Lessons from Regulating Automated Management: What Automation Teaches Us about Supporting Workers

Europe and the United States are both navigating fierce debates about how to support workers in increasingly automated environments. Governments are trying to determine how to classify outsourced workers completing jobs through platforms like Uber, DoorDash, and Deliveroo, even as they face constant pressure from lobbying groups to water down the worker protections offered in legislative efforts like the European Union’s Platform Work Directive.

In this context, we work on automated and outsourced work environments from different angles. One of us focused their academic research on automated and outsourced work environments in the United States while working toward design and strategy solutions to close the last mile in technology systems and build tools for human-machine partnerships. The other has worked for nearly a decade in policy and policy-oriented research, working in the EU with governments and workers to think about how technology impacts the world of work. Together, we are thinking about what automated work environments teach us about supporting workers more broadly, including through policy, strategy, and design for better technology tools. What we didn’t expect to find in our work on automated management tools were important lessons about supportive work environments that could be applied with human managers, too.

Why some workers still prefer working with an automated middle manager compared to other low-wage work environments

It’s always tricky to respond to a contradiction. When we each asked, "How does this job compare to others that you’ve had?" We both had several interviewees say, "I like it better than some of my other low-wage, unstable jobs."

The American gig workers we spoke to talked about the constantly unstable schedules, shifting working expectations, and emotional abuse they endured from their managers in other low-wage work environments. The promise of “be your own boss” was a little closer in this platform role than in jobs where their managers reminded them they were expendable and used their power to force them into difficult working schedules that negatively impacted their health and well-being. It wasn’t always perfect, but those who had not navigated platform deactivation before noted that they felt like they had a little more control in these roles than they had in others. Their platform work gave them some sense of stability because they could set their own hours, and the expectations/rules were more consistent (until they weren’t).

In Europe, the adverse effects of algorithmic management in platform work are well known and documented: increased feelings of precarity, increased stress levels, fragmented income streams, deliberate design choices by the platform companies to hinder worker organizing and communication amongst workers, difficulties in seeking redress for automated or semi-automated decisions by platforms, power asymmetries in setting remuneration levels and “renting” platform accounts to those in vulnerable positions, e.g., migrants, forcing them into undocumented work agreements with third parties. Yet, when we interviewed platform workers just last spring, many still found that, for their current situation, the fragmented and precarious, yet flexible, working arrangements were suited to their needs.

Although these views have eerie echoes of the promotional recruitment material from platform companies themselves (“be your own boss”), some of the workers we talked to in Finland genuinely appreciated the straightforwardness of their gig work. Despite reporting similar issues as those listed above, the workers found that the flexibility and convenience offered by the platforms were a big draw for them in their current situations, at least in the short term.

Can these two seemingly opposite viewpoints be squared – the well-documented precarity and powerlessness in the face of algorithms and the draw of clarity and flexibility for some? In our different research contexts, we each found stories from the workers themselves explaining that some of them still preferred their automated work environments to their previous roles with human managers. This is a tricky thing to make sense of in our current moment, where algorithmic managers are criticized for being black boxes of surveillance, oppression, and negligence. So, what does the automated manager offer us as we think about how to improve working experiences for all workers? When executed with the intent to communicate clearly on its design and expectations, the automated manager requires less emotional management and “guessing” than their human managers did.

What we’ve learned as we try to regulate work environments with automated management tools

One critical problem platform workers encounter in their routine roles is the absolutism of an algorithmic judgment and the stone walling they face when they need to appeal a decision or receive help in their work. In any data system, some data can be mislabeled or misinterpreted. Engineers spend time developing tests for predictable environments and occasions to try to rule out errors. Still, a lot of human environments, rather than lab-controlled ones, have more variation and diversity than we can predict successfully (even with advanced tools like AI). Countless stories from workers under strict and mechanistic algorithmic management share the leitmotif of a lack of contextual information in the design and operation of these sociotechnical systems. Often these results in absurd situations for workers: a driver having to decide between obeying the route optimization algorithm despite knowing that the road is blocked, a courier being told she can cycle a trip in a short amount of time despite there being an immense elevation difference on the route or a shift scheduling algorithm understaffing the opening shift because it doesn’t understand all the work involved. In cases like these, often simply asking a worker could have provided the necessary sanity check for what is and is not feasible.

