Melting Labour: Asking about My Work's Future

Sleepless "Ghostworkers" Behind Portals and Social Media

2020.01.23 15:17
Choi Mi rang, Sim Yoon-ji

Illustration by Kim Sang-min

Illustration by Kim Sang-min

Web portals, online shopping malls, and various platforms∼. The online world is “open” 24/7. Someone has to keep an eye on them to see if they are working properly. But when you tap your smartphone, it’s not easy to think about the “workers” behind the screen.

Mary L. Gray and Siddharth Suri of Microsoft Research studied the labour behind artificial intelligence (AI) and Internet technology and they dubbed the people developing and complementing AI “ghostworkers.” They argue that bringing their work--which is shattering existing employment relations and spreading--to the surface and protecting these workers is our most urgent task.

In connection with this issue, the Kyunghyang Shinmun focused on the low-wage workers in South Korea working at web portals. Compared with the employees of Google and Facebook scattered throughout the world, these workers have a relatively clear employment relationship, but their work can be referred to as “ghostwork” because their work of complementing AI and machines is not quite visible. The people who responded to the interview said that such a term used to describe their work was “appropriate.” However, they did not want to be named, so they are mentioned under assumed names.

■ The “Gatekeepers” Behind the Screen

Source of photo: O-dan.net, a website providing free images

Source of photo: O-dan.net, a website providing free images

Yi Hye-eun has worked at an affiliate of a South Korean web portal operator for years. When she enters the office, she turns the computer monitor on and stares at images and videos for hours. “I have to check and see if they are normal behavior or sexual abuse, slang that can be used or not, those kind of things. At first, I had a hard time because I wasn’t used to it, but if you keep looking, you kind of get numb.”

AI first filters posts, images, and videos that users post on blogs and online communities. But there is always an area that AI cannot determine. Workers like Yi have to decide whether or not such content can be exposed by working in shifts 24/7. AI then learns from these man-made decisions, improving its accuracy. “When I think about how kids could access these contents, I really want to get rid of them quickly. Once I fiercely deleted the videos thinking, ‘Okay, it’s either you or me.’”

Yi and her colleagues use the expression, “turning off,” to describe how they try to forget work once they have completed their tasks. The problem is that some scenes keep popping up in their minds even if they turn their monitors off. Yi, who has a daughter, said her hands still trembled and tears still poured out when she recalled a child abuse video she saw when she was on duty years ago. “No matter what I do, I just can’t ‘turn off.’ The world just seems so terrifying.”

“Who in the world would post this thing? Are there people like that around me, too?” Such thoughts once made it difficult for her to meet people. She had trouble sharing her problems with her friends. “’Lucky you.’ ‘You get to watch things we can’t even access.’ I thought I would be hurt if someone just made a joke like that.”

In her department there are many twenty-something “newbies.” Although they were aware of what they would be doing when they joined the company, Yi remembers several employees who “just didn’t know what to do because it (toxic content) was too intense.” The salary for the entry-level staff is slightly higher than the minimum wage. “It’s a job that has to be done, but it’s not a job where one’s performance really stands out. It isn’t a highly valued task.” Workers who crumbled under the stress left in search of other jobs.

The amount of images and videos they had to watch was drastically reduced in the mid 2010s, when the headquarters outsourced some of the task to a Chinese affiliate. It was possible because images are less subject to language barriers. The development of AI also automatically filtered out more and more toxic contents.

Jo Hyeon-ae examines contents at an affiliate of another web portal. She once got cystitis keeping her eyes on her monitor all day without even a bathroom break. When she worked in day/night shifts, she suffered from sleep disorders and tinnitus. When she applied sanctions on users who violated the operation policy, she would receive phone calls and e-mails full of profanity.

Digital companies all over the world including Google and Facebook also hire people to filter out inappropriate content. The standards for blocking such disturbing content may differ from country to country, but the process is the same. AI first filters out the contents and then people filter out contents where any risks have been detected. Scholars call such labour, “data gatekeeping,” “hidden digital labour,” or “digital cleaning labour.”

Naver and Kakao have affiliates and partner firms in and outside the country handle these tasks. Google and Facebook distribute the job to workers in all corners of the world through firms that supply IT labour. Workers in the Philippines and India are assumed to be moderating most of the English contents.

