Trade Unions: AI Brings Higher Job Quality But Trust Issues

OECD researchers surveyed 14 large trade union confederations across OECD countries to better understand current sentiments about the future adoption of AI in the workplace.

According to their findings, published in the OECD Employment Outlook 2023, the main concerns among trade unions right now center around the trustworthy use of AI (which includes, among other things, addressing risks to do with privacy, transparency, security and accountability), as well as the rapidly changing skill demands (associated with high re- and upskilling costs for both businesses and workers) and the physical and mental health risks (for example through excessive AI-based surveillance).

In addition to this, over a third of respondents said that AI poses the risk of weakening collective action and social dialogue. The report cites fears held by the British Trade Union Congress (TUC) that the use of AI could change the employment relationship in a way that “blurs accountabilities of decisions”, which may ultimately affect trade unions and other social partners’ ability to represent workers’ and employers’ interests.

But trade unions are weighing up these cons with the recognition that AI can bring some positive changes too. As the following chart shows, a majority of respondents said that its adoption could improve jobs for workers through effective collaboration between workers and AI, for instance, resulting in a reduction of working time, a possible focus on more interesting tasks and increased work autonomy. At the same time, this data shows that many working within the trade unions sphere consider AI to be creating new tasks and jobs, such as the roles of data analysts, computer engineers, scientists and network experts.

Reductions in human bias in HR and management processes are another perceived benefit among just under a third of respondents. Where it is argued a machine may be able to increase fairness for discriminated groups and minorities, critics would argue that caution is still needed here since AI is usually modeled on biased data sets.