How workers and unions should tackle algorithmic management is a question we are only just starting to get to grips with. It is an increasingly important one, as algorithmic management seeps ever deeper into the daily life of workers, including those far beyond the gig economy. According to the OECD, 79% of companies use at least one form of algorithmic management in Europe today: this is the new normal of work.
Gent’s argument rests on two main stools: first, by focusing on contractual elements, such as bogus self-employment and temporary contracts, unions are not paying attention to the work process itself, in particular the ways in which control is exerted through algorithmic management. Even if workers overcame “precarity” and achieved contractual security, working conditions would not be transformed without pushing back against algorithmic control directly.
Secondly, Gent argues that when unions are paying attention to algorithmic management, they are making the wrong arguments. A focus on transparency and explainability is deemed to be a “liberal” approach to algorithmic management because it "amounts to little more than explaining the reasons behind algorithmic decisions, rather than giving [workers] direct input into those decisions”. Technology has class-relations and labour exploitation built into its very fabric; it won’t be made “fair” by being made transparent.
Gent’s case contains a lot of good sense, but he also misses or downplays some crucial elements which might frame the focus of unions in a somewhat different light.
It is undoubtedly true, in the gig economy at least, that unions have focused on the employment status question: significant resources have been dedicated to legal actions and political lobbying on this issue. It’s also true that this focus is not necessarily conducive to building worker autonomy and collective action at the workplace itself, key elements in constructing workers’ power. Despite this, we would defend the focus on tackling bogus self-employment as a strategic necessity.
Bogus self-employment is seriously damaging to workers’ organisation because of the fluidity of labour supply under those contractual conditions. Digital labour platforms have a permanent over-supply of labour because it costs them almost nothing to register more workers onto their platform and they do not have to pay workers for waiting time. That means workers can be disposed of rapidly (all it takes is the click of a button) while they can also be replaced just as easily due to the low barriers of entry to platform work jobs like food delivery. A fluid labour supply significantly reduces the bargaining power of workers.
Gent does mention labour over-supply in passing, but the focus of his book is very much on the industrial sociology of algorithmic management. While such an analysis is clearly valuable, it has to be allied to a realistic assessment of the broader, macro dynamics which impinge upon industrial relations in the food delivery sector. As well as labour (over)supply, other crucial factors include the role of finance (venture capital in particular), the legal status of workers (whether they are undocumented or working on student visas), and the (precarious) conditions in the labour market more broadly.
This does not mean that resistance is not possible in the context of bogus self-employment. Indeed, the most powerful strikes we have seen in food delivery have been under these contractual conditions, exactly because these workers can deploy their lack of employment status to carry out wildcat actions which take the platforms by surprise, and can use tactics (like blockading restaurants, for instance) which are more difficult to deploy when striking through the limits of (anti-)trade union laws. But what tends to happen is that these wildcat actions go up like a rocket and come down like a stick, often leaving little trace of permanent worker organisation behind, exactly because of their informal character. Their short-term strength is also their long-term weakness.
On the other hand, there is evidence that riders are more likely to join and engage actively in unions when they are employees. This may not lead to the same explosiveness in struggles as we have seen under conditions of bogus self-employment, but it carries with it greater capacity to build enduring collective power. It is also true that the reality of employment in a sector like food delivery can be as brutal as bogus self-employment, and can also include illegal labour practices (for instance under conditions of sub-contracting). But the wager is that the potential for unionisation is significantly higher when you at least have an employment contract you can defend, and that it will be union power which can be the game-changer over the long-term when it comes to challenging algorithmic control.
The second point of Gent - that algorithmic transparency is a red-herring for workers - fails to consider how transparency can and should be viewed not as an end in itself, but as a means to strengthening the organisational capacity of workers. Indeed, Gent goes as far as to suggest that workers may be better off with opacity, arguing that: “In the absence of transparency and a rationale for the information given to them, drivers are inadvertently liable to act in ways the system finds suboptimal.”
It would be interesting to test that hypothesis (PhD thesis, anyone?), but everything we have learnt about how workers engage with algorithmic management would suggest to us that this is not accurate. Groping around in the dark tends to lead to demoralisation and submissiveness, not resistance. When workers don’t know the extent to which they are being surveilled, the default position is to fear that the company can see everything, making them less likely to engage in activism and/or trade-unionism for fear of the consequences.
On the otherhand, when workers are able to grasp information which allows them to, for example, compare their pay rates with that of their colleagues or to understand their average pay rate over a month or a year, it provides them with informational weapons that they can use to dispute the claims of their platform bosses. It follows that if the information asymmetry between workers and platforms is a major source of industrial relations power for the latter over the former, the reduction of this asymmetry will help, not hinder, workers to build collective power.
What is certainly true is that algorithmic transparency will not fundamentally alter power relations in the gig economy. When the Platform Work Directive is transposed and workers in the EU have ‘the right to know’ and the right to an explanation on automated decisions from December of this year, we suspect that, at least initially, the real-world consequences will be very limited indeed. But that doesn’t mean we should make a virtue out of a necessity: it’s better for workers to have access to information than to be in the dark, and indeed Gent’s own research evidences this. He cites, for example, cases where workers have sought to access the information on their supervisors’ screen to better understand their productivity targets.
Where we are in full agreement with Gent is that if unions limit their demands to transparency, they will be doing workers a disservice. In the book’s epilogue, Gent highlights the case of film workers in the US who have sought to “suppress” the use of AI in their industry. ‘Negotiating the algorithm’ should also mean negotiating whether and what types of algorithmic management enter the workplace. We have argued in a study on the risks of Uberisation that unions should actively seek to block the introduction of algorithmically determined ratings/performance evaluation, pay, shift/task allocation and punishments, and to remove these forms of algorithmic control if they are already in place.
‘Cyber Boss’ is a useful contribution because Gent stakes out his position with clarity, thus facilitating a debate about work, unions and algorithmic management. We shouldn’t fear such debate, indeed it’s healthy as a movement to question our intellectual assumptions and to test ideas in theory as well as practise, especially when we are dealing with something as new to the world of work as algorithmic management. That’s how we will learn and evolve.
Ben Wray, Gig Economy Project co-ordinator