Last year, the Washington Post covered the status of Filipino content moderators who suffered from suicidal impulses. Last December, one content moderator in Ireland filed a lawsuit against Facebook and a contractor claiming that he suffered from post-traumatic stress disorder because of the cruel images.

In South Korea, there are workers who moderate contents for Google. Gim Min-a is one of them. Gim’s job is to select the websites that should be exposed at the top when a certain word is searched. This is to improve the accuracy of the search. She signed a contract that stated a condition that she would not have to review hazardous contents. Because of a confidentiality agreement with the company, she could not share details, such as her salary. She “has no idea” who else is doing what she does. She receives an e-mail from the company when the work has not been properly handled. “I read the e-mails carefully every time, because the company could ‘annul the contract’ (she used this expression instead of the words ‘lay off/fire’) via e-mail.”

■ “Manual Labor” in a Digital Company

The inside of a data center in Switzerland. Getty Images

The inside of a data center in Switzerland. Getty Images

Chae Myeong-min works at a data center of a South Korean web portal. His main task is to add or remove servers and repair flaws. He also carries the heavy servers in and out. He joined the company after receiving IT training funded by the state, but he describes his task as “literally physical labor.” When a problem occurs, he has to respond immediately, so he usually eats in front of the monitor.

Chae was employed shortly after he had completed his mandatory military service, and at the time he agreed with the words of some politicians who argued that the young people had trouble finding jobs because they had their goals set too high. He thought if he could get a job by lowering his standards, he would be able to move on to a better job after struggling for a few years and gaining experience. Once he entered the company, he said he was surprised at the high turnover rate, but now a year later, he can see why. “They don’t pay you a lot, and there’s no future.” Even if he worked the night shift from 10 p.m. to 10 a.m. his annual salary is just 200,000 won more than the minimum wage (20,940,000 won based on 2019 wages). “The people who were here before me advise, ‘There is no future in a place like this, where they don’t know the value of people. You’ll be better off in China.’”

Han Yu-ri works at the control tower of a web portal. When an alert indicates a problem, she delivers the information to the person in charge according to the manual. She has worked 12-hour night shifts alone for up to six times a month. Her colleagues are mostly in their early twenties and the wage is no different from the minimum wage. “When I tell people I work in an affiliate of a web portal, no one thinks that I would be doing this kind of work. Only people who are young, have no experience, and who have never worked before seem to come here. It’s a job that has no ‘next.’”

Gim Dong-geun, who works at the data center of another web portal also described his job as similar to “manual labor.” “The number of employees are so few, they probably can’t reduce it any further. Sometimes five or six of us tend to tens of thousands of servers.”

Their workplace is an administration firm that handles tasks outsourced by headquarters (Naver, Kakao). Last year, the Korea Labour & Society Institute analyzed the wages of Naver and Kakao employees who had joined the National Chemical, Textile and Food Industry Union, a branch of the Korean Confederation of Trade Unions, and the results showed a distinct wage gap between headquarters and the administration firm. Among union members at Naver headquarters (392 respondents), 93% received an annual salary exceeding 50 million won, but when it came to members in the administration firm (132), 82% said their annual salary was less than 50 million. The situation was similar at Kakao―among union members at Kakao (80), 84% earned more than 50 million won a year, while 100% of the members in the administration firm (29 respondents) received less than 40 million won.

If AI develops further, will such arduous low-wage labour disappear? Scholars predict the continuous need for human labour even for automation. Gray and Suri referred to this as the “paradox of automation’s last mile.” There is no “end” to automation, because AI developers continue to modify their goals. Thus we will need someone to continue to complement AI. Workers will always have to newly learn this or that task.

The workers on the field seem to know this already. Jo Hyeon-ae said, “We might disappear in the distant future. But machines and programs are man-made, so they will need people to continue to take care of them in invisible places. Unfortunately, companies will keep trying to save labour costs∼ so I don’t think the pay will get any better.”

추천기사

바로가기 링크 설명

화제의 추천 정보

    오늘의 인기 정보

      추천 이슈

      이 시각 포토 정보

      내 뉴스플리에 저장