Priya's survey had four thousand and twelve responses, a margin of error she could defend, and a graph so tidy that her supervisor had printed it and pinned it to the wall. It showed that people in the city spent an average of thirty-four minutes travelling to work, that satisfaction rose with the length of the journey up to a point and then fell sharply, and that the worst-affected group was parents of young children. Priya had spent five months on the sample. She was proud of it, in the specific way of someone who has cleaned a very dirty file.
普里娅的调查收到了四千零一十二份回复，误差范围是她能够辩护的，图表也漂亮得让导师打印出来钉在墙上。图表显示，这座城市的人平均花三十四分钟通勤；满意度会随路程变长而上升，到某一点之后急剧下降；受影响最严重的是有小小孩的父母。她在这个样本上花了五个月。她为它骄傲，那种骄傲很具体，属于一个刚刚把一份极脏的文件清理干净的人。
The data had been collected through an app, which meant that respondents needed a smartphone and, more subtly, needed to be awake at a reasonable hour to answer the notification. Priya had noted this as a limitation in the final paragraph, in the standard language of limitations, which is designed to be read and forgotten. Then, in March, she missed her usual bus and had to wait at the depot at five in the morning, where she met a woman in scrubs eating rice from a plastic box and looking at the departures board as though it had personally insulted her.
数据是通过一个手机应用收集的，这意味着受访者得有一部智能手机，而且更微妙的是，得在合乎常理的时间醒着，才能回应那条通知。普里娅在最后一段把这一点写成了局限，用的是写局限的标准措辞——那种措辞就是为了被读到、然后被忘掉而设计的。后来在三月，她错过了常坐的那班公交，清晨五点在车场等车，遇见一位穿手术服的女士，正从一个塑料盒里吃饭，盯着发车时刻表，好像那张表得罪了她本人。
The woman's name was Blessing, and she had been a night nurse for six years. Her journey took fifty minutes on a good night and two hours on a bad one, because the first bus did not run until half past five. She had not received Priya's survey. She did not have the app. She described her commute with a precision that made Priya's categories look like children's drawings: the walk, the wait, the change at the market, the driver who sometimes stopped for her and sometimes did not, and the hour of sleep she took before her children came home from school.
那位女士叫布莱辛，做了六年夜班护士。顺利的夜晚，她通勤要五十分钟；不顺利的时候要两小时，因为第一班公交五点半才发车。她没有收到普里娅的问卷。她没有装那个应用。她描述自己通勤的方式精确得让普里娅的分类看起来像儿童的涂鸦：走路、等车、在市场的换乘、那位有时会为她停、有时不停的司机，还有孩子们放学回家之前她补的那一个小时觉。
Priya sat in the depot until the light came up, writing. It was not a single omission. Every shift worker in the city had been excluded by a design choice she had made in an afternoon, and the omission had a shape: it removed precisely the people whose journeys were longest and whose wages were lowest, and it did so while producing a graph that seemed to prove the transport system was working. She thought about the final paragraph, the standard language of limitations, and understood that she had written a sentence intended to protect her rather than to warn anyone.
普里娅在车场一直坐到天亮，一直在写。这不是一处遗漏。这座城市里每一个轮班工作者，都被她某个下午做出的一个设计选择排除在外，而这次排除是有形状的：它删掉的恰恰是路程最长、工资最低的那群人，同时却生成了一张看起来能证明交通系统运转良好的图表。她想起最后那一段，那套写局限的标准措辞，明白自己写下的那句话，本意是保护她，而不是提醒任何人。
She did not throw the survey away. She kept it, and spent the summer standing at bus depots and hospital entrances between ten at night and six in the morning, with a clipboard and a bag of cheap pens. It was slow and cold, far messier than the orderly research she had planned at her desk. She got three hundred and forty-one responses. Her supervisor said the sample was too small to be representative, which was true, and Priya replied that it was large enough to be embarrassing, which was also true.
她没有把那份调查扔掉。她留着它，然后用一个夏天，晚上十点到早上六点，站在公交场站和医院门口，拿着一块写字板和一把廉价圆珠笔。这件事又慢又冷，远比她坐在书桌前规划的有序研究更杂乱。她拿到了三百四十一份回复。导师说样本太小，不足以代表整体，这是真的；普里娅回答说，它已经大到足够让人难堪，这也是真的。
The final report had two graphs instead of one. The second was ragged and full of outliers, and it showed that a quarter of the city's workers were making journeys that the first graph had simply not measured. The council read it, or read the summary, and extended the night bus by two hours. Priya did not claim credit; the campaign had been running for years. But she rewrote her limitations paragraph, and this time she put it first, and she learned to be suspicious of any dataset that had been cleaned until it was beautiful.
最终的报告有两张图，而不是一张。第二张参差不齐，满是离群值，它显示这座城市四分之一的劳动者正在走的路程，第一张图根本没有测量。市政会读了它，或者说读了摘要，把夜班公交延长了两个小时。普里娅没有把功劳归给自己；这件事已经有人推动了很多年。但她重写了那一段关于局限的文字，这一次把它放在开头，并且学会了对任何一份被清理到漂亮的数据库保持怀疑。