美国国家公共电台 NPR--When your boss is an algorithm

When your boss is an algorithm

Transcript¹

Two brothers who drive for Uber recently conducted an experiment. They opened their Uber apps while sitting in the same room, and tested which brother could earn more money to do the same work.

In a video published on The Rideshare Guy YouTube channel, the brothers recorded themselves looking for rides on the app. They found that Uber showed them nearly identical jobs, but offered to pay one of them a little better. The siblings² could only guess why. Had Uber's algorithm somehow calculated their worth differently?

University of California College of the Law professor Veena Dubal says that's exactly what's going on. In a recent paper, she says rideshare apps promote "algorithmic wage discrimination" by personalizing wages for each driver based on data they gather from them. The algorithms are proprietary³, so workers have no way of knowing how their data is being used, Dubal says.

"The app is their boss," Dubal told Morning Edition's A Martinez. "But unlike a human boss who you can negotiate with or withhold⁴ information from, the algorithms know so much about these workers."

Uber says workers who drive electric vehicles get a $1 bonus per ride, but the company does not use drivers' personal data to set their pay rates. "Uber does not personalize fares to individual drivers, and a driver's race, ethnicity, acceptance rate, total earnings⁵, or prior trip history are not considered when calculating fares," a spokesperson writes in a statement. A representative for Lyft calls Dubal's paper "biased⁶", saying it relies on cherry-picked data and debunked⁷ anecdotal information.

Personalized digitized pay is already the new normal in some workplaces, according to Dubal, and it's begun to attract attention from regulators.

This conversation has been lightly edited for clarity and length.

Rideshare drivers say the app is their boss. And unlike a human boss who you can negotiate with or withhold information from, the algorithms know so much about these workers. They know how much a worker is willing to accept for a particular ride. They know how much workers try to earn on any given day. They can really personalize how much that worker makes in order to influence their behavior in particular ways.

I look specifically at ride-hailing firms to discuss the phenomenon of digitalized variable pay, but it's happening across the on-demand economy and even maybe beyond it. Basically, these firms, because they treat their workers as independent contractors⁸, cannot tell them what to do and where to go. Instead, they use these pay mechanisms⁹ to influence their behavior. They learn everything that they can about particular workers and use that knowledge to shape how workers get paid.

On how this differs from other forms of unequal pay

In a more familiar employment setting where workers make different amounts of money [to do similar jobs], we still have a legal norm: equal pay for equal work. In those contexts, there is often some logic¹⁰ to why people are earning more, whether it's seniority or experience or skill, and that is often transparent¹¹. There are also laws that ensure that companies check their own practices to make sure people are earning roughly the same amounts. What's complicated about algorithmic pay is that there is no logic. Instead, it might be that the person who works for a really long time, works really hard, and has the most experience is earning the least, and we just can't know. The logic is all hidden behind black box algorithms.

On how algorithms can reproduce discrimination

Some of these firms have, in their own research, found that these practices can lead to women earning less than men. They ascribe these differences to the algorithms. But if these algorithms are recreating traditional wage differences that are illegal under employment laws, then something is deeply wrong.

On whether workers can access how algorithms calculate their pay

In Europe, under GDPR [General Data Protection Regulation], some workers have, after litigation, won the right to have some access to what data companies are extracting from their work to determine particular prices. None of that has been revealed yet. In the U.S. there are similar privacy laws, but none of this has been litigated yet, and attempts to get at it from regulators has largely been met with resistance. Companies maintain that this is, oddly, about privacy, that they don't want to unveil their practices because it might lead to information about workers being leaked. They also maintain that these systems they've developed are their intellectual property.

On whether variable digitized pay is illegal

It might be illegal under antitrust laws. There is the potential to say that some of this is price fixing, if all of these workers are independent contractors. That is being litigated in California courts right now. But absent a finding on an antitrust violation¹², this isn't necessarily illegal. It's a brave new world.

On whether regulators could intervene

The Federal Trade Commission is looking into this. They're very interested in whether or not this violates antitrust laws. And I think that any number of lawmakers who are generally interested in economic equality – from Sen. Elizabeth Warren to Sen. Bernie Sanders – are likely very interested in what amounts to a dystopic system of work.

1 transcript   

n.抄本，誊本，副本，肄业证书

参考例句：

• A transcript of the tapes was presented as evidence in court.一份录音带的文字本作为证据被呈交法庭。
• They wouldn't let me have a transcript of the interview.他们拒绝给我一份采访的文字整理稿。

2 siblings   

n.兄弟，姐妹( sibling的名词复数 )

参考例句：

• A triplet sleeps amongst its two siblings. 一个三胞胎睡在其两个同胞之间。 来自《简明英汉词典》
• She has no way of tracking the donor or her half-siblings down. 她没办法找到那个捐精者或她的兄弟姐妹。 来自时文部分

3 proprietary   

n.所有权，所有的；独占的；业主

参考例句：

• We had to take action to protect the proprietary technology.我们必须采取措施保护专利技术。
• Proprietary right is the foundation of jus rerem.所有权是物权法之根基。

4 withhold   

v.拒绝，不给；使停止，阻挡

参考例句：

• It was unscrupulous of their lawyer to withhold evidence.他们的律师隐瞒证据是不道德的。
• I couldn't withhold giving some loose to my indignation.我忍不住要发泄一点我的愤怒。

5 earnings   

n.工资收人；利润，利益，所得

参考例句：

• That old man lives on the earnings of his daughter.那个老人靠他女儿的收入维持生活。
• Last year there was a 20% decrease in his earnings.去年他的收入减少了20%。

6 biased   

a.有偏见的

参考例句：

• a school biased towards music and art 一所偏重音乐和艺术的学校
• The Methods: They employed were heavily biased in the gentry's favour. 他们采用的方法严重偏袒中上阶级。

7 debunked   

v.揭穿真相，暴露( debunk的过去式和过去分词 )

参考例句：

• His theories have been debunked by recent research. 最近的研究揭穿了他的理论的真相。
• Some advertising slogans should be debunked. 某些夸大的广告用语应予揭露。 来自《简明英汉词典》

8 contractors   

n.（建筑、监造中的）承包人( contractor的名词复数 )

参考例句：

• We got estimates from three different contractors before accepting the lowest. 我们得到3个承包商的报价后，接受了最低的报价。 来自《简明英汉词典》
• Contractors winning construction jobs had to kick back 2 per cent of the contract price to the mafia. 赢得建筑工作的承包商得抽出合同价格的百分之二的回扣给黑手党。 来自《简明英汉词典》

9 mechanisms   

n.机械( mechanism的名词复数 )；机械装置；[生物学] 机制；机械作用

参考例句：

• The research will provide direct insight into molecular mechanisms. 这项研究将使人能够直接地了解分子的机理。 来自《简明英汉词典》
• He explained how the two mechanisms worked. 他解释这两台机械装置是如何工作的。 来自《简明英汉词典》

10 logic   

n.逻辑(学)；逻辑性

参考例句：

• What sort of logic is that?这是什么逻辑？
• I don't follow the logic of your argument.我不明白你的论点逻辑性何在。

11 transparent   

adj.明显的，无疑的；透明的

参考例句：

• The water is so transparent that we can see the fishes swimming.水清澈透明，可以看到鱼儿游来游去。
• The window glass is transparent.窗玻璃是透明的。

12 violation   

n.违反(行为)，违背(行为)，侵犯

参考例句：

• He roared that was a violation of the rules.他大声说，那是违反规则的。
• He was fined 200 dollars for violation of traffic regulation.他因违反交通规则被罚款200美元。