Algorithmic control

Work in the social sciences on algorithmic systems can inform how unions address their impact on the power balance between workers and employers­.

Algorithmic technologies have provided employers with new tools to exercise control. An influential study which popularised the term algorithmic management—based on seminal research in the 1970s by the American management professor Henry Mintzberg—found that they did so by allocating, optimising and evaluating work. These algorithmic practices are spreading from digital platforms to conventional firms and the public sector where the vast majority of employees work.

As the use of algorithms and artificial intelligence in working life has proliferated, a related policy discussion has developed. Social Europe has played an active role, with its series respectively on The transformation of work and Artificial intelligence, work and society.

So far, the scholarly input to the debate has been dominated by law and economics, especially by descriptive and quantitative research. While this is and will continue to be important, there is an increasing need to develop theory, to understand better and to shape the influence of algorithms and the various forms of control of workers in the digital economy. To do this, we must also take stock of other findings from the social sciences and humanities.

Different perspective

Algorithms at Work: The New Contested Terrain of Control is a comprehensive literature review by researchers from Stanford and the Massachusetts Institute of Technology. The researchers analysed more than 1,100 peer-reviewed, empirical, social-science articles published since 2005, concerning algorithmic, crowd or platform technologies.

Their purpose was to provide a different perspective than that prevailing in management and economics, which is primarily ‘the impact of algorithms in terms of economic value derived from greater efficiency, revenue, and innovation’. Instead, the authors wanted to explore how algorithms provide managers and employers with new ways of controlling workers, and the consequences.

They find that algorithms are more encompassing, instantaneous, interactive and opaque than previous technological and bureaucratic systems of control. Of course, this does not make earlier research irrelevant—quite the contrary. The Stanford/MIT researchers use existing theory to structure the reviewed literature.

More specifically, they draw on labour-process theory from a 1979 book, Contested Terrain: The Transformation of the Workplace in the Twentieth Century. This argues that employers use three control mechanisms to obtain desired outcomes from workers: direction, evaluation and discipline.

The Stanford/MIT researchers expand this typology by identifying new control mechanisms they call the ‘six Rs’. They find that algorithms are used by employers to:

• direct workers by restricting and recommending,
• evaluate workers by recording and rating and
• discipline workers by replacing and rewarding.

This ‘new contested terrain between employers and workers’, the authors argue, is specific to algorithmic settings and has many consequences. These include workers’ experiences of being controlled through the six Rs, new ways of collectively and individually resisting this control, and the emergence of new algorithmic occupations.

Similar challenges

The challenges for scholars who conduct research on the social effects of algorithms are actually similar to those faced by labour activists and trade unionists who attempt to shape and counteract the use of algorithmic systems in working life. For instance, access to algorithms is often restricted due to their proprietary nature and the particular skills needed to understand them. They are often embedded in larger systems which make algorithms difficult to scrutinise and contingent in ways that are hard to disentangle. Due to these similarities, the methodological approaches used by social scientists to study algorithms might serve as templates for trade unions’ strategies to counteract algorithmic control.

In a widely cited article, the British geographer Rob Kitchin lays out six methodological approaches to the empirical study of algorithms, which could just as easily be taken as tentative advice for trade unions:

• examine pseudo-code/source code,
• reflexively produce code,
• reverse engineer,
• interview designers or conduct an ethnography of a coding team,
• unpack the full socio-technical assemblage of algorithms and
• examine how algorithms work in the real world.

Any one or a combination of these methodological approaches could provide useful guidance for unions in making sense of and influencing how employers use algorithms in ways that affect their respective sectors and members.

Complex systems

A final, but fundamental, insight for trade unions can be drawn from critical algorithm studies —a repudiation of the common view of algorithms as static technological artefacts, or ‘black boxes’ to be opened up and explained.

Instead, as suggested, for instance, by the American anthropologist Nick Seaver, algorithms should be understood as complex and dynamic systems. These change over time and in context and consist of interrelated components, of which the computational code (what is commonly referred to as ‘the algorithm’) is only one. Other aspects include cultural and social factors which shape how humans interact with and form part of such systems, legal and regulatory issues, and the organisational and economic contexts in which the systems are implemented.

Algorithmic systems provide new ways of exercising power in labour relations. This insight has direct relevance for trade unions, policy-makers, think tanks and researchers who wish to comprehend and shape their use in working life.