Put simply, an algorithmic system is a set of computer codes that works inside a computer. It is this algorithmic code that creates algorithmic management. Therefore, in essence, without algorithmic systems, there would be no algorithmic management.
In whatever way it is used, algorithmic management continues to be of intense interest to managers and trade unions alike. And the use of algorithmic systems by management has serious implications for trade unions.
To alleviate such implications, algorithmic management also demands that trade unions develop new competencies and skills. Accordingly, in the wake of algorithmic management, one might like to think of trade unions’ two main areas of activities:
- Explaining AI: first of all, unions are challenged by having to explain algorithmic systems and the use of artificial intelligence (AI) in companies, and what all this means for workers.
- Shaping AI: secondly, and perhaps more importantly, unions will have to engage, participate, and shape the process of introducing and running algorithmic management.
All too often, artificial intelligence (the code), algorithmic systems (the entire setup that makes algorithmic systems useful to management), and algorithmic management (the application) are entirely new undertakings by trade unions.
Yet, unions should also be aware of what management needs to do to comply with local laws and regulations. This, for example, can include rules like the EU’s “AI Law”. Such laws and local regulations could set narrow parameters for the corporate usage of AI.
This could include strong administrative as well as criminal sanctions on companies for, activities such as violating privacy and for creating biased (racist and sexist) and discriminatory algorithmic-managerial regimes.
For trade unions, this often means not just the monitoring of how management applies an algorithmic system, but also the cooperation with management on the usage of artificial intelligence.
Despite all the AI-tech-talk in a corporate working environment, the principles of human prominence, human agency, and what philosophers call personhood should be paramount.
Beyond the often-rehearsed claims that artificial intelligence and algorithmic rules are currently obstructed by too much state regulation and bureaucracy, the very opposite is the case.
In other words, there is no serious regulation that would protect workers and would also slow down management’s attempt to move towards algorithmic management.
Instead of AI being way ahead of state regulations, algorithmic management in companies should be regulated. So far, such regulative action is not in sight, at least not in terms of labor relations.
Put differently, there is “no” protective umbrella shielding workers from the negatives of authoritarian algorithmic management.
Worse, the apostles of corporate ideologies like Managerialism all too often sideline the algorithmic exploitation of workers by overplaying an all too optimistic managerialist dream of an Uber-perfected efficiency at work.
The key corporate weasel-word is “optimization” which essentially means work intensification and algo-stress. In short, work – not even when working under the dictate of algorithms – will not love you back!
In virtually all of this, the essential obligation to protect workers fall, unsurprisingly, onto trade unions. They play a key role in defending working conditions in corporate areas where AI and algorithmic management are becoming invasive elements.
To do that, trade unionists engaged with shaping algorithmic-managerial rules will need to acquire new skills. This, in turn, means that trade union education needs to be adjusted according to the new realities at work.
Simultaneously, traditional competencies like organizing, securing workers’ rights, the right of participation in the workplace, collective bargaining, grievances, guiding dispute resolutions, etc. still exist. Algorithmic systems will not eliminate them.
In addition, trade unions also play a central role in the monitoring and supervision during the actual operation of algorithmic systems, and even more so when algorithmic management assumes power.
Before all that, it remains the responsibility of trade unions to assess corporate-algorithm rules throughout its three phases of algorithmic management, namely: conception, introduction, and operation. This starts even before the initial design processes are conceptualized and orchestrated by management.
As always when human beings are involved, virtually all AI activities become, inevitably, an ethical issue. Thus, trade unions can make use of what management likes to call business ethics. They can do so rather particularly well when setting up an “AI ethics committee”.
In all this, trade unions could play an essential role in consultation, co-decision-making, and organizational opinion-forming by getting involved at an early stage.
This opens up a third option which supports the regulated, far-sighted, and agreed introduction of algorithmic systems.
Such a third option operates between (1) managerial despotism and (2) managerial whims. In addition, the third option operates with the participation of trade unions while, simultaneously, also safeguarding what is known as AI ethics. The third option is neither 1 or 2:
- Managerial Despotism: The first extreme is management’s authoritarianism leading to despotic management, and
- Managerial Whims: Secondly, neo-liberalism’s hyper-unregulated laissez-faire policy that often comes under the ideology of fundamentalist libertarianism. This gives, in the absence of a state, corporate bosses a free hand to act at will and on whims.
