographic effects play a role in determining not only platform uptake but both motivations for information seeking and the action of information sharing on social media, (2) The type of social media platform influences the type of information people seek, (3) Households from lower socioeconomic and minority backgrounds were more likely to use social media platforms to sought out different information on social media than their peer, (4) perceptions of information reliability are also influenced by social divides, where households in rural areas, lower income groups, and racial minorities were more likely to report greater unreliability in social media information. These findings provide new insights into the roles of social media use in creating or dismantling the digital divide during disasters.

人口如何在灾难中使用社交媒体对于评估亚人群的代表性，同时分析社交媒体数据以进行应急响应和灾难研究至关重要。现有的机器学习模型可以从社交媒体中提取，表征和理解数字跟踪数据，但无法解释人口群体的多样性和社交媒体的使用。因此，他们决策能力的可靠性仍然值得怀疑。本文对2017年至2018年美国发生的三场重大飓风暴风雨期间在不同人群中社交媒体平台上的信息搜寻，共享活动和信息可靠性感知进行了实证家庭调查数据的探索性分析。