OIT - Global case studies of social dialogue on AI and algorithmic management

X Global case studies of social dialogue on AI and algorithmic management Authors / Virginia Doellgast, Shruti Appalla, Dina Ginzburg, Jeonghun Kim, Wen Li Thian

July / 2025 ILO Working Paper 144© International Labour Organization 2025 Attribution 4.0 International (CC BY 4.0) This work is licensed under the Creative Commons Attribution 4.0 International. See: https:// creativecommons.org/licenses/by/4.0/. The user is allowed to reuse, share (copy and redistribute), adapt (remix, transform and build upon the original work) as detailed in the licence. The user must clearly credit the ILO as the source of the material and indicate if changes were made to the original content. Use of the emblem, name and logo of the ILO is not permitted in connection with translations, adaptations or other derivative works. Attribution – The user must indicate if c hanges w ere made and must cite the wor k as follows: Doellgast, V., Appalla, S., Ginzburg, D., Kim, J., Thian, WL. Global case studies of social dialogue on AI and algorithmic management. ILO Working Paper 144. Geneva: International Labour Office, 2025.© ILO. Translations – In case of a translation of this work, the following disclaimer must be added along with the attribution: This is a translation of a copyrighted work of the International Labour Organization (ILO). This translation has not been prepared, reviewed or endorsed by the ILO and should not be considered an official ILO translation. The ILO disclaims all responsibility for its content and accuracy. Responsibility rests solely with the author(s) of the translation. Adaptations – In case of an adaptation of this work, the following disclaimer must be added along with the attribution: This is an adaptation of a copyrighted work of the International Labour Organization (ILO). This adaptation has not been prepared, reviewed or endorsed by the ILO and should not be considered an official ILO adaptation. The ILO disclaims all responsibility for its content and acalong with the attribution: This is an adaptation of a copyrighted work of the International Labour Organization (ILO). This adaptation has not been prepared, reviewed or endorsed by the ILO and should not be considered an official ILO adaptation. The ILO disclaims all responsibility for its content and accuracy. Responsibility rests solely with the author(s) of the adaptation. Third-party materials – This Creative Commons licence does not apply to non-ILO copyright materials included in this publication. If the material is attributed to a third party, the user of such material is solely responsible for clearing the rights with the rights holder and for any claims of infringement. Any dispute arising under this licence that cannot be settled amicably shall be referred to arbitration in accordance with the Arbitration Rules of the United Nations Commission on International Trade Law (UNCITRAL). The parties shall be bound by any arbitration award rendered as a result of such arbitration as the final adjudication of such a dispute. For details on rights and licensing, contact: rights@ilo.org. For details on ILO publications and digital products, visit: www.ilo.org/publns. ISBN 9789220421673 (print), ISBN 9789220421680 (web PDF), ISBN 9789220421697 (epub), ISBN 9789220421703 (html). ISSN 2708-3438 (print), ISSN 2708-3446 (digital) https://doi.org/10.54394/VOQE4924 The designations employed in ILO publications, which are in conformity with United Nations practice, and the presentation of material therein do not imply the expression of any opinionwhatsoever on the part of the ILO concerning the legal status of any country, area or territory or of its authorities, or concerning the delimitation of its frontiers or boundaries. See: www.ilo. org/disclaimer. The opinions and views expressed in this publication are those of the author(s) and do not necessarily reflect the opinions, views or policies of the ILO. Reference to names of firms and commercial products and processes does not imply their enorg/disclaimer. The opinions and views expressed in this publication are those of the author(s) and do not necessarily reflect the opinions, views or policies of the ILO. Reference to names of firms and commercial products and processes does not imply their endorsement by the ILO, and any failure to mention a particular firm, commercial product or process is not a sign of disapproval. Information on ILO publications and digital products can be found at: www.ilo.org/researchand-publications ILO Working Papers summarize the results of ILO research in progress, and seek to stimulate discussion of a range of issues related to the world of work. Comments on this ILO Working Paper are welcome and can be sent to research@ilo.org.

Authorization for publication: Caroline Fredrickson, Director, Research Department ILO Working Papers can be found at: www.ilo.org/research-and-publications/working-papers Suggested citation: Doellgast, V., Appalla, S., Ginzburg, D., Kim, J., Thian, WL. 2025. Global case studies of social dialogue on AI and algorithmic management, ILO Working Paper 144 (Geneva, ILO). https:// doi. org/10.54394/VOQE492401 IL O Working Paper 144

