Identifying drug and alcohol displays on social media using a machine learning approach, and mechanisms that impact adolescent substance use

项目总结 此应用程序的总体目标是支持和促进必要的培训以开发 独立研究生涯专注于检验线上和线下影响青少年的社会背景 物质使用(SU)。从长远来看，申请者寻求开发一项专注于使用的研究计划 数字技术研究导致青少年酗酒和吸毒的社会风险因素 提供预防规划。通过培训目标、指导性指导和互补性 经验，拟议的项目将战略性地促进申请者对社交媒体和SU的了解 研究。申请者还将接受最先进的量化方法的培训，以增强设计， 进行大数据分析，以提高我们目前对社会化对青少年苏的影响的理解。 预防青少年饮酒和吸毒仍然是一个关键的研究领域，因为 酒精和毒品的使用可能会导致长期的负面后果。事实上，十分之九的人 经历过SU问题的人在18岁之前就开始使用(CASAColumbia，2011)。在印心的时候 SU通常发生在青少年时期，青少年也花了大量的时间 使用社交媒体。青少年使用社交媒体作为与社交网络连接的一种方式，以及查看和 显示SU行为。然而，评估社交媒体对青少年SU影响的研究一直在进行 研究不足，现有的研究有几个方法论上的局限性。也就是说，先前的工作具有 主要关注大学样本，以及在青少年中不太受欢迎的社交媒体平台(即， Facebook)。此外，以前的工作使用自我报告数据或人工编码来评估在线SU内容。这个 拟议的研究旨在提高对社交媒体作用的认识，特别是接触SU 电子烟内容和用户生成的电子烟内容在青少年中酒精和毒品使用的升级中。 作为培训奖学金中心的拟议项目包括两个目标，这两个目标提出了次要目标 数据分析，以及原始数据的收集。拟议项目的目标1(n=243)将使用第二 数据分析确定SU态度、主观规范和知觉行为控制是否起中介作用 接触同行发布的SU相关内容与上有影响力的人物之间的预期关联 离线SU行为。拟议项目的目标2(n=200)将前瞻性地审查两者之间的联系 使用一种新的方法在Instagram上在线用户生成电子烟内容和离线使用电子烟 接近。将开发机器学习算法来检测Instagram个人资料中的电子烟内容，以及 将与自我报告数据进行比较，以评估这种联系的强度是否因方法而异。 这个项目将增加对青少年SU行为的了解，这些行为在线下看不到，这可能是独一无二的 通过确定与当今青年相关的可修改的干预目标，为预防计划提供信息。

PROJECT SUMMARY The overall goal of this application is to support and facilitate the necessary training to develop an independent research career focused on examining online and offline social contexts that influence adolescent substance use (SU). In the long-term, the applicant seeks to develop a program of research focused on the use of digital technologies to examine social risk factors that lead to alcohol and drug use among adolescents and to deliver prevention programming. Through the training goals, guided mentorship, and complementary experiences, the proposed project will strategically advance the applicant's knowledge of social media and SU research. The applicant will also be trained in state-of-the-art quantitative methodologies to enhance the design, conduct, and analysis of big data to improve our current understanding of socialization effects on adolescent SU. The prevention of alcohol and drug use among adolescents remains a critical area for research, as experimentation with alcohol and drug use can lead to long-term negative consequences. In fact, 9 out of 10 people that experience SU problems began using before the age of 18 (CASAColumbia, 2011). While initiation of SU typically occurs during the teenage years, adolescents are also spending a substantial amount of time using social media. Adolescents use social media as a way to connect with a social network, as well as view and display SU behaviors. However, research evaluating the impact of social media on adolescent SU has been understudied, and the available research has several methodological limitations. Namely, prior work has primarily focused on college samples, as well as a less popular social media platform among adolescents (i.e., Facebook). Furthermore, prior work has used self-report data or human coding to assess online SU content. The proposed study seeks to advance the knowledge regarding the role of social media, specifically exposure to SU content, and user-generated e-cigarette content, in the escalation of alcohol and drug use among adolescents. The proposed project at the center of this training fellowship includes two aims that propose secondary data analyses, and the collection of original data. Aim 1 of the proposed project (n = 243) will use secondary data analyses to determine whether SU attitudes, subjective norms, and perceived behavioral control mediate the prospective association between exposure to SU-related content posted by peers and influential figures on offline SU behaviors. Aim 2 of the proposed project (n = 200) will prospectively examine the association between online user-generated e-cigarette content on Instagram and offline e-cigarette use using a novel methodological approach. Machine learning algorithms will be developed to detect e-cigarette content on Instagram profiles, and will be compared to self-report data to assess whether the strength of this association varies based on approach. This project will increase knowledge of adolescent SU behaviors that are not visible offline that could uniquely inform prevention programs by identifying modifiable targets for intervention that are relevant for today's youth.