From Energy Crises to Earnings Cuts: the Syrian Data Worker Experience

3 min read

In today’s world, we are surrounded by technological advancements and the incredible capabilities of AI and its many applications. From self-driving cars that smoothly navigate city streets to advanced medical imaging systems that can detect diseases with remarkable accuracy, AI has become an integral part of our daily lives. These marvels of AI are transformative agents that shape our future. Yet behind these marvellous advancements, there is an army of data workers who remain unseen in the darkness. The data workers who power these AI systems often grow tired in the shadows, toiling over complex tasks to bring these technologies to life. Working long hours, striving to make a living with little to no alternative, data workers’ contributions usually stay unseen and underappreciated. As a data worker and co-researcher in the Data Workers’ Inquiry (DWI) project, I always advocate for the well-being, decent working environment and better living conditions for data workers like me who play a crucial role in powering AI applications.

Data workers in Syria in arduous conditions

In the Middle East, war-torn Syria has become the ideal spot to exploit people. With a lack of legal protection, many data workers there find themselves vulnerable to exploitation. In Syria, data workers face several hardships. Living almost without electricity and with no stable internet connection makes it hard to continue our work. We have to continually search for places that can provide us with electricity and access to the internet. Fortunately, there are some places in Damascus that charge $0.50 to $1.50 per hour, which is roughly equivalent to our earnings. What makes all this worse is that the jobs come with strict deadlines and high accuracy requirements, often leading to our work being rejected. It has even been known for people to lose their jobs.

Imagine working long hours, doing repetitive, nerve-wracking and complex tasks that strain your eyes, all while paying out-of-pocket for electricity and internet, only to find out that the work you have been striving to complete was rejected without you knowing why or being able to ask for an explanation. Many data workers complain about their work being rejected. When a batch of data is rejected, the whole team have their earnings docked, which feels like a shared penalty. One example of this is the story of Hind, who clocked up eight hours of work a day for an entire month on a self-driving car project, even working at weekends, adhering to strict rules and daily targets. She lived in fear of being fired if she did not meet those targets. At the end of the project, due to a change in compensation, she received less money than she had spent on electricity and internet. This decision was made without any explanation or the opportunity to appeal or receive an adequate explanation. Typically, workers calculate their expenses versus earnings before accepting a new project to ensure it is profitable. Therefore, unexpected changes in compensation are their biggest fear, as it directly affects their livelihood.

Unpaid work is a pervasive issue affecting data workers in Syria. Take Moiad, for instance, a dedicated data annotator who was promised a flexible five-hour workday. However, he soon found himself working additional hours without compensation, struggling to meet relentless project demands. Despite the time and effort he spent on the work, he was only paid for the initial five hours, a betrayal that left him feeling deeply exploited and demoralised. The realisation that his extra labour was dismissed without acknowledgement or fair pay amplified his sense of injustice.

Spending excessive time on inadequate training programmes without compensation amounts to wage theft. Additionally, “testing tasks”, which are integral to the project dataset, are another form of exploitation. These tasks determine whether a data worker will be accepted for a specific project and can take hours, if not days, to complete. Despite the significant time investment, these tasks are unpaid, further highlighting the injustice faced by data workers.

Inadequate training and its consequences

Poor training is a primary cause of rejected work among Syrian data workers. Instead of well-structured guidance, workers must rely on lengthy, repetitive videos or complex documentation in foreign languages – materials that often create more confusion than clarity. The impact is severe. Tasks completed with unclear instructions face rejection without compensation or feedback. Workers navigate assignments through guesswork, fearing penalties or dismissal. When mistakes occur, entire teams may face wage deductions, amplifying the unfairness. This reflects a deeper issue: companies transfer productivity costs to workers, treating their time as disposable. The solution requires clear, accessible training and constructive feedback systems that enable worker success rather than failure.

What steps can we take to ensure fair treatment and compensation for data workers?

The well-being of data workers spans crucial physical, emotional and financial dimensions and is often overlooked. These workers frequently endure cycles of excessive work and low wages, causing significant stress and exhaustion. Ensuring fair treatment and proper compensation for data workers is necessary to acknowledge the real worth of their labour. It is essential to implement comprehensive training programmes that are clear, concise and in accessible languages, ensuring that data workers have the necessary skills and knowledge to perform their tasks efficiently. Supervisors and team leaders should provide regular, constructive feedback to guide improvement, acknowledge hard work and help workers understand their mistakes and improve their performance. These measures will not only enhance the well-being of data workers, but also improve overall productivity and job satisfaction.

Most data workers operate in countries in the global majority, some of which are subject to sanctions. Data workers in sanctioned regions often have limited access to the necessary resources and face significant challenges when transferring compensation to Syria, which results in financial loss. Even with contracts, it is difficult to acquire rights because living in Syria renders these contracts ineffective and barely worth the paper they are written on[FW1] . Additionally, no international organisations can protect these workers as they are unable to operate with Syrian residents. These restrictions add an extra layer of hardship, making it difficult to maintain a stable and productive working environment. 

It is imperative for organisations to distinguish political issues and sanctions from data workers’ rights. This separation is crucial to ensure that data workers are not unfairly impacted by geopolitical circumstances. International organisations need to take swift action to monitor and enhance the conditions of data workers to better their working and living environments. Providing such oversight guarantees that data workers are treated fairly and compensated justly for their essential work in AI. Addressing these concerns is vital for both ethical considerations and the sustained progress and integrity of AI.

What data workers want to say to AI buyers (requesters)

It is crucial that AI buyers recognise and acknowledge that behind every dataset and algorithm, there are data workers labouring in challenging conditions. They work long hours, often without basic necessities or a safe working environment. AI requesters must understand the processes behind obtaining and handling their data, which involve real people who are frequently living in dire circumstances, striving to make a living and performing repetitive tasks for hours on end. It is not ethical to turn a blind eye to the exploitation of these individuals, who pour their heart and soul into their work. Even if AI buyers pay substantial amounts to outsourcing firms, this does not necessarily translate to fair compensation for the data workers themselves. AI requesters have a critical role to play in upholding the ethics of AI by ensuring and monitoring how their work is done, how data workers are compensated and the working conditions they endure. By taking these steps, they can force companies to adhere to fair practices and prevent the exploitation of data workers.

 [FW1]This is another useful expression which I think works well in this context. An alternative option here might be “mere words on a page”.