Artificial intelligence is now finding application across the world of work in all kinds of technologies. This has implications for products and work equipment, as well as the tools used to design activities, processes, workplaces, and management systems. It has the potential to help shape work and workplace OSH services productively and beneficially. At the same time, its impacts on employees and organisations need to be interrogated critically. The challenges for human-centred work design that arise in this contested field make it necessary to think about occupational safety and health from a new perspective.

Sweeping changes in the world of work

AI is penetrating organisations and, as it does so, coming to be used in many different contexts. Assessing its opportunities and risks for occupational safety and health and implementing it in human-centred ways are particular challenges because AI applications are continuing to develop at a dizzying speed and the range of technologies and systems classifiable as AI is hugely diverse.

Apart from the debate about potential effects on the labour market, it is therefore necessary to investigate numerous other questions relating to AI’s deployment in the world of work, which will require AI technologies to be surveyed comprehensively and systematically. Both digital and ecological transformation processes and the demands of sustainable implementation will have to be taken into consideration in such investigations.

This evolution is becoming particularly apparent in digitally mediated service sectors such as platform-based delivery and logistics services, where AI-supported systems are increasingly taking over personnel management duties, including assignment planning, performance monitoring, and payment. This form of work organisation is referred to as “algorithmic management” (AM). AM is characterised by the use of algorithmic systems to automate, support, or control management functions that have traditionally been carried out by managers. Against this background, the central question is how algorithmic management systems can be designed so they are fair and transparent, and administer employees’ work equitably.

How can algorithmic management (AM) be designed fairly?

The paper “Fairness Engineering in the Algorithmic Management of Platform Work: A Comparison of Four Approaches in the Food Delivery Industry” draws on a systematic review of recent academic literature to examine how algorithmic management can be designed fairly. To do this, it utilises the concept of fairness engineering, which denotes the targeted integration of fairness constraints into the development, implementation, and application of algorithmic systems. In the context of algorithmic management, fairness engineering is aimed at identifying bias in data-driven decision-making processes, reducing discrimination risks, and enabling transparent, comprehensible management decisions. The paper discusses examples of employees in platform-based delivery and logistics services who are affected to a particular degree by AI-supported personnel management systems, and analyses how fairness engineering approaches could contribute to fair work design. Platform-based delivery and logistics services are particularly relevant because algorithmic systems control central aspects of work organisation in this sector such as task assignment, performance monitoring, and payments. They consequently have direct impacts on employees’ working conditions, incomes, and autonomy.

Challenges from AI

Changes and new design goals driven by AI technologies are found at various levels of OSH action: AI systems have to be designed safely, as required by product safety law for example. Furthermore, their deployment within organisations is to be implemented safely to ensure operational safety standards are maintained. But AI technologies also have to be used to design work more safely and healthily, for example by targeting how they are deployed to ease employees’ workloads, to support risk assessments, or as research tools, particularly for the analysis of risks.

AI is transforming work systems at the micro, meso, and macro levels. This has implications for work tasks, activities, processes, and organisation, as well as novel kinds of work equipment. The latter include collaborative robots, self-learning systems, intelligent protective clothing, work assistance systems, and driverless transport systems. In addition, AI-supported systems are increasingly being deployed for algorithmic management as entities that control the organisation of work by influencing the assignment of tasks, the pacing of work, and performance targets. In view of such far-reaching innovations, it is necessary to take a human-centred approach when planning the assignment of roles and designing the interactions between humans and (AI-supported) work equipment.

The use of AI in the world of work may have the consequence that employees experience new or more-severe kinds of pressure. Examples include rising work intensity, a growth in repetitive work, and deskilling, not to mention the declining transparency of technical systems and growing complexity, greater dependence on technology, the lower levels of control with which this is associated, and the danger of social isolation. It is necessary to track these changes with empirical data, which was one of the motivations for conducting the . OSH criteria and instruments have to be scrutinised, adapted, and refined in line with real-world developments in order to keep pace with innovations in AI. These make corresponding adjustments necessary in the fields of regulation and standardisation as well.

AI's potential benefits

Nonetheless, AI applications can also provide employees with meaningful support and strengthen OSH provision. Today, cognitive assistance systems are already capable of supplying information about workloads and stressors and so positively influencing employees’ health habits. AI is making it possible for body sensor networks to improve the measurement of physical stresses. AI-based image evaluation techniques will soon be assisting the detection and classification of microscopic hazardous substances. If they are well designed, AI technologies can hold out the prospect of alleviating the pressures on workers who perform repetitive tasks. In principle, it is possible to configure AI systems transparently and comprehensibly. New opportunities are also opening up in particular to design work inclusively by creating options to individualise the support that is provided. In the context of algorithmic management, AI systems could help to distribute workloads and the stresses to which they give rise more responsively to individual needs, provided these systems are designed humanely and factor in employees’ health and well-being instead of exclusively concentrating on efficiency optimisation targets.