CRII: HCC: Practical Steps Toward Integrating the Tools of Emergency Management with Crisis Informatics Techniques

该项目旨在改进应急管理者如何利用经历自然灾害和其他危机的公民创建的社交媒体帖子和其他信息。这些信息可以帮助记录正在展开的危机并揭示需求；如果使用得当，这可以帮助应急管理人员更好地评估危机情况，并改进他们关于在哪里以及如何应对的决策。然而，尽管最近在使用人工智能技术分析公民创建的信息的工具方面取得了进展，但这些工具尚未有效地应用于实际的应急情况。该项目的目标是通过分析应急人员的做法和培训，找出这些做法与现有信息分析工具的技术能力之间的差距，并通过在设计过程中拉近应急管理人员和工具开发人员的距离，努力缩小差距，从而使有效部署更加可行。项目组计划与应急管理组织密切合作，这种合作有可能挽救生命和财产，减少痛苦和社区危机的影响。这项研究将分两个主要阶段进行。首先，项目组将调查和采访应急管理从业者的日常技术生活。通过评估其各自领域(包括执法、消防科学、国土安全和紧急医疗服务)的应急管理技能，该团队将为其他研究人员提供有价值的背景，帮助他们在哪里找到利益相关者、合作者和发展空间。该团队还将收集和分析应急管理教学大纲，以了解如何教授应急管理学生使用技术。在第二阶段，项目组将与应急管理人员一起进行参与性设计练习，探索利用机器学习和信息检索的工具如何适用于他们的实践，使用模拟实际分析准确度的方法，以说明这些工具中不可避免的错误存在及其对人们在人在回路系统中工作能力的影响。在开发参与式设计材料时，项目组将评估TF-IDF、主题建模、上下文中的关键字等技术以及它们所依赖的基础工具和数据集，以了解它们在不同于培训时的环境中安装、维护和应用的情况。这两个阶段都将着眼于许多背景，使研究团队能够评估跨信息来源、人口和不同类型的灾难事件的结果的一致性。该项目由以人为中心的计算和既定的激励竞争研究计划(EPSCoR)共同资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This project seeks to improve how emergency managers can make use of social media posts and other information created by citizens experiencing natural disasters and other crises. This information can help document unfolding crises and reveal needs; used well, this could help emergency managers better assess crisis situations and improve their decisions about where and how to respond. However, despite recent advances in tools that use artificial intelligence techniques to analyze citizen-created information, these tools have not been effectively deployed in actual emergency response situations. This project’s goal is to make effective deployment more feasible, through analyzing the practices and training of emergency responders, identifying gaps between those practices and the technical capabilities of existing information analysis tools, and working to close the gaps through bringing emergency managers and tool developers closer together in the design process. The project team plans to work closely with emergency management organizations, a collaboration that has the potential to save lives and property, reducing suffering and the impact of crises in communities. The research will proceed in two main phases. First, the project team will survey and interview emergency management practitioners about their everyday technological life. By evaluating the skills of emergency management across its respective domains (including law enforcement, fire science, homeland security, and emergency medical services), the team will provide valuable context to other researchers about where they can find stakeholders, collaborators, and space for development. The team will also collect and analyze emergency management syllabi in order to understand how emergency management students are taught to use technologies. In the second phase, the project team will conduct participatory design exercises with emergency managers that explore how tools that leverage machine learning and information retrieval could fit into their practices, using methods that simulate realistic levels of analysis accuracy in order to account for the inevitable presence of error in these tools and its effect on people’s ability to work in human-in-the-loop systems. In developing the participatory design materials, the project team will evaluate techniques like TF-IDF, Topic Modelling, Keyword-in-Context, and the underlying tools and datasets they depend on, in terms of how well they can be installed, maintained, and applied to contexts different than those they were trained on. Both phases will look at a number of contexts, allowing the research team to evaluate consistency of results across sources of information, populations, and disparate kinds of disaster events. This project is jointly funded by Human Centered Computing and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.