A webinar exploring ‘what does it means to have an algorithm as your manager?’ has heard calls for a new EU Directive specifically aimed at employment protections and rights from the increasingly pervasive role of Artificial Intelligence and data algorithms at work.
The Gig Economy Project, led by Ben Wray, was initiated by BRAVE NEW EUROPE enabling us to provide analysis, updates, ideas, and reports from all across Europe on the Gig Economy.
This series of articles concerning the Gig Economy in the EU is made possible thanks to the generous support of the Andrew Wainwright Reform Trust
Webinar hears call for new EU Directive to protect workers from AI and algorithmic power
A webinar exploring ‘what does it means to have an algorithm as your manager?’ has heard calls for a new EU Directive specifically aimed at employment protections and rights from the increasingly pervasive role of Artificial Intelligence and data algorithms at work.
Aida Ponce Del Castillo, senior researcher at the European Trade Union Institute, argued at the webinar hosted by Dutch MEP Kim van Sparrentak on Friday [9 July] that the EU’s Artificial Intelligence Act, unveiled in April, was “completely absent” when it comes to labour rights.
Castillo described it as an “AI regulatory winter” in the EU, because the Act “had high expectations, but was disappointing with many critical gaps”.
She added: “It was designed to promote AI as an industrial sector, but not really to protect rights [in general] or labour rights, which are completely absent. It’s focused on a limited number of high risk AI systems and the approach is based on self-assessment by the provider, with no prior validation by any external authority.”
Castillo said that a Directive focused on protecting employees from AI would include:
-
The responsibility of the employer in managing AI systems and ensuring AI does not intrude on workers’ rights, such as privacy and health & safety.
-
To ensure workers data was safeguarded and that algorithms were explainable and transparent to workers.
-
Workers should also have clear rights to contest automated decisions.
-
The ‘human in command’ principle in use of algorithms should be enshrined in law, and that should include not just company bosses but also worker representatives.
-
The prohibition of algorithmic worker surveillance, where an algorithm can build a profile of a worker (based on evaluating biologically, emotionally and behaviourally) and make decisions affecting the worker on that basis.
Jeremias Adams-Prassl, a professor at the University of Oxford who specialises in EU and labour law with a focus on labour law and the digital economy, agreed that the EU’s AI Act was “disappointing” when it came to workers’ rights.
He added: “GDPR has been important in providing some protection, but the EU has been operating in silos when it comes to algorithms. The Artificial intelligence Act could have tackled that when it comes to labour law but it hasn’t.”
Adams-Prassl, author of ‘Humans as a service: The problems and perils of Work in the gig economy’ (2018), also said that “enforcement” of rights around the algorithmic power of corporations was also a problem.
“Even if we have these rights, how do we enforce them and make them real is very difficult. There is a duty on member states to ensure effective enforcement, but translating that into practise is difficult.”
He explained that many of the algorithms currently deployed by corporations to assess things like productivity and innovation were “bizarre and just out-right silly” and had the danger of systematising the biases of those in management who have signed-off on the algorithm.
“For example, talent retention and pay setting, algorithms are designed to work out who is most likely to leave and therefore to give big pay rises to them and not others,” Adams-Prassl said. “The result of that will often be highly gendered…we end up not just replicating biases but actually making them worse, systematising them across the entire workforce [which] is something deeply problematic.
The question of employer responsibility was also key, Adams-Prassl argued, because there had been a “control-accountability paradox” since AI emerged as a key part of corporate operations: traditionally, the more control exercised by a business the more accountability they had, but now with algorithms the control exercised by companies has been massively intensified, but the accountability has got “lost in the cloud”.
Castillo agreed: “There’s always a human behind an algorithm. Blaming an algorithm for mistakes is to dilute or deny the responsibility of management. We should not shift the responsibility from human management to an algorithm.”
James Farrar, director of the Worker Info Exchange who was part of the team which led the successful UK Supreme Court case against ride-hail giant Uber in February, made the case for workers having full access to the data which is kept on them by platform companies, citing his own case when a driver at Uber.
“When we went to court there was an argument about whether or not I deserved to have the minimum wage, because Uber said I was cancelling so much of the work that is being offered to me, it wasn’t possible to provide me with the minimum wage, because surely I’m just at home cancelling the work. Uber said ‘this guy cancels about half the work we send him, why should we pay him the minimum wage?’
“Luckily, I did get access to enough data to actually turn that argument around. Because what Uber was not saying was the work that I was actually doing, how much work I was actually completing, how productive I was. They were just talking about the work that I didn’t do and not the work that I did do.
“Uber in their training documents were saying drivers should be doing between 1.4-1.6 jobs per hour for a 40-60 hour week. But the data showed I was doing an 90 hour week, and I was doing 1.5 jobs per hour. So at 90 hours a week I was at the upper end of productivity for somebody supposed to be working 40-60 hours a week, but yet they were telling the court that I didn’t deserve minimum wage for that.”
Farrar, who is also co-leader of the App Drivers & Couriers Union and led a successful court case in Amsterdam in April against the “robo-firing” of six Uber drivers, also said he was concerned about a growing “cross-over” between worker algorithmic surveillance and state surveillance.
“The police backed Uber recovering its London license…because Uber is a great source of intelligence to the police. In London, Uber processed 2,500 data requests from the Metropolitan police without a warrant last tear. That’s state access without warrant to that surveillance.”
You can watch the full video of the event below:
Be the first to comment