Abstract Employers are adopting and refining artificial intelligence (AI) and algorithm-based tools in the workplace, with wide-ranging implications for work and employment. This working paper examines case studies of social dialogue on AI at national, regional, sectoral, company, and workplace levels in Europe, North America, Asia, South America and the Caribbean, and Africa. Findings are organized around three distinct ‘action fields’ in which worker representatives have sought to influence strategies and outcomes associated with the growing use of AI and algorithms in the workplace. These include the employment and skill impacts of AI, algorithmic management practices, and working conditions and rights in AI value chains. Across these action fields, social dialogue is playing a crucial role in encouraging an alternative, high road approach to AI investments and uses, based on complementing rather than replacing worker skills, empowering rathpractices, and working conditions and rights in AI value chains. Across these action fields, social dialogue is playing a crucial role in encouraging an alternative, high road approach to AI investments and uses, based on complementing rather than replacing worker skills, empowering rather than controlling the workforce, and embedding rather than displacing new jobs in labor and social protections. Comparative findings suggest that these social dialogue initiatives are more effective where there are constraints on employer exit, support for collective worker voice, and strategies of inclusive labor solidarity.

Key words: AI, algorithms, algorithmic management, social dialogue, labor unions, skills, job quality, surveillance, global value chains, outsourcing About the authors Virginia Doellgast is the Anne Evans Estabrook Professor of Employment Relations and Dispute Resolution in the ILR School at Cornell University. She is President of the Society for the Advancement of Socio-Economics (SASE) 2024-25, a Senior Research Fellow at the at the WSIHans Böckler Stiftung, and Co-Editor of the ILR Review. Her research focuses on the comparative political economy of labor markets and labor unions, inequality, precarity, and democracy at work. Publications include Exit, Voice, and Solidarity (Oxford University Press, 2022), Disintegrating Democracy at Work (Cornell University Press, 2012), International and Comparative Employment Relations (Sage, 2021), and Reconstructing Solidarity (Oxford University Press, 2018).

Shruti Appalla is a PhD student in the ILR School at Cornell University. She studies the impact of emerging technologies on workers and firms in global supply chains. She has previously been a Predoctoral Fellow in Economics and holds a master’s degree in public policy from National Law School, India. Dina Ginzburg is an MS Labor Research and Policy student in the ILR School at Cornell University. She studies labor policy, union strategy, and collective action. Jeonghun Kim is a PhD student in the ILR School at Cornell University. His research explores how precarity is generated through new work arrangements and technologies, and how workers build

She studies labor policy, union strategy, and collective action. Jeonghun Kim is a PhD student in the ILR School at Cornell University. His research explores how precarity is generated through new work arrangements and technologies, and how workers build solidarity to regulate them. He has conducted a comparative study of two unions’ organizing strategies in the South Korean food delivery platform sector. For his dissertation, he is working on a project that compares call center unions in the public health insurance sector in South Korea and the United States, focusing on how they respond to the challenges posed by AI adoption.02 ILO Working Paper 144 Wen Li Thian is a PhD student in the ILR School at Cornell University. Her research focuses on mechanisms of labor control, the labor process, and technologies at work. She has done research on the lived experiences of factory workers and platform gig workers in Singapore. She holds a master’s degree in Sociology from the National University of Singapore.03 ILO Working Paper 144 Abstract 01 About the authors 01 X Introduction 07 Analytical framework 08 Case selection and research approach 11 X 1 International, national, and regional social dialogue 13 1.1. Europe 13 1.1.1. Social dialogue at EU-level 13 1.1.2. Social dialogue at national level 16 1.2. North America 19 1.3. Asia 23 1.4. South America and the Caribbean 25 1.5. Africa 27 1.6. Summary 27 X 2 Sectoral, company, and workplace social dialogue 30 2.1. Social dialogue over employment and skill impacts of AI: from labor replacing to labor complementing 30 2.1.1. Europe 31 2.1.2. North America 35 2.1.3. Asia, South America, and Africa 41 2.2. Social dialogue over algorithmic management: From labor controlling to labor empowering 44 2.2.1. Europe 45 2.2.2. North America 50

2.1.2. North America 35 2.1.3. Asia, South America, and Africa 41 2.2. Social dialogue over algorithmic management: From labor controlling to labor empowering 44 2.2.1. Europe 45 2.2.2. North America 50 2.2.3. Asia, South America, the Caribbean, and Africa 53 2.3. Social dialogue over working conditions and rights in AI-enabled fissuring: From labor displacing to labor embedding 56 2.3.1. Re-embedding the AI value chain 57 2.4. Summary 63 X Conclusion 65 Annex 1. List of interviews and email communication 68 References 70 Table of contents04 ILO Working Paper 144 Acknowledgements 8605 ILO Working Paper 144 List of Figures Figure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity 10 Figure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity 6506 ILO Working Paper 144 List of Boxes Case study 1: SAG-AFTRA – US actors mobilize to establish AI guidelines in film, television, and game development 36 Case study 2: Negotiating ‘labor empowering’ agreements in the German ICTS industry: Deutsche Telekom and IBM 46 Case study 3: Regulating algorithmic management in Las Vegas casinos and hotels: the Culinary Union 51 Case study 4: Challenging facial recognition technologies: HD Hyundai Heavy Industries versus various unions in South Korea 54 Case study 5: Embedding AI labor in Africa: the Kenyan Content Moderators Union 6007 ILO Working Paper 144 X Introduction Employers worldwide are adopting or expanding their use of artificial intelligence (AI) and algorithm-based tools in the workplace, with transformative impacts on work and employment.1 Labor unions and other worker representatives are responding to both the opportunities and threats of these new technologies with strategies that seek to establish more fair and productive

