AI and Cognitive Automation
According to a recent proposal by the EU Commission, ‘artificial intelligence system (AI system)’ means ‘software that is developed with one or more of the techniques and approaches […] and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with’1.
We see more and more automation of cognitive tasks through AI-based systems in the workplace, some more prevalent and others more subtle. When applied in workplaces, AI-based systems are mostly narrow in function and lack general intelligence. This means that their incorporation into workplaces is likely to make particular tasks of an occupation redundant for humans to perform, rather than eliminate entire occupations outright, as they need critical thinking and decision making on a level of complexity not yet achievable by AI.
Impact on tasks, jobs, and sectors
A taxonomy of tasks based on extensive literature reviews, reflect those which are being, or are likely to be automated by AI-based systems: a) person-related, b) information-related, and c) object- related. Person-related tasks involve a worker interfacing with a person (customer, patient); information-related tasks with information (data-processing, software programming, etc.); and object-related tasks with objects (vehicles, drones, etc.). While AI-based systems are able to automate tasks of all three types, the literature demonstrates that, for now, information-related tasks are the most suited for automation by AI-based systems, followed by person-related tasks.
Occupational safety and health (OSH) considerations
Technological advancement is often a double-edged sword in that it presents both risks and opportunities. Many are hopeful that advancements in AI-based systems will continue the historical trend of eliminating dangerous jobs. A prominent example of this would be the advent of self-driving vehicles. Approximately 9.3 individuals per 100,000 die each year in traffic-related fatalities in Europe. A considerable proportion of people on the road at any given time are commuters driving to work, ride service providers, or truck drivers transporting goods and services. It is widely believed that the rise of self-driving vehicles could dramatically minimise this cause of premature death.
The automation of cognitive tasks by AI-based systems will continue to eliminate repetitive and boring kinds of clerical or administrative work. As intelligent programmes more efficiently process forms, applications, claims, legal documents etc., it will no longer be necessary for humans to complete these ‘mind-numbing’ and alienating tasks. If analyses and recommendations made by AI systems prove to be effective and accurate enough to be considered worthy of trust and more broadly followed, administrators could potentially either supervise more projects or focus more on the human-centred side of their job. This would constitute a shift to potentially more engaging kinds of work.
Another related hope is that AI-based systems can reduce the burdensome and emotionally demanding nature of some occupations. Care work, for example, is currently a very high-touch occupation. That is, carers have to constantly engage in physical and emotional interaction with patients when completing the entirety of their job duties. If some aspects of providing care can be offloaded onto smart devices, this could transform care work into an increasingly low-touch labour process, and thereby curb the emotionally challenging dimensions of the work as it is performed now.
Considerations on an organisational and legislative level should also go towards the topic of cybersecurity. AI will increasingly handle threats in that area2 by taking over tasks like high accuracy threat detection with higher efficiency compared to human intervention3, effectively supporting IT workers and cybersecurity specialists. However, it also can become a target of such attacks. The OSH related consequences of this can vary widely. The risk of a system being targeted needs to be considered especially if the AI is handling sensitive or personal data. However, it can also play a key role in protecting said data2.
Additionally there is a lack of tools supporting companies in conducting a thorough risk assessment of AI-based systems when it comes to OSH. While there are some publications addressing emerging risks associated with AI in the workplace, like EU-OSHA policy brief “Impact of artificial intelligence on occupational safety and health”4, and the European Commission regulatory framework5 on artificial intelligence, which defines four levels of risk in AI applications, these provide a first overview. Specific tools developed to assess and AI and its surroundings to highlight possible risks specific to this set-up could support companies in ensuring OSH.