End-User Audits: A System Empowering Communities to Lead Large-Scale Investigations of Harmful Algorithmic Behavior

End-User Audits: A System Empowering Communities to Lead Large-Scale Investigations of Harmful Algorithmic Behavior
复制标题

最终用户审计:一个使社区能够领导对有害算法行为进行大规模调查的系统

DOI:
--
复制
发表时间:
2022
期刊:
Proc. ACM Hum. Comput. Interact.
影响因子:
--
通讯作者:
Michael S. Bernstein
Michael S. Bernstein
中科院分区:
--
文献类型:
--
作者:
Michelle S. Lam;Mitchell L. Gordon;D. Metaxa;Jeffrey T. Hancock;J. Landay;Michael S. Bernstein

文献摘要

参考文献

被引文献

相似文献

因为算法审计是由技术专家进行的,所以审计必然限于专家认为要测试的假设。最终用户承诺扩大这一范围,因为他们居住在空间中,并见证了审计师不会看到的算法影响。为了追求这一目标,我们提出了最终用户审计--由非技术用户领导的系统规模审计--并提出了一种在假设生成、证据识别和结果交流方面为最终用户搭建脚手架的方法。今天,执行系统规模的审计需要用户花费大量精力来标记数千个系统输出,因此我们引入了一种协作过滤技术,该技术利用算法系统自己的分类训练数据来将少量最终用户标签投影到完整的测试集上。我们的最终用户审计工具IndieLabel使用了这些预测标签,因此用户可以快速探索他们的意见与算法系统的输出有何不同。通过突出显示系统对用户表现不佳的主题领域,并显示可能的错误案例集,该工具可指导用户编写审计报告。在对有17名非技术参与者参与的广受欢迎的评论毒性模型的最终用户审计进行的评估中,参与者既重复了正式审计以前发现的问题,也提出了以前报告不足的问题,如对延续污名的隐蔽形式的仇恨标记不足,以及对被边缘化社区回收的诽谤标记过多。
Because algorithm audits are conducted by technical experts, audits are necessarily limited to the hypotheses that experts think to test. End users hold the promise to expand this purview, as they inhabit spaces and witness algorithmic impacts that auditors do not. In pursuit of this goal, we propose end-user audits-system-scale audits led by non-technical users-and present an approach that scaffolds end users in hypothesis generation, evidence identification, and results communication. Today, performing a system-scale audit requires substantial user effort to label thousands of system outputs, so we introduce a collaborative filtering technique that leverages the algorithmic system's own disaggregated training data to project from a small number of end user labels onto the full test set. Our end-user auditing tool, IndieLabel, employs these predicted labels so that users can rapidly explore where their opinions diverge from the algorithmic system's outputs. By highlighting topic areas where the system is under-performing for the user and surfacing sets of likely error cases, the tool guides the user in authoring an audit report. In an evaluation of end-user audits on a popular comment toxicity model with 17 non-technical participants, participants both replicated issues that formal audits had previously identified and also raised previously underreported issues such as under-flagging on veiled forms of hate that perpetuate stigma and over-flagging of slurs that have been reclaimed by marginalized communities.
让社区了解情况:了解维基百科利益相关者对基于机器学习的系统的价值观
DOI: 10.1145/3313831.3376783
发表时间: 2020
期刊: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems
影响因子: --
作者:
Smith, C. Estelle;Yu, Bowen;Srivastava, Anjali;Halfaker, Aaron;Terveen, Loren;Zhu, Haiyi
通讯作者: Zhu, Haiyi
保守派、跨性别者和黑人社交媒体用户的不成比例的删除和不同的内容审核体验:边缘化和审核灰色区域
DOI: 10.1145/3479610
发表时间: 2021
影响因子: --
作者:
Haimson, Oliver L.;Delmonaco, Daniel;Nie, Peipei;Wegner, Andrea
通讯作者: Wegner, Andrea
审计新闻策展系统:检查 Apple News 中算法和编辑逻辑的案例研究
DOI: --
发表时间: 2020
期刊: Proceedings of the International AAAI Conference on Weblogs and Social Media
影响因子: --
作者:
Jack Bandy, Nicholas Diakopoulos
通讯作者: Jack Bandy, Nicholas Diakopoulos
迈向用户驱动的算法审计:调查用户发现有害算法行为的策略
DOI: 10.1145/3491102.3517441
发表时间: 2022
期刊: CHI '22: CHI Conference on Human Factors in Computing Systems
影响因子: --
作者:
DeVos, Alicia;Dhabalia, Aditi;Shen, Hong;Holstein, Kenneth;Eslami, Motahhare
通讯作者: Eslami, Motahhare
日常算法审计:了解日常用户在发现有害算法行为方面的力量
DOI: 10.1145/3479577
发表时间: 2021
影响因子: --
作者:
Shen, Hong;DeVos, Alicia;Eslami, Motahhare;Holstein, Kenneth
通讯作者: Holstein, Kenneth