ChinAI #112: Part II, The Human Cost of 30-min Food Deliveries

Plus, must-reads on forecasting research community growth and the effects of foreign-language proficiency

Greetings from a world where we play infinite games…

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: Delivery Drivers, Trapped in the System

*We’ve finished out the rest of this 15,000-word piece introduced in last week’s issue. Big thanks to ChinAI readers Gabriele and San for helping out with parts of the translation!

DISCUSSED IN PART TWO OF THE FULL TRANSLATION (in the style of how Believer magazine intros their articles)

  • The unchecked power (and impatience) of Chinese consumers, backed up by Ipsos survey data

  • Food delivery as a customer-centric “social performance” in which drivers are performing emotional labor, which is more draining than manual labor

  • Combined, and Meituan “employ” nearly 6 million drivers. Just let that sit for a minute.

  • Drawing Peppa Pig memes, begging for five-star reviews, and keeping spare Coke

  • How Meituan and implement gamification. One driver, "I was a Black Gold Knight last month. If I want to maintain it, I still need 832 points. There is still a lot of work to do."

  • Meituan’s effort to develop a windproof, waterproof, noise-removing, and smart Bluetooth headset with intelligent voice interaction so that workers stop looking at their phones while driving.

  • A “more deadly surprise” than being forced to watch driving safety videos — Operation Smile: a spot-check to make sure drivers are wearing work clothes, a helmet, badge, and a mask (in a time of Covid). Some false positives and negatives with this computer vision application.

  • Platform companies as “hands-off husbands” who care little about housekeeping; in Meituan and’s case they contract out the “housekeeping” of ensuring driver and traffic safety to police officers and insurance companies.

  • Nick Seaver’s concept of “algorithms as culture” — “constituted not only by rational procedures, but by institutions, people, intersecting contexts, and the rough-and-ready sensemaking that obtains in ordinary cultural life."

  • Amidst all the hubbub about Meituan’s market value exceeding US$200 billion, some people brought up Wang Xing’s (Meituan’s CEO) fascination with James Carse’s book Finite and Infinite Games. The purpose of the former is to achieve victory, while the latter aims to keep the game going forever. The article concludes, “The delivery platform system is still running, and the game continues, but the drivers know almost nothing about their role in this ‘infinite game.’ They are still sprinting ahead, for the possibility of a better life.”

*I’ve posted the first few sections of the second half translation at the bottom of this issue. If you have the time, the full translation is well-worth a read, especially if your diet of China-related content mostly views “China” as an abstraction, rather than an entity composed of many actual human beings.

FULL TRANSLATION: Delivery Drivers, Trapped in the System

ChinAI Links (Four to Forward)

Must-read: A Novel Approach to Predicting Exceptional Growth in Research Communities

Klavans, Boyack, and Murdick develop a novel approach to forecast growth in highly specific research communities. Based on one of their models, they forecast the top research communities in AI applications, focusing on who was the research leader rin each community:

Table 10

Table 10 shows the dominance of Alphabet (Google’s parent company) and Tsinghua University. Chinese and American institutions make up 9/10 of the research leaders in these research communities which are forecasted to show high growth.

Must-read: Effect of Foreign Language Proficiency on Attitudes Toward a Former Aggressor State (Hu and Liu, 2020, in Journal of East Asian Studies)

Very relevant to ChinAI’s value-prop, the authors find a channel through which individuals proficient in a foreign language have an “alternative channel through which they are exposed to positive narratives put forth by other parties regarding the former aggressor state. And as a result, their attitudes towards the former aggressor state are more positive than those held by their linguistically limited counterparts.” For ChinAI purposes, it’s not just about positive narratives but a greater diversity of narratives.

The authors examine public attitudes towards the Japanese in Mainland China, Singapore, and Taiwan—three Chinese-ethnic majority political units that experienced Japanese aggression leading up to and during World War II. Using survey data, they demonstrate that individuals who are proficient in the English language are much more likely to hold positive attitudes towards the Japanese. These results are robust even when controlling for alternative factors. H/t to Kaiser Kuo for recommending.

Should-read: The True Story of Lee Kuan Yew’s Singapore

Haonan Li and Victor Yaw write: “The Western student of international politics knows to nod approvingly when Lee’s name is mentioned. Frustrated by the sludge of partisan politics in his own country, he sees in Lee’s legacy a kind of exotic escape. If asked, he remarks sagely: Singapore is proof of what enlightened authoritarianism can achieve.”

