论文标题
各个年龄段的隐私政策：1996 - 2021年隐私政策的内容和可读性
Privacy Policies Across the Ages: Content and Readability of Privacy Policies 1996--2021
论文作者
论文摘要
众所周知，大多数用户没有阅读隐私政策，但是几乎所有用户都会勾选盒子以同意它们。在本文中，我们使用透明度研究，机器学习和自然语言处理的方法分析了25年的隐私政策历史。具体而言，我们从1996年至2021年收集了大规模的隐私政策纵向语料库，并根据其描述的数据实践，授予用户的权利以及为组织保留的权利分析隐私政策的长度和可读性及其内容。我们特别关注响应最近的隐私法规（例如GDPR和CCPA）的变化。我们的结果表明，政策越来越长，越来越难以阅读，尤其是在新法规生效之后，我们发现了一系列有关数据实践的范围。我们的结果使我们能够推测为什么很少阅读隐私政策并提出更改，从而使隐私政策为其读者而不是作家提供服务。
It is well-known that most users do not read privacy policies, but almost all users tick the box to agree with them. In this paper, we analyze the 25-year history of privacy policies using methods from transparency research, machine learning, and natural language processing. Specifically, we collect a large-scale longitudinal corpus of privacy policies from 1996 to 2021 and analyze the length and readability of privacy policies as well as their content in terms of the data practices they describe, the rights they grant to users, and the rights they reserve for their organizations. We pay particular attention to changes in response to recent privacy regulations such as the GDPR and CCPA. Our results show that policies are getting longer and harder to read, especially after new regulations take effect, and we find a range of concerning data practices. Our results allow us to speculate why privacy policies are rarely read and propose changes that would make privacy policies serve their readers instead of their writers.