X Introduction Employers worldwide are adopting or expanding their use of artificial intelligence (AI) and algorithm-based tools in the workplace, with transformative impacts on work and employment.1 Labor unions and other worker representatives are responding to both the opportunities and threats of these new technologies with strategies that seek to establish more fair and productive uses of these tools. In this report, we examine case studies of worker representative involvement in social dialogue over AI and algorithms in a range of countries and industries, with the goal of both documenting these cases and learning from them.2 What models of or experiments with social dialogue are developing in different contexts? What are their goals? How do they seek to pursue these goals, for example, through increasing workers’ ability to have a say in adoption and deployment decisions, securing good jobs with decent pay and conditions, or preventing intensified surveillance and insecurity? And where have they shown the most success in meeting these goals: what factors explain those successes? Our analysis is organized around three distinct areas in which AI affects workers and working conditions, which we argue constitute different ‘action fields’ for labor unions and other worker representatives. These include social dialogue over: 1) the employment and skill impacts of AI – with the goal of encouraging employers to move from replacing jobs and tasks with technology to complementing worker skills, 2) algorithmic management practices - with the goal of encouraging employers to use new management tools to empower rather than intensify worker control and biased decision-making; and 3) working conditions and rights in AI-enabled fissuring – with the goal of encouraging a shift from displacing jobs from social and labor market protections to embedding them in encompassing collective and social standards.3 In each area, the tools available differ, as do the main outcomes various stakeholders care about, affecting opportunities for mobilization or coalition building.

First, new AI-based tools are often feared to automate jobs and tasks, in a way that leads to downsizing or ‘replacement’ of workers and worker skills. There may also be a deskilling effect, essentially replacing skilled workers and allowing the downgrading of jobs to more repetitive or lower valued tasks. Alternatively, these tools can complement workers’ existing skills or help them to develop new skills and modes of working - augmenting rather than replacing workers. We ask how worker representatives have sought to influence these potential impacts of AI on employment and skills, and their success in encouraging ‘labor complementing’ applications through social dialogue. Second, new AI-based tools are used in automated algorithmic management systems to manage workers’ performance via monitoring, coaching, and decision-making around dismissals, incentive pay, and promotions; as well as to automate decisions related to hiring, scheduling, and training. This can intensify discipline, reduce workers’ individual control over the pace and content of their work, and increase worker burnout. These effects are often experienced unequally 1 ‘Artificial intelligence’ refers to computer systems that can perform tasks traditionally requiring human intelligence, including advanced pattern recognition and problem-solving. ‘Generative AI’ tools such as ChatGPT are a class of machine learning technologies that generate new content based on this analysis of patterns in data. 2 We use the ILO’s definition of social dialogue as including 'all types of negotiation, consultation or simply exchange of information between, or among, representatives of governments, employers and workers, on issues of common interest relating to economic and social policy. It can exist as a tripartite process, with the government as an official party to the dialogue or it may consist of bipartite relations only between labour and management (or trade unions and employers' organizations). Workplace cooperation, collective bargaining at company, sector or cross-industry levels, and tripartite consultation processes are common forms of social dialogue.'