Should-watch: DoD Joint AI Center Explainer on AI and ML

Greg Allen, head of strategy and comms for the JAIC, breaks down the technical concepts of AI and Machine Learning for a non-technical audience. Check out some really great stuff on their blog as well.

Thank you for reading and engaging.

These are Jeff Ding's (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is a PhD candidate in International Relations at the University of Oxford and a researcher at the Center for the Governance of AI at Oxford’s Future of Humanity Institute.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Any suggestions or feedback? Let me know at or on Twitter at @jjding99

Peppa Pig and Coke

Because of a disagreement with a customer, Meituan delivery driver Xiaolin discovered a ‘secret’ hidden in the system – the delivery time shown to the customer was different from the one displayed on the driver’s interface.

At the time, he had just started delivering Meituan orders. One time, he received an order, but as soon as he got to the store, a customer pelted him directly with a question: “How come you haven’t delivered this yet? It’s been delayed for a while.” Xiaolin thought that the customer was being unreasonable, because according to his smartphone there still were nearly 10 minutes available for delivery. After delivering the meal, he and the customer clashed again over time, and they both took out their phones to compare – the customer’s “expected delivery time” was exactly 10 minutes shorter than the driver’s “demanded delivery time”.

After he discovered this ‘secret’, Xiaolin has been calling the Meituan customer service every month for nearly four years; every time there is a different serviceperson, but the response is always the same: “Explain to the customer how that is only the expected delivery time.”

This is not only Xiaolin’s individual experience; many drivers have mentioned this problem to People. In their view, this is how the system pleases and retains customers, and this is also one of the principal causes of conflict between customers and drivers.

In his book Consumer Behavioral Science: Perspectives on Chinese Consumers, scholar Lu Taihong points out that the convenience afforded by the digital age makes consumers become increasingly critical, as they pay more and more attention to service quality and product experience; as their loyalty towards brands and products weakens, they become more ready to change suppliers at any time, and so “customers have a larger guiding influence on the market than they had in the past.”

Facing this kind of influence, delivery platforms that focus on large numbers of users and orders have also used algorithms to construct a kind of power structure; in this system, the customer is at the absolute top, and has unchecked power.

But customers can make mistakes. “Sometimes customers just can’t really tell” – on this topic, Gansu driver Wang Bing has a lot on his mind: “A lot of people don’t know where they live… They clearly live at the number 804, but they write 801. They clearly are at the South Gate, but they write North Gate. And there are some customers who forget they ordered a meal: when you call them, no one answers, but on the next day they remember and they give me a call, asking about their food… there are also people who order without looking at the address at all, and when I receive the order I immediately notice that the address is not right, that it’s somewhere in another province…”. The point is that customers do not need to pay for their mistakes, and if their order is delivered too late, it is always the driver who is punished.

As a sociologist who has extensively researched the professional difficulties of delivery drivers, Sun Ping has also discussed this kind of “unchecked customer power” in one of her articles. While the driver delivers a meal, the customer can examine everything about the driver: their real name, phone number, punctuality rate, how many times they have been praised, meal pick-up time, delivery route, and how long it will take to deliver. The customer also has the right to cancel the order while the order is being processed.

“They can see everything, the entire process, but we don’t know who they are. And if there is any problem, we can’t cancel the order like they can,” a driver complained to Sun Ping. They also shared a personal experience about having an order cancelled:

“I had two orders on my hands, one was 1.5 kilometers away, and had 45 mnutes left; the other was 3 kilometers away, and had only 20 minutes left, so I delivered the more distant one first. The 1.5 km customer got angry because he saw me pass their place through the GPS tracking without delivering their order. He was fuming, he canceled the order and complained about me with the platform…”

In the survey run by People, there also are drivers sharing similar experiences: on one day, upon receiving their meal, the customer asked the driver “Aren’t you just delivering one order to me?”

As the speed of delivery increases, the rating system becomes completely skewed, and the system’s pampering treatment of customers makes them more and more impatient.

