Mining real-time social media big data to monitor HIV: Development and Ethical Issues

挖掘实时社交媒体大数据来监测艾滋病毒：发展和道德问题

• 批准号: 9987247
• 负责人: Sean Young
• 金额: $ 79.46万
• 依托单位: UNIVERSITY OF CALIFORNIA-IRVINE
• 依托单位国家: 美国
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-04-01 至 2021-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9987247
• 关键词: AIDS prevention; Address; Affect; Algorithms; Artificial Intelligence; Behavior; Behavioral; Benefits and Risks; Big Data; Big Data Methods; Centers for Disease Control and Prevention (U.S.); Characteristics; Code; Communities; Computer software; Computers; Continuity of Patient Care; County; Data; Data Analyses; Data Collection; Data Engineering; Development; Drug usage; Epidemiology; Ethical Issues; Ethics; Facebook; HIV; HIV diagnosis; HIV risk; Hand; Health; Health Personnel; Human; Incidence; Individual; Informed Consent; Infrastructure; Interview; Label; Learning; Letters; Machine Learning; Manuals; Measures; Methods; Mining; Minority; Modeling; Monitor; Moods; Natural Language Processing; Outcome; Participant; Population; Privacy; Public Health; Research; Research Personnel; Risk; Risk Behaviors; Scientist; Site; Social Characteristics; Social Interaction; Socioeconomic Status; Structure; Technology; Testing; Text; Time; Twitter; Work; behavior observation; combat; computer science; design; health organization; high risk; high risk sexual behavior; insight; interest; machine learning algorithm; novel strategies; psychologic; public health priorities; social; social media; software development; statistics; tool

Social “big data” holds information with wide-ranging implications for addressing issues along the HIV care continuum. Social big data refers to information from social media and online platforms on which individuals and communities create, share, and discuss content. One in four people worldwide, or over a billion people, are publically documenting their activities, intentions, moods, opinions, and social interactions on these sites. They are doing so with increasing volume and velocity, including 400 million “tweets” per day on Twitter and 4.75 billion content items shared per day on Facebook. With an increasing number of these platforms supporting access to publicly-available user data, social big data analysis is a promising new approach for attaining organic observations of behavior that can be used to monitor and predict real-world public health problems, such as HIV incidence. New tools such as social data are therefore needed to supplement existing HIV data collection methods. In preliminary research, our team developed the first approach that identifies psychological and behavioral characteristics from social big data (>550 million tweets) found to be associated with HIV diagnoses. Since groups at the highest risk for HIV (e.g., minority populations) are the fastest growing Twitter users, and because social media users have been found to publicly share personal information, we identified and collected tweets suggesting HIV risk behaviors (e.g., drug use, high-risk sexual behaviors, etc.) and modeled them alongside CDC statistics on HIV diagnoses. We found a significant positive relationship between HIV- related tweets and county-level HIV cases, controlling for socioeconomic status measures and other variables. The problem is that this approach is not currently scalable for use by HIV researchers and public health organizations. Although public health agencies are interested in mining social data to address HIV, current tools are not accessible to most health scientists, as the tools require advanced computer science expertise. For example, analyzing 500 million tweets a day requires expertise in big data engineering, advanced machine learning, natural language processing, and artificial intelligence. Developing a single platform for mining social data that has been designed and tested by and for HIV researchers could provide a significant impact on HIV prevention, testing, and treatment. We seek to create a single automated platform that collects social media data; identifies, codes, and labels tweets that suggest HIV-related behaviors; and ultimately predicts regional HIV incidence. Because of the potential ethical issues associated with mining people's data, we also seek to interview staff at local and regional HIV organization and participants affected by HIV to gain their perspectives on the ethical issues associated with this approach. The software developed from this application will be shared with HIV researchers and health care workers to provide additional tools that can be used to combat the spread of HIV.

社会“大数据”具有对解决艾滋病毒护理问题的广泛含义的信息 连续。社交大数据是指社交媒体和在线平台的信息 社区创建，共享和讨论内容。全球四分之一的人，或超过十亿人 正在公开记录他们在这些网站上的活动，意图，情绪，观点和社交互动。 他们的体积和速度增加，其中包括每天4亿个“推文”，在Twitter上 每天在Facebook上共享47.5亿个内容项目。随着这些平台的越来越多 支持访问公开可用的用户数据，社交大数据分析是一种有希望的新方法 获得可用于监视和预测现实世界公共卫生的行为的有机观察 问题，例如HIV事件。因此，需要新工具（例如社交数据）来补充现有的 HIV数据收集方法。 在初步研究中，我们的团队开发了第一种识别心理和行为的方法 社交大数据（> 5.5亿推文）的特征发现与HIV诊断有关。自从 艾滋病毒风险最高的群体（例如少数民族人口）是增长最快的Twitter用户，并且 由于已经发现社交媒体用户公开共享个人信息，因此我们确定并 收集的推文暗示了艾滋病毒风险行为（例如，毒品使用，高风险性行为等）并建模 他们与CDC统计数据有关HIV诊断。我们发现HIV- 相关的推文和县级艾滋病毒案件，控制着社会经济地位措施和其他变量。 问题在于，这种方法目前无法扩展用于HIV研究人员和公共卫生 组织。尽管公共卫生机构有兴趣挖掘社交数据以解决艾滋病毒，但目前 大多数健康科学家都无法使用工具，因为工具需要高级计算机科学专业知识。 例如，每天分析5亿条推文需要大数据工程，高级机器的专业知识 学习，自然语言处理和人工智能。开发一个用于采矿社会的平台 由HIV研究人员设计和测试的数据可能会对HIV产生重大影响 预防，测试和治疗。我们试图创建一个收集社交媒体的单个自动化平台 数据;标识暗示与HIV相关行为的推文，标识，代码和标签；并最终预测区域 艾滋病毒事件。由于与采矿人员的数据相关的潜在道德问题，我们也试图 面试当地和区域艾滋病毒组织的人员以及受艾滋病毒影响的参与者以获得他们的观点 关于与这种方法相关的道德问题。该应用程序开发的软件将是 与艾滋病毒研究人员和卫生保健工作者共享，以提供可用于战斗的其他工具 艾滋病毒的传播。