- Negotiated Algorithmic Systems: The third option is about finding common agreement between labor unions and management. It is the negotiated introduction, monitoring and management of algorithmic systems.
Yet, a trade union influencing “workplace regulation” does not issue a clearly defined package to be stringently followed. Instead, it is a cooperative attitude that enables trade unions and management to negotiate algorithmic systems. For responsible pro- union regulation, there is no “one best way”. It is a cooperative solution inside a pluralist setting.
In other words, pro-union workplace regulation is about applying local labor laws and labor practices to algorithmic systems.
Moreover, it promotes the freedom of workers in the absence of managerial domination. In that, workplace regulation that is negotiated between unions and management becomes a rather hands-on tool for the implementation of algorithmic management.
In all this, trade unions are not alone. Classical labor relations operate with three sets of actors: a) management, employers and companies; b) workers and trade unions; and c) the state and external agencies.
Such a third-party actor between labor and management can be the regulating and sanctioning state, but also other civic institutions like universities, a workplace ombudsman, health and safety institutions, and advisory bodies.
These can assure that potential pathologies like algorithmic bias and discrimination can be prevented “before” management creates, accidentally or purposefully, a dehumanizing bias or discrimination.
Workplace regulations enforced by the state or trade unions – that are, for example, achieved through collective bargaining – can compel companies to draw up a set of workplace rules which can be implemented. For that, it is useful to have established trade union competencies, and, if possible, rely on recognized collective bargaining practices.
Regulations that are shaped by trade unions for algorithmic systems is a monitoring strategy that builds on two key parts: union competencies and collective bargaining.
For trade unions, it remains imperative to set the scope, the meaning, and the purpose of algorithmic management as well as adjacent algorithmic systems. Inside such a consultative and co-decision-making framework, unions need to be clear about the “new” role of trade unions when representing workers’ collective interests.
This, of course, occurs inside the context of algorithmic management embedded inside a work environment. Rather inevitably, trade unions will have to identify the types of algorithms that are applied. Basically, these can take any one of six forms of algorithmic management (AM):
1) No AM: Standard management prevails as no algorithms are used.
2) Assistant AM: Algorithmic management is used to assist management.
3) Partial AM: Parts of management uses algorithms while managers remain core decision-makers.
4) Conditional AM: Most management decisions are handed over to algorithms.
5) High AM: Most of corporate control and worker discipline is handed over to algorithms.
6) Full AM: Algorithms define performance and control with minimal management involvement.
Based on this, it helps trade unions to be clear about which type of algorithmic management (1-6) is to be used. These six levels of algorithmic management indicate a diverse set of interfaces between management (human) and technologies (machine).
Even when used for “optimization” (work intensification) or advanced worker surveillance, algorithmic management remains not only as a human-made invention, but it also happens inside a traditional work environment shaped by the imperatives of capitalism as well as management’s twelve almost “unalterable” and semi-pathological characteristics. Some of the worst aspects of doing business are:
Profits: Management must maximize profits – this is the only real raison d’etre for a business to exist.
Amorality: Management sees itself as an amoral institution shielded from obligations of personal morality.
Growth: Management and managers are hooked on the profitable growth of a company or corporation.
Quantification: Management must convert subjective values into objective quantities – only numbers count.
Exploitation: By definition, management engages in the exploitation of workers and the environment.
No Limits: Many corporate apparatchiks truly believe that “the sky is the limit” – they can do what they want.
Hierarchy: Management is always a hierarchical organisation: kiss up and kick down remains the motto.
Dehumanizing: Corporate apparatchiks see workers as ciphers, cogs of a wheel, and replaceable like a tool.
Aggression: Managers like to place every person in fierce competition with each other – competition is good!
Virtually, all of these can and in fact, are likely to be enhanced (read: made worse) by algorithmic management and adjacent monitoring software applications as well as something called “people analytics”.
Management sees this as “the systematic identification and quantification of workers to drive business results” (read: corporate profits).
Algorithmic management as well as algorithmic systems require access to large data. Such enormous volumes of data, that are way too large for standard computer programs, are commonly known literally as “big data”. Algorithmically created “big data” are an absolute and unavoidable imperative for effective algorithmic management.