Jing Jing, who lives in Shanghai, admits that he has been "spoiled." He is usually busy with work, does not know how to cook, and almost entirely relies on deliveries to fill his stomach. He often ordered food at a “light-meal” restaurant not far away. According to his memory, in the past, it took about 45 minutes from placing an order to eating the first small tomato in a Caesar salad. To pass the time, he usually watched a 45-minute TV series while he waited. Recently, the waiting time has stabilized at 26 minutes, but not long ago, the meal delivery time exceeded 30 minutes, and he became unbearable, making 5 calls to check on the order.

In 2017, Ipsos, a French research institute, conducted a survey on the “impatience” of consumers in 12 provinces and cities in China. The results showed that the development of mobile technology has made consumers more and more impatient in all aspects. This phenomenon is more and more prominent in economically developed areas and young people. Among them, "consumers in Beijing are the most impatient."

Faced with increasingly impatient customers, the drivers have no choice but to try every means to comfort them.

Speaking of this, Wang Bing also had a lot to say — when the delivery times of orders are close, he will pick the expensive one to deliver first, because customers with high-priced orders are usually more likely to lose their temper. “No matter how you explain it, they don’t listen, and they’ll suddenly get angry, saying that they will return the order. This 100 RMB takeaway -- how do I have the money to pay for this every day?”

They also try to meet customer needs other than food delivery, such as buying cigarettes and water, or "bringing a razor to an Internet cafe." For a while, under the influence of Douyin, there were always customers who asked Wang Bing to draw a piggy Peppa when delivering the meal. Wang Bing was very angry about this but he still had to draw it. "I bought a piece of cowhide/kraft paper, drew a Peppa Pig, and also wrote the sentence, ‘are you stupid?’"

"Delivery is a kind of customer-centric social performance." Sun Ping wrote in the survey report. She called the behavior of drivers to please customers and strive for five-star praise as "emotional and sentimental labor." In her opinion, this part of labor is often overlooked, but its damage and consumption to the driver are far greater than manual labor.

In the conversation with "People", she mentioned oneo driver who left the deepest impression. "Two vehicles were stolen from him within three days, and his battery was stolen three times. In talking about this he started crying and said that the platform requires us to say ‘Have a nice meal.’ Nobody knows this but I came from the countryside. I did farming work before. I am really embarrassed to say this (“have a nice meal!”) Also to ask for five stars --  I am a man. How can I say this?”

In an interview with Jiemian News about the "SKP incident," Shen Yang, associate professor of the Department of Public Economics and Social Policy at Shanghai Jiaotong University, said that although delivery drivers may have a monthly salary of over 10,000, they are still mired in class inequality. Those who make more money at the expense of time and health must do more intense work-both physically and emotionally-to get more wages.

Wang Bing is still developing new tricks to soothe customers. In summer, many people will order an extra cup of Coke with meals, but this summer there was a lot of rain. He often crashes due to way too many orders. When this happens, the Coke is basically gone and can’t be saved, but go back to the merchant and get another one, you will not only have to pay for it yourself but the order will also inevitably be late. In order to prevent customers from getting angry, he always keeps a bottle of Coke in his takeaway box. If the customer’s Coke is spilled, he finds a place where nobody is around to fill the spare Coke into the original paper cup, and wipes around the cup to leave no traces. He thinks this method is great.

At the same time as these delivery drivers are going to these lengths, some anxious customers appeared on several legal consulting websites. Someone posted a message asking, "I hurried the delivery driver to bring my order faster, causing an accident. Do I have to bear legal liability?" Below the question, there was a lawyer’s reply: "No responsibility."


Not long ago, Meituan and successively announced their 2020 Q2 earnings. In this quarter, turned the corner and reached positive profits per order, while Meituan completed a net profit of 2.2 billion RMB, a year-on-year increase of 95.5%. Among this, the delivery business is the biggest contributor to Meituan’s profitability.

On August 24, 2020, Meituan’s share price also reached a new high, with a market value exceeding US$200 billion, making it the fifth-largest company in Hong Kong stocks by market capitalization.

In this half-year-long investigation, People came into contact with nearly 30 takeaway drivers, and the one phrase they frequently brought up was yi mao qian (ten cents/a dime).

 A Meituan driver in Hunan said, "If your punctuality rate is less than 98%, then you lose ten cents per order; if it’s less than 97%, then you lose twenty cents per order. Isn’t this just forcing the drivers to speed up? For us, the difference of a dime in each order is huge.”

