Artificial intelligence (AI) workers in Kenya have launched the Data Labelers Association (DLA) to fight for fair pay, mental health support and better overall working conditions.
Employed to train and maintain the AI systems of major technology companies, the data labellers and annotators say they formed the DLA to challenge the “systemic injustices” they face in the workplace, with 339 members joining the organisation in its first week.
While the popular perception of AI revolves around the idea of an autodidactic machine that can act and learn with complete autonomy, the reality is that the technology requires a significant amount of human labour to complete even the most basic functions.
Otherwise known as ghost, micro or click work, this labour is used to train and assure AI algorithms by disseminating the discrete tasks that make up the AI development pipeline to a globally distributed pool of workers.
Despite Kenya becoming a major hub for AI-related labour, the DLA said data workers are massively underpaid – often earning just cents for tasks that take a number of hours to complete – and yet still face frequent pay disputes over withheld wages that are never resolved.
Screenshots shared by the DLA show that in the worst instances, data workers have been paid nothing for around 20 hours of work.
“The workers power all these technological advancements, but they’re paid peanuts and not even recognised,” said DLA president Joan Kinyua, adding that while the labellers work on everything from self-driving cars to robot vacuum cleaners, many of the products their largely hidden labour powers are not even available in Kenya.
Given the job specs of a lot of data work – which requires workers to be graduates and have high-speed internet connections and quality machines – DLA vice-president Ephantus Kanyugi said workers are being forced to making big upfront investments in their education and equipment, only to be paid a few cents per task.
The workers power all these technological advancements, but they’re paid peanuts and not even recognised Joan Kinyua, Data Labelers Association
He added employers in the sector are disincentivised to pay more – or even follow through on paying people for the work they’ve done – because the large surplus labour pool means when people inevitably get frustrated and leave, “they already have someone in the pipeline” to replace them.
DLA secretary Michael Geoffrey Abuyabo Asia added that weak labour laws in Kenya are being deliberately exploited by tech companies looking to cheaply outsource their data annotation work.
“A contract is supposed to be agreed within the confines of the law, but they know the law is not there, so it becomes a loophole they’re utilising,” he said.
DLA members added the lack of formal employment contracts with clear and consistent terms throughout the sector also leads to a lack of longer-term job security, as work can change unpredictably when, for example, jobs are randomly taken offline, and allows sudden account deactivations or dismissals to take place without warning or recourse.
On the contracts, Kinyua added there is no consistency, as while some are indecipherable due to the legal jargon, others will cover just a few days, and in some instances there is no contract at all.
The lack of security is heightened by the fact the workers do not have access to healthcare, pensions or labour unions. The workers said all of this combines to create highly variable workloads and income, which then makes it difficult for the workers to plan for the future.
On top of the day-to-day precarity faced by data workers, the DLA said many also have to deal with “content moderation trauma” as a result of having to consistently interact with disturbing and graphic images, as well as retribution from companies when they do raise issues about their work conditions.
“Any time we raise our voice, especially the ‘taskers’ who are on the lowest level, we are dismissed automatically,” said one DLA member, who added contracts were no help in these instances because they would typically only specify a start and end date without any other information, and that she herself was dismissed when speaking up on behalf of others.
To help alleviate these issues, the DLA will focus on getting workers mental health support, giving them legal assistance to deal with pay or employment disputes, providing them with professional development opportunities, and running advocacy campaigns to highlight common problems faced by data labellers.
The DLA will also seek to put in place collective bargaining agreements with the data companies that sit between workers and the large tech firms whose models and algorithms they are ultimately training.
As part of its efforts to push for better working conditions in the sector, the association is already working with the African Content Moderators Union and Turkopticon – which largely works with data labellers on Amazon’s Mechanical Turk platform – as well as the Distributed AI Research (DAIR) Institute.
The organisation added it is already in touch with Kenyan politicians and is communicating with the Ministry of ICT to help lawmakers better understand the nature of their work and how conditions for platform workers can be improved.