Trade unions should remember that these data come from workers and are produced by workers (IR workers). Most importantly, these data are often used “against” workers.
Like a drug addict hooked on crack-cocaine, algorithmic management depends on a never-ending flow of information about workers that can be processed via algorithmic systems.
This distinguishes “algorithmic” management from “conventional” management. While seemingly personal relationships are replaced by algorithmic systems, the creation, collection, analysis, interpretation, and eventual use of these data is “always” done by people, manager, and corporate apparatchiks.
Despite the diversity of algorithmic systems and the ranges of algorithmic management (1-6) that are used, the technology applied in algorithmic management still has one common factor: it is made by humans.
It is made “for” or “against” humans. And this is so, even though corporate apparatchiks like to give algorithmic management the appearance of being objective, technical, engineering-like, neutral, and independent.
In spite of being called artificial intelligence (AI), algorithmic systems are tools to collect data in order to find patterns, trends, and correlations. This is pretty much it.
Yet, algorithmic management uses algorithmic systems to achieve a more efficient workflow optimization (read: intensification), improved performance (read: control) and the reduction of costs (read: lower wages). In sum, higher profits. Ultimately, this is what is driving algorithmic management.
To achieve work intensification and increased profits, algorithms are essential for algorithmic management because they can identify useful patterns in datasets.
They do this “for” management – not “for” workers. In the next step, they assist management in making decisions based on patterns that AI has found. In other words, algorithmic systems can be used to monitor and ultimately control workers and workers’ behavior.
Pretty much for or all of this, employee behavior needs to be interpreted by management given it is management that uses algorithmic systems and the data it produces as performance measures.
All in all, the advent of algorithmic management almost forces trade unions to become more active, to gain knowledge about algorithmic systems and artificial intelligence.
On algorithmic management, unions need to push forward the concept that the world of work remains principally a human environment. However, such systems are made to appear technical, mathematical, objective, and neutral by corporate apparatchiks. In reality, however, algorithmic management is a human endeavor. This, inevitably, involves human ethics – a moral code.
Yet, this also includes the managerially (workplace) and mass-media (society) induced hallucination that algorithmic systems are flawless and that moral and legal responsibilities are now handed over to algorithmic systems.
In reality, however, there still are the managerially created algorithmic discrimination, algorithmic dehumanization, and the managerial despotism at work – now enhanced through algorithmic management.
Challenges to workers and trade unions also include a sheer uncontrolled stream of data that are processed by managers using AI-technologies and are often used by management in non-transparent decision-making processes.
Under algorithmic management, these remain hidden inside mathematical equations, computer coding, and algorithms. Worse, corporate apparatchiks use algorithmic systems to present their decision as neutral and objective.
For the practical defense of workers against algorithmic management and for the safeguarding of collective labor rights, trade unions need to work toward organizational transparency during the implementation of algorithmic systems.
This includes the original planning and subsequent design stage as well as the “personalization” stage (read: its application to workers).
The final stages are the actual operation of algorithmic systems and the evaluation of its impact (read: profitability effectiveness) as organized by management.
In any case, trade unions might remember that they are the only organizational institution that can force management to deliver transparency about algorithmic systems. It remains imperative for trade unions to realize that algorithmic management has two key raison d’etre:
- it assists management in decision-making; and
- it is introduced to enhance corporate profit-making.
Corporate legitimizer and management bedfellow McKinsey believes that using algorithmic systems will be a profitable endeavor. Globally, this will be in the vicinity of $4.4tr per year ($4,400,000,000,000) – no small change. This is more than the GDP of Japan ($4.2tr).
All of this indicates that the increased introduction of algorithmic-corporate regimes will assist management in controlling workers. Algorithmic management is definitely a fundamentally new form of working environment.
As a side effect, top-management might even be able to shrink or eliminate middle-level managers as this level of management might no longer be needed.
In all of this, state-sanctioned regulatory frameworks are a good way to negotiate organizational issues like the shift from traditional management towards algorithmic management.
This might diminish the much-favored laissez-faire ideology of corporate apparatchiks that allows them to rule their little kingdoms of techno-feudalism at will.
Born on the foothills of Castle Frankenstein, Thomas Klikauer (PhD) is the author of 999 publications, including a book on Managerialism and The Language of Managerialism.
