经济与社会科学2024link.springer.com
不遭受基于公开可用数据的 AI 画像的权利
The right not to be subjected to AI profiling based on publicly available data
入门 · 约 35 分钟读完前置知识:隐私法基础、数据画像概念
导读
这篇论文探讨是否应确立一种新权利:个人有权不被基于公开可得数据进行的 AI 画像所评判。文章论证了 AI 画像带来的新型伤害,主张现有隐私法不足以应对,需要新的法律权利来保护个人免受自动化画像的影响。
HN 关注点:HN 读者长期关注隐私与 AI 伦理议题,讨论焦点在于:当数据本身是公开的,法律是否应该、以及能否限制基于这些数据的画像行为。
HN 评论精选
这几条评论里:支持 1 质疑 1 补充 1 亲历 1
支持隐私保护过去靠监控成本高昂这一事实,如今成本骤降,需要新法律应对
我认为隐私的量变会带来质变。上世纪70年代警察当然也能在公共场合跟踪你,但全天候监控需要6名以上警察,年成本高达六位数。普通人的隐私其实不是靠法律或隐私理念保护的,而是靠简单的数量限制——警察根本不够用,连0.1%的人口都盯不过来。但在智能手机、车牌识别和征信机构的时代,全天候跟踪的成本降低了几个数量级,我们确实需要新的法律和权利来反映这一新现实。
michaelt查看英文原文
I think there are qualitative differences that arise from incremental, quantitative changes in privacy. Sure, in the 1970s the cops could follow you around any time you were out in public. But they needed 6+ cops to provide round-the-clock surveillance, with an annual cost in the high six figures. So the average citizen's privacy was protected not by law or high-minded ideals about privacy, but by simple numbers. There weren't enough cops to follow even 0.1% of the population. But in the modern age of smartphones and license plate recognition and credit reference agencies? I think there's a strong argument to be made that we need new laws and new rights to reflect the new reality, where tracking people around the clock is several orders of magnitude cheaper.
亲历亲历:AI 画像可能严重失准,若被雇主采用会造成实际伤害
一位合作伙伴给我看了他们的 CRM 工具,其中 AI 组件为每个联系人建立了画像,信息相当完整,明显来自 LinkedIn 等来源。人格总结部分令人担忧:某些方面非常准确,另一些则错得离谱。最让我在意的是它说我'厌恶风险'、'难以在不完整信息下做决策'——而我做了35年创业和独立承包工作,风险承受力和模糊决策正是我的强项。更糟的是,如果这份画像被潜在雇主看到,可能会直接毁掉机会。这就像被 MBTI 测试结果审判一样。
Ensorceled查看英文原文
A partner showed me their CRM tool, it had an AI component that created a profile of each of their contacts. It was pretty complete and clearly had info from LinkedIn and other sources. The personality summary was concerning as it was both deeply accurate in some respects and deeply inaccurate in others. Most concerning, it said I was "risk adverse" and "struggles to make decisions with incomplete data". With 35 years at startups and independent contracting, risk tolerance and ability to make decisions with incomplete data are kind of in my wheelhouse. Worse, if this profile was being shown to potential employers, it could (would?) be a deal breaker. It's kind of like being judged by your MBTI results.
补充呼吁将此类权利扩展到不透明的信用评分体系
我希望这项权利也能适用于征信机构,它们从不公布信用评分的计算方法。除了众所周知的粗略框架,没人知道具体机制是什么,而这些评分却在实实在在地影响人们的生活。
epgui查看英文原文
I hope this right gets applied to credit scoring agencies as well, which don't publish the methodology in the calculation of credit scores. God knows what the exact mechanics of these calculations are (beyond the well-known but very hand-wavy broad strokes), and they impact people in serious ways.
质疑质疑:数据已公开则画像无法阻止,应聚焦数据删除与安全权利
恕我直言,这是个愚蠢的想法:如果数据已经在别人手里,你根本无法阻止画像。真正需要保护的是个人数据的删除权和数据安全。这种权利无异于马跑出马厩后才去关门。
faeriechangling查看英文原文
This is an absolutely stupid notion IMHO because if somebody has the data there’s NOTHING you can do to stop profiling. The right to the deletion and security of your own personal data is what needs to be protected. This sort of right is closing the barn door after the horse leaves.
相关资源
从这篇论文所有 HN 讨论帖的全部评论里整理出来的链接。
历次讨论
- 2024Hacker News 讨论帖277 分 · 212 评
同领域还值得读
一切的回报率,1870–2015
The Rate of Return on Everything, 1870–2015
利用1870–2015年多国的长期数据,估算股票、住房等各类资本的回报率,并将其与经济增长率进行比较。
经济与社会科学2019 · 487 分 · 260 评进阶 · 约 100 分钟编辑精选一切资产的回报率,1870–2015
The Rate of Return on Everything, 1870–2015
构建了 1870–2015 年 16 个国家的长周期资产回报数据集,覆盖股票、住房、政府债券和短期票据,系统比较各类资产的真实回报与波动性。
经济与社会科学2018 · 215 分 · 151 评进阶 · 约 120 分钟编辑精选一切资产的回报率,1870–2015(2019)
The Rate of Return on Everything, 1870–2015 (2019)
这篇发表于《经济学季刊》的论文构建了 1870–2015 年间 16 个主要发达国家的股票、债券、房地产等资产收益率的长周期数据库。
经济与社会科学2024 · 190 分 · 148 评进阶 · 约 80 分钟编辑精选想法是否越来越难找?
Are ideas getting harder to find?
探讨“想法是否越来越难找到”:通过衡量多个领域的“研究生产率”,发现维持相同的增长需要投入指数级增长的研究资源,即研究效率在持续下降。
经济与社会科学2016 · 182 分 · 106 评入门 · 约 60 分钟编辑精选比特币与区块链的经济极限
The Economic Limits of Bitcoin and the Blockchain
芝加哥大学 Eric Budish 的论文从经济学角度分析比特币与区块链的根本限制:核心论点是只要攻击网络(双花或砸盘做空)的收益超过挖矿成本,51% 算力攻击就有利可图,而挖矿…
经济与社会科学2018 · 176 分 · 37 评进阶 · 约 90 分钟编辑精选