An driver from Shanghai said, " The lowest unit price is 4.5 yuan. The more you run, the higher the unit price. Sometimes the extra dime feels very touching. It looks different -- 4.9 RMB or 5 RMB."

 In order to keep this "dime", drivers not only need to run faster, but also run more.

 This is also what the system hopes to see, because there is another secret hidden in the system — a "game" with levels.

 Whether it’s Meituan or the system has set a point level system for the driver -- the more orders you run, the higher the punctuality rate, the better the customer evaluation, the higher the driver’s points will be. Higher points lead to higher levels and more reward income -- the system will also package this evaluation system like a monster-fighting video game. Drivers of different levels have different titles. Taking Meituan’s system as an example, the titles from low to high are: ordinary, bronze, silver, gold, diamond, and king.

 A Meituan crowdsourced driver from a city in the southeast described the specific level setting: within one week, he completed 140 valid orders, with a punctual rate of 97%. He will become a "Silver driver" and receive an additional reward of 140 RMB per week. If you complete 200 valid orders and the punctuality rate reaches 97%, you will become a "Gold driver" with an additional bonus of 220 RMB per week. In, the order quantity is directly linked to the delivery fee. If the number of orders completed per month is less than 500, then it’s 5 RMB per order; 500 to 800 orders, 5.5 RMBper order; 800 to 1,000 orders, 6 RMB per order... and so on. In the game rules, the points will be cleared on a weekly or monthly basis.

In the research report "Orders and Labor: Research on Algorithms and Labor from the Economic Perspective of Chinese Food Delivery Platforms," ​​Sun Ping said that in addition to overtime punishment, the system also uses this gamified evaluation method to involve many drivers. An unstoppable cycle, "They want us to work day and night," a driver said to her, but they can’t get out of it. "I was a Black Gold Knight last month. If I want to maintain it, I still need 832 points. There is still a lot of work to do."

 "The higher the level, the greater the pressure that drivers face to maintain the level." In Sun Ping's view, this gamified packaging not only presents the possibility of addiction, but also cleverly combines the driver's self-worth realization with capital management, and the coat of gamification "provides a universal, internalized, and reasonable explanation for the exploitation of algorithms."

According to the "Job Employment Report for the First Half of 2020" published by Meituan, the total number of drivers in Meituan has reached 2.952 million. The fengniao official website shows 3 million drivers. Facing the systematic survival of nearly 6 million drivers, Zheng Guanghuai, a sociologist at Central China Normal University, proposed the concept of "downloaded labor".

In the investigation report "Wuhan City Delivery Workers Group Survey: Platform Workers and ‘Downloaded Labor’", Zheng Guanghuai's team gave an in-depth explanation of this concept—Drivers "download" the app to start work. On the surface, this app is just a production tool to assist them in their work, but in fact, what the drivers "download" is a set of sophisticated labor control modes. Under this mode, "The original subjectivity of workers has been comprehensively shaped and even replaced." They seem to work in a more free way, but at the same time they "suffer deeper control."

"The platform creates "platform workers" through downloaded labor." Zheng Guanghuai's team wrote, and the characteristics of this labor model are: strong attractive force, weak contract force, high supervision, and low resistance.

The medium that assists the system in completing the "downloaded work" is the drivers' own mobile phones-as the most important work tool, in public reports, the takeaway platform has been working hard to help the drivers get rid of their mobile phones.

"We are afraid that the driver will have trouble taking orders." In an interview with 36Kr in April 2018, He Renqing, the head of Meituan's delivery algorithm team, specifically mentioned, "For Meituan, the most difficult problem is how to prevent the driver from looking at their mobile phone while riding."

For this reason, Meituan spent 7 months developing a Bluetooth headset with a built-in intelligent voice interaction system. According to He Renqing, this headset is windproof, waterproof, noise-removing, and smart. Drivers can complete all operations by speaking as long as they wear it, ensuring that they can get ride of the mobile phone during the food delivery process.

In reality, none of the Meituan drivers who have communicated with People have received or used this smart Bluetooth headset, and none of the drivers can really get rid of their mobile phones.

