CHS: Large: Collaborative Research: Pervasive Data Ethics for Computational Research

CHS：大型：协作研究：计算研究的普遍数据伦理

• 批准号: 1704369
• 负责人: Katherine Shilton
• 金额: $ 90.32万
• 依托单位: University of Maryland, College Park
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1704369&HistoricalAwards=false
• 关键词: CHS; Large; Collaborative; Research; Pervasive

This project promotes the progress of science and technology development by providing the empirical knowledge needed to advance fair, just computational research. Big, pervasive data about people enables fundamentally new computational research, but also raises new ethical challenges, such as accounting for distributed harms at scale, protecting against the risks of unpredictable future uses of data, and ensuring fairness in automated decision-making. National debates have erupted over online experiments, leaked datasets, and the definition of "public" data. Investigators struggle to advise students on engaging vulnerable populations or navigating terms of service. Regulators debate how to translate traditional ethical principles into workable policy guidance. Research addressing these challenges has hit roadblocks caused by a lack of empirical knowledge about emerging norms and expectations. This project discovers how diverse stakeholders - big data researchers, platforms, regulators, and user communities - understand their ethical obligations and choices, and how their decisions impact data system design and use. It also compares stakeholder perspectives against the risks and realities of pervasive data itself, answering fundamental questions about the fairness and ethics of such research. Understanding how computing researchers adapt their practices in the big data era, and highlighting points of convergence or conflict with data realities, user expectations, and regulatory practices, will produce concrete guidance for pervasive data ethics. In addition to improving ethical approaches for studying people in computing contexts, this work empowers researchers with actionable information about emergent norms and risks. Outputs, such as decision-support tools, guidance on measuring risk, public educational material and bibliographies, and reusable empirical data, are designed to support the wide range of stakeholders in data ethics. To meet these goals, this project enables a collaboratory - a virtual center combining data and analytical resources - to collect empirical data on research ethics at diverse scopes and scales. The research includes including attention to multiple ethical issues (privacy, risk, respect, beneficence, justice) as well as the full network of stakeholders involved in research ethics (user communities, computing research communities, technical platforms, and regulations). The project conducts interviews with, and surveys of, 1) user communities, 2) computing researchers, 3) data ethics regulators, and 4) commercial platform providers. The project also gathers numerous shared document sets, including 1) pervasive data research publications, 2) pervasive computing curricula and degree requirements, 3) news articles and public discourse about pervasive data research, 4) a corpus of existing data ethics training, 5) pervasive data grant summaries and data management plans, and 6) corporate ethics guidelines and regulatory documents. The project uses these resources to: discover metrics for assessing and moderating risks to data subjects; document how user attitudes and media reactions shape subjects' willingness to participate in pervasive data research; model user concerns in ways accessible to computational researchers; discover how existing ethical codes can be adapted and adopted for the real-world working conditions of sociotechnical and cyber-human research; determine how the changing practices of academic and corporate regulators impact users and researchers; and illuminate implementable and sustainable best practices for research ethics.