A key feature of the EU Platform Work Directive requires introducing a human back into the automated loop around important workplace decision-making, like firing or deactivating a worker from the platform. Keeping a person as an anchor in the overall technical system helps overcome some of the limits of an automated environment to make the system more effective and for human workers to get the support they need in their roles. It also dulls some of the absolutism and cruelty workers face when an algorithm unfairly labels them as bad actors.

In Europe, both proponents of the Platform Work Directive and those in favor of an “AI at work” directive that would apply similar provisions to all workplaces call for a baseline of transparency in how algorithms work and to ensure that humans are making the automated or semi-automated decisions that concern workers.

Transparency, however, is only a baseline: just seeing how things work is not enough, but workers and their representatives in the EU also know that workers must be able to influence and negotiate how algorithms are developed and deployed. Although legislation pertaining to collective bargaining on algorithms at the workplace already exists on the national level and some collective bargaining agreements in Europe have now included related clauses, experts, unions, and political parties are pushing for a clearer reaffirmation of collective bargaining rights through the EU level. Although many of these demands concern workers' rights to transparency, participation, negotiation, and redress, the examples above also speak to another related facet: negotiation over algorithms is also about co-design, ensuring that the algorithms implemented work the way they should. This should be a shared concern of both companies and workers, guaranteeing a transparent and clear set of rules.

What automation regulation teaches about ways to support workers more broadly

As we discussed our results, we realized there is something important we can learn from this feedback and how we think about labor reform. The platform workers challenged us both to think about what this arrangement teaches us about other kinds of jobs in the economy.

The highest performing teams inside are often a mix of strict rule and process followers mixed with teammates who are willing to encourage a little more risk-taking and improvisation depending on situational context. Suppose algorithms can be made more transparent about the rules they follow, and organizations that replace managers with these automations engage in better co-design with their teams. In that case, people can provide an important form of improvised problem-solving and creativity that makes their work more effective and interesting.

We learned that people are necessary decision-makers and designers within the larger technical systems that make platform work possible. So we asked: is the lesson from the happier platform workers about places where processes and expectations could be more clearly designed and implemented between workers and their employers in their non-automated job environments, and what could this look like?

In the United States, the implication may be that workplaces have gone too far towards just-in-time scheduling and staffing. This staffed model eroded and made it difficult for firms with this structure to maintain reliable services. Workplaces like Amazon have churned through so many workers that they are running out of people to hire. Workers picked jobs with better working conditions. Customers reacted to unreliable services and distrusted these businesses. Our research implies that there is value in communicating and offering clearer, consistent expectations for workers in many different environments. There is a balance to an adaptable working environment and one with clearly defined expectations and consistent conditions that produce overall better results, for the workers and the broader market. Workers can and should be included in process design and implementation – this keeps the overall system more sustainable and effective.

In Europe, where unions are more likely to participate in co-designing their workplace conditions, the conversation is looking further ahead to who counts as self-employed within the gig work system vs. who operates like a full-time employee at these platforms. In these environments, we consider those working full time (40+ hours a week) as employees of the firm and entitled to some of the regular benefits they would receive as employees. Gig work platforms, like other technical tools, keep clear numbers on user activity and have information on who would be these “super users,” compared to those who work part-time roles (15-25 hours) and operate as “consistent” users within their churn calculations. The others working on a couple of projects a week (5-10 hours) look like the normal contractors and are likely “moderate” users. There are clear ways to align some of these internal churn calculations that the platforms already monitor, and use them to guide some of our legislation to support these workers. Especially those in full-time gig work arrangements should be included in the co-design of their work environment, whether with an automated manager or a human-led team.

In both cases, the evidence is clear: low-wage work is often a miserable experience that relies on a massive supply of workers so these companies can churn and burn through the workforce. What we propose is a look at what a sustainable firm model looks like for workers and the platforms themselves. Smart managers will include workers deeply in the co-design of workplace processes – it benefits your team, your workplace, and your clients.