Although she has only experienced the life of a delivery driver for a few days, Cao Dao still has lingering fears of the fear of being controlled by her mobile phone. "As you’re navigating, the system will ceaselessly remind you that Meituan crowdsourcing has another new order, please check the order in time and that will get mixed together with the navigation sounds. And then once you’re almost past time, customers will call you to ask where you are. You may have to take another order while navigating, and then answer the phone to explain to the customer why the order is not delivered on time..." Cao Dao said that this feeling made her feel that every minute is important and that she is being chased every day, "Can only go fast, go faster."

Electric Vehicles

"We can never lose time on the road. The time on the road is the fastest." An driver told People, and another Meituan driver said that it’s only when on the road that an order is truly in his own hands., "Unless there is traffic police riding behind me and saying don’t speed, don’t speed. Otherwise when there are too many orders, all the drivers want to let it fly." After that, he added another sentence, “even if you fly, you can’t make it.”

At this time, the only thing that can help them is the electric bike that they are riding.

Before taking the job, drivers need to solve the problem of electric vehicles by themselves. Usually, the distribution stations have long-term cooperative third-party companies that provide drivers with electric vehicle rentals. In order to save costs, most drivers will choose a car with a rent of several hundred RMB, and the conditions of these cars are mostly difficult to say — some do not have a rearview mirror, and some pedals and front of the car are wrapped seven or eight times by rubber strips. A driver said that after running deliveries, he became an "electric car repair master."

If you don't want to rent, some sites will also guide drivers to buy a vehicle in installments.

A Meituan driver in Chengdu, at the request of the distribution site, bought an electric vehicle of an unknown brand at a price of 1,000 RMB higher than the market price. Another driver said that he had spent thousands on an electric vehicle through the distribution site. Just two days after running, the vehicle’s battery broke.

Meituan driver Wang Fugui felt lucky compared with those peers who mistakenly spent too much money. The only thing that happened to him was that he flew (along with the battery) out of the vehicle and got his head stuck in a guardrail in the middle of the road -- on his first day of becoming a driver. He had rented the vehicle through the station, and the monthly rent was 200 RMB. "Basically, it's a bunch of pieces put together." With no lights, the brake pads are worn out. Sometimes when you step on the brake, it will move forward, but when you step on the accelerator, it will reverse.

But this is not a problem. On the second day after the crash, he spent 10 RMB to install a foot brake by himself. When he ran the night shift, he would put a small flashlight in his mouth instead of the lights, or glue the flashlight to the front of the car. After all, this car also has its advantages. "It's extremely fast and can run up to 65 kilometers per hour," said Wang Fugui.

According to data released by the Ministry of Public Security in 2018, between 2013 and 2017, there were 56,200 road traffic accidents involving electric bikes causing casualties, resulting in 8,431 deaths and direct property losses of 111 million RMB. In order to further regulate the use of electric vehicles, in April 2019, the country officially implemented the new national standard for electric vehicles-according to regulations, the speed of electric vehicles must not exceed 25 kilometers per hour, and an electric vehicle that meets the new national standard must be sold for at least 1,000 RMB.

However, among nearly 30 delivery drivers contacted by People in this survey, no matter whether they were with Meituan or, none of their electric vehicles meet the new national standard. These electric vehicles can generally run up to 40 kilometers per hour, far exceeding the speed limit. In groups and Tiebas of drivers, there are still many people discussing how to remove the speed limit for newly purchased electric vehicles through modification.

After working as a driver for more than a year, that car started to break down more and more frequently. Wang Fugui sometimes has to take a taxi to deliver meals. Fortunately, he’s located in a small county town in the northwest, so instead of having delayed deliveries on orders due to vehicle breakdowns, it’s more affordable to take a taxi. You can easily send more than a dozen orders for 50 RMB.

Later, in order to run faster, he gritted his teeth and bought a new vehicle by himself. As for the broken-down vehicle from before, he doesn’t know how many parts were dismantled and installed on however many electric vehicles waiting to be rented.

Whether riding an old or new vehicle, Wang Fugui’s performance has always been ranked in the top five and top three in the region, but he resigned not long after joining because he couldn’t stand the platform’s new requirements. “ Meituan, in order to expand, had us go on the streets to solicit customers. Every day we had to sign up two people who had never logged in to the Meituan app. At first, I had to endure it for a few days. Later, I couldn’t stand it anymore, so I ran away.”

Operation Smile

(continued in) FULL TRANSLATION: Delivery Drivers, Trapped in the System