该项目通过提供促进公平、公正的计算研究所需的经验知识来促进科学和技术发展的进步。关于人的无处不在的大数据从根本上实现了新的计算研究，但也提出了新的伦理挑战，例如对大规模分布式损害进行核算，防范未来不可预测的数据使用风险，以及确保自动化决策的公平性。关于在线实验、泄露的数据集和“公共”数据的定义，已经爆发了全国性的争论。调查人员很难就与弱势群体接触或导航服务条款向学生提供建议。监管机构就如何将传统的道德原则转化为可行的政策指引展开了辩论。应对这些挑战的研究遇到了障碍，原因是缺乏对新兴规范和预期的实证知识。该项目了解不同的利益相关者--大数据研究人员、平台、监管机构和用户社区--如何了解他们的道德义务和选择，以及他们的决策如何影响数据系统的设计和使用。它还将利益相关者的观点与无处不在的数据本身的风险和现实进行比较，回答了有关此类研究的公正性和伦理的基本问题。了解计算研究人员如何在大数据时代调整他们的实践，并强调与数据现实、用户预期和监管实践的汇合点或冲突点，将为普适数据伦理提供具体指导。除了改进在计算环境中研究人的伦理方法外，这项工作还赋予研究人员关于紧急规范和风险的可操作信息。决策支持工具、风险衡量指南、公共教育材料和书目以及可重复使用的经验数据等产出旨在支持数据伦理方面的广泛利益攸关方。为了实现这些目标，该项目使一个协作室--一个结合了数据和分析资源的虚拟中心--能够在不同的范围和规模上收集关于研究伦理的经验数据。研究包括对多个伦理问题(隐私、风险、尊重、慈善、正义)的关注，以及参与研究伦理的利益相关者(用户社区、计算研究社区、技术平台和法规)的全面网络。该项目对1)用户社区、2)计算研究人员、3)数据伦理监管机构和4)商业平台提供商进行访谈和调查。该项目还收集了大量共享文件集，包括1)普及数据研究出版物，2)普及计算课程和学位要求，3)关于普及数据研究的新闻文章和公共讨论，4)现有数据伦理培训语料库，5)普及数据拨款摘要和数据管理计划，以及6)企业伦理指导方针和管理文件。该项目利用这些资源：发现评估和缓解数据对象风险的指标；记录用户态度和媒体反应如何影响对象参与普及数据研究的意愿；以计算研究人员可以访问的方式模拟用户关切；发现现有伦理准则如何适应和采用社会技术和网络人类研究的真实工作条件；确定学术和公司监管机构不断变化的做法如何影响用户和研究人员；以及说明可实施和可持续的研究伦理最佳实践。

项目成果

From Human to Data to Dataset: Mapping the Traceability of Human Subjects in Computer Vision Datasets

从人类到数据到数据集：在计算机视觉数据集中绘制人类受试者的可追溯性

• DOI: 10.1145/3579488
• 发表时间: 2023
• 期刊: Proceedings of the ACM on Human-Computer Interaction
• 作者: Scheuerman, Morgan Klaus;Weathington, Katy;Mugunthan, Tarun;Denton, Emily;Fiesler, Casey
• 通讯作者: Fiesler, Casey

Surveillance and the future of work: exploring employees’ attitudes toward monitoring in a post-COVID workplace

监控与工作的未来：探索员工对后疫情工作场所监控的态度

• DOI: 10.1093/jcmc/zmad007
• 发表时间: 2023
• 期刊: Journal of Computer-Mediated Communication
• 影响因子: 7.2
• 作者: Vitak, Jessica;Zimmer, Michael
• 通讯作者: Zimmer, Michael

Governing with Algorithmic Impact Assessments: Six Observations

通过算法影响评估进行治理：六项观察

• 发表时间: 2021
• 期刊: and Society (AIES
• 作者: Watkins, Elizabeth;Moss, Emanuel;Metcalf, Jacob;Singh, Ranjit;Elish, Madeleine Clare
• 通讯作者: Elish, Madeleine Clare

Taking Algorithms to Courts: A Relational Approach to Algorithmic Accountability

将算法告上法庭：算法问责的关系方法

• DOI: 10.1145/3593013.3594092
• 发表时间: 2023
• 期刊: and Transparency
• 作者: Metcalf, Jacob;Singh, Ranjit;Moss, Emanuel;Tafesse, Emnet;Watkins, Elizabeth Anne
• 通讯作者: Watkins, Elizabeth Anne

When does data collection and use become a matter of concern? A cross-cultural comparison of American and Dutch people’s privacy attitudes

数据收集和使用何时成为人们关注的问题？

• 发表时间: 2023
• 期刊: International Journal of Communication
• 影响因子: 1.7
• 作者: Vitak, J.;Liao, Y.;Mols, A. D.;Zimmer, M.;Kumar, P. C.;Pridmore, J.
• 通讯作者: Pridmore, J.