Workers must have a collective voice in the design of artificial-intelligence systems for the workplace.
Last month, the United States trade-union centre AFL-CIO and the technology giant Microsoft announced a partnership on artificial intelligence and its effects on workers. The partnership will have three goals: sharing information with workers and their representatives, incorporating worker opinions and expertise into AI systems and proposing public policies that equip workers with the skills needed for an increasingly AI-powered economy. The partners said workers’ voices had to be central for the AI transition ‘to improve work while creating richer possibilities for our lives on the whole’.
Although little is known about the terms of the partnership beyond the joint press release, this agreement represents a significant step in the regulation of AI. As the professor of labour law Valerio De Stefano has pointed out, most discussion of AI’s impact on workers has focused on ‘quantitative’ aspects—how many jobs will be created or lost through its deployment. ‘Qualitative’ aspects—the effects AI systems (especially algorithmic management technologies) have on job quality in almost every sector of the economy—have not been to the fore. Yet automated human-resource systems, notably in the ‘gig’ economy, have had significant implications for workers’ occupational safety and health and fundamental rights (such as to privacy and personality at work and access to collective bargaining).
The AFL-CIO/Microsoft partnership does not simply equate the challenge of AI in the workplace with cushioning occupational shocks but also aims to promote worker-centred designs in these systems. The mechanisms of consultation and information—learning sessions, on-demand digital resources for union leaders and Microsoft-hosted labour summits—will be the keystones of this partnership.
Dependent on goodwill
Due to the precise nature of the agreement not being made public, it is however unclear how the AFL-CIO can ensure that the information it receives from Microsoft is of sufficient quality and that the feedback it provides will be taken seriously. This partnership is neither a collective agreement nor, it seems, any other sort of binding contract, meaning it will depend entirely on the corporation’s goodwill.
Although Microsoft has in recent years recognised unions in a series of its subsidiaries and in 2022 declared itself neutral towards efforts to unionise its US-based employees, it is a profit-driven company which could change its attitude towards organised labour in the future. If that were to happen, it is unclear if the AFL-CIO would have any redress.
Moreover, there are questions related to the scope of the partnership. Microsoft is the principal backer of the poster child of AI development, OpenAI, which is valued at $86 billion. The complex and opaque relationship between Microsoft and OpenAI however raises doubts as to whether the developers of ChatGPT will be involved in the partnership with the AFL-CIO—not involving OpenAI would though represent bad faith on Microsoft’s part.
Information asymmetry
European workers are far better protected than their American counterparts when it comes to surveillance and automated decision-making in the workplace, thanks to the General Data Protection Regulation and higher rates of collective-bargaining coverage. While this does not mean the use of AI systems at work should be ignored, it does constrain deployment of some of the more egregious elements of algorithmic management in the US.
Europe is also a ‘net importer’ of AI technologies, with much less development taking place than across the Atlantic. So there are few, if any, AI ‘producers’ with which European unions can engage. Yet the European enterprise survey on the use of AI, published in 2020, found that 42 per cent of European firms had adopted at least one AI technology, with another 18 per cent planning to adopt these systems in the next two years.
The key takeaway of the American labour-technology partnership for European trade unionists should be the importance of securing technical understanding of AI systems. That can come through bipartite information and consultation mechanisms but also, recognising the above concerns, by developing in-house AI expertise.
Collective bargaining
Collective bargaining is the most effective mechanism for regulating AI in the workplace, especially considering the shortcomings of the European AI Act. Beyond the 2020 European Framework Agreement on Digitalisation, collective agreements on the use of AI in the workplace have been reached in Spain and in insurance at European level. Given however that information asymmetry is one of the key determinants of inequality of bargaining power, unions must ensure that they do have the expertise to negotiate the most favourable agreements possible.
On a wider canvas, meanwhile, the AFL-CIO/Microsoft partnership confirms that social dialogue is a catalyst for human-centred innovation. It is not a roadblock to be swept away.
Philip Freeman is pursuing a research traineeship with the European Trade Union Institute.