Responsible adoption of artificial intelligence through collective bargaining agreements

The PSI Digital Bargaining Hub is a free and publicly accessible online resource designed to assist trade unions in negotiating the complexities of digital transformation, including AI and algorithmic management through collective bargaining. The hub provides information about eight key themes and 28 related sub themes with respect to digitalisation of work. It also provides access to real-world bargaining clauses via the database of existing clauses collected from unions across the world. The Hub was launched in April 2023 with around 140 clauses, the hub now includes more than 500. It is available in English, Spanish and French and is classified under the following themes:

• Definitions and Scope
• Involvement, information and consultation
• Equity, diversity, and inclusion
• Employment, jobs, skills and lifelong learning
• Telework, working time, work-life balance and platform work
• Data rights and data protection
• Digital tools, artificial intelligence, and algorithms

-Technologies and tools: Information and scope of use

-Technologies and tools: Restriction clauses

-Transparency and design of tools, artificial intelligence (AI) and algorithms

-Intervention in AI and algorithmic systems

• Health and safety
• Union communication, organising and representation

Section 6 of the Hub, specifically focuses on digital tools, artificial intelligence, and algorithms. It offers comprehensive guidance on how to address the challenges and opportunities presented by these technologies. It has four sub themes in the context of workers’ rights, digital tools, AI and algorithmic management.

• The first sub theme is related to information-sharing practices with respect to the use of digital tools at work and the importance of clearly defining their use.
• The second sub theme addresses possible limitations on the use of digital tools and technologies in the context of electronic monitoring and surveillance equipment.
• The third sub theme concerns the importance of transparency when designing and using tools, AI and algorithmic management. The design and functioning of automated systems at work should be clearly defined and transparent for all.
• The fourth sub theme deals with intervention mechanisms in AI and automated systems. The human in control principle should become the norm in this respect.

PSI also notes that clauses addressing the use of AI should be accompanied by clauses covered in the other sections of the database, particularly in regard to Involvement, information and consultation (theme 1) and Data rights and data protection (theme 5).

Below clause from the collective bargaining agreement of the Writers Guild of America that sets the boundaries on the use of generative AI technologies shows how the database looks like:

In March 2024, UNI Europa and Friedrich-Ebert-Stiftung created a database of collective agreement clauses regarding the negotiation of AI and algorithmic management. This database serves as a repository of negotiated agreements that address the use of AI and algorithms in the workplace and highlights the role of collective bargaining in managing the impact of these technologies. The database includes 23 collective bargaining agreements, which can be browsed selecting specific topics, sectors, countries and validity.

Algorithmic management refers to the use of algorithms to oversee, manage, and sometimes control workers’ activities, often through automated decision-making processes. This practice is becoming increasingly prevalent across various sectors, driven by advancements in AI and big data. While algorithmic management can enhance efficiency and productivity, it also raises significant concerns about privacy, fairness, and the potential for worker control and exploitation. In 2020, UNI Global Union developed a comprehensive guide to algorithmic management aimed at equipping trade unions with knowledge and tools needed to address the challenges it poses and ensure workers share in the gains made possible by technological advances. The guide provides practical advice on how to negotiate fair and transparent use of algorithms in the workplace so that all the parties can bargain over and mitigate any potential harm and seize the opportunities brought by algorithmic management. These include:

• Reduce rather than increase bias and discrimination
• Improve worker flexibility and autonomy
• Improve the quality and fairness of management decisions by providing managers with independent, data-driven advice

The guide explains basic concepts and mechanisms of algorithmic management focusing on three main areas of algorithmic use: recruitment, workplace decisions and performance management. It outlines the potential risks associated with algorithmic management, including issues of privacy, data security, bias, and the erosion of worker autonomy. Additionally, it explains how unions around the world have pushed back against these challenges.

UNI Global Union published a new edition of the guide in 2023 focusing on performance management, an emerging topic in the global trade union movement. It offers practical strategies for trade unions that can be used in collective bargaining negotiations. The Guide outlines strategies for collective resistance based on already existing protections with examples from different countries, including data protection laws, fair work standards, health and safety laws, and obligations to bargain over algorithms. Through collective bargaining, unions can negotiate terms that protect workers’ rights in the face of automated decision making. This includes ensuring that workers have access to the data used for evaluating their performance and ability to contest decisions made my algorithms. The guide critically evaluates certain algorithmic management systems used by companies such as performance algorithms.

The Guide concludes with ten key negotiating demands for trade unions to use during collective bargaining over the question of algorithmic management. These demands focus on notice and assessment periods, decision making, the right to know, discrimination, discipline, health and safety, data collection and access, monitoring and surveillance, sharing benefits, and training.