ILO (2019) Social Dialogue. https://www.ilo.org/resource/social-dialogue-0 3 Doellgast, V. (2023). Strengthening social regulation in the digital economy: comparative findings from the ICT industry. Labour and Industry, 33(1), 22-38.08 ILO Working Paper 144 by different employee groups, where AI-based tools reproduce biased decision-making embedded in training data or model design. At the same time, AI-based technologies may be a tool for empowering workers, or giving them more control over working methods and practices. These ‘labor empowering’ uses of AI are more likely where workers have voice in and oversight over how digital and AI-based tools are used to organize, for example, schedules and training; and where negotiations limit invasive forms of monitoring and performance evaluation. We ask how worker representatives have sought to influence these potential impacts of AI on worker control and bias through social dialogue, and the role of this dialogue in encouraging ‘labor empowerment’ via strengthened worker voice in management decision-making. Third, AI-based tools open opportunities for or encourage experimentation with new location decisions and organizational models. Employers used new technologies associated with the first wave of digitalization to consolidate, outsource, and offshore a range of service and manufacturing jobs in the 1990s to 2000s. The combination of cloud computing, faster data speeds, and algorithm-enabled management tools are now expanding location options and permitting a new set of strategies to restructure jobs, which we describe as AI-enabled fissuring. One important trend involves developments in the ‘AI value chain’ to produce and refine AI-based technologies - including the growing number of energy-hungry data centers. New jobs are being created and restructured in data coding, labelling, maintenance, and engineering. These jobs are organized through complex global production networks of lead firms, suppliers, and platforms, with power significantly concentrated at major tech companies. Workers performing the most labor-intensive ‘data janitorial’ and content moderation work are often low paid and subject to intensive algorithmic control. Key questions that these trends raise include how worker representatives are rethrough complex global production networks of lead firms, suppliers, and platforms, with power significantly concentrated at major tech companies. Workers performing the most labor-intensive ‘data janitorial’ and content moderation work are often low paid and subject to intensive algorithmic control. Key questions that these trends raise include how worker representatives are responding to the labor displacing impacts of AI-based restructuring - and their success in ‘embedding’ these new and restructured jobs within collective agreements and national systems of social protection. This report examines case study examples of social dialogue in these three ‘action fields’, across Europe, North America, Asia, South America and the Caribbean, and Africa. Findings draw on stakeholder interviews, archival sources, academic research, and news reports. In the following sections, we first summarize our analytical framework (Section 1.1) and then discuss our approach to identifying and analyzing our case study examples (Section 1.2). In Section 2, we examine international, national, and regional cases of social dialogue. We then compare company-, workplace-, and industry-level examples of social dialogue between employers and unions (Section 3), organized by the three themes discussed above: labor replacing to complementing (Section 3.1), labor controlling to empowering (Section 3.2), and labor displacing to embedding (Section 3.3.). We conclude with a discussion of the conditions for social dialogue to promote a more equitable and just transition to an AI-enabled digital economy (Section 4). Analytical framework Past research suggests that social dialogue will be most effective in establishing a socially equitable approach to new technology investments - including in AIand algorithm-based toolswhere it helps to move employers toward strategies and practices that take a longer-term view concerning the goals of these investments. This means making commitments to creating good jobs with benefits and security, sharing productivity gains with workers, investing in skills and worker discretion, limiting invasive data collection and monitoring, and establishing fair and transparent opportunities for workers to challenge and change technology-enabled decisions. Where labor unions and other worker representatives have engaged in social dialogue over

jobs with benefits and security, sharing productivity gains with workers, investing in skills and worker discretion, limiting invasive data collection and monitoring, and establishing fair and transparent opportunities for workers to challenge and change technology-enabled decisions. Where labor unions and other worker representatives have engaged in social dialogue over AI, they have sought to institutionalize these goals in collective agreements, laws, and policies.09 ILO Working Paper 144 The social and worker impacts of technological change have been a focus of worker mobilization, consultation, and collective bargaining for over a hundred years.4 The pace of this change has increased over the past five decades. From lean production and the introduction of micro-computing in the 1980s and 1990s to the first wave of digitalization and widespread adoption of industrial robots and CNC machines in the 1990s and 2000s, labor unions have sought to encourage alternative, high road approaches to integrating new technologies into the workplace.5 And they have faced the same challenges, as many employers adopted these technologies in ways that cut costs or centralized control through automating and deskilling work, intensifying and individualizing performance monitoring, and ‘displacing’ work and workers through temp agency contracting, outsourcing, and offshoring - or workplace fissuring.6 At the same time, alternative high road approaches were developed by employers who sought to compete based on longer-term investments in skills and high-quality products and services, rather than on short-term cost savings. Comparative research suggests that these alternatives were most consistently pursued where legislated minimum standards and collective bargaining institutions placed ‘productive constraints’ on employers - effectively closing off low road strategies, while supporting investments in skills and participation.7 The case studies discussed in this report can be read as different efforts across world regions and countries to establish a new set of ‘productive constraints’ that make it more difficult to pursue what Acemoglu and Restrepo describe as the ‘wrong kind of AI’ - focused on short-term cost savings, with risks of stagnating employment and productivity.8 Under what conditions have these efforts been most successful? We apply the framework in

sue what Acemoglu and Restrepo describe as the ‘wrong kind of AI’ - focused on short-term cost savings, with risks of stagnating employment and productivity.8 Under what conditions have these efforts been most successful? We apply the framework in Doellgast’s book Exit, Voice, and Solidarity to consider three factors or conditions that play an important role in supporting more effective social dialogue on AI: constraints on employer exit, support for collective worker voice, and strategies of inclusive solidarity.9 Constraints on employer exit. The first condition is negotiated, legal, or skill-based restrictions on employers’ willingness and ability to exit internal employment relationships. This can include minimum employment standards and social protections at national level, within industries, in companies, or along firms’ supply chains or supplier networks. It can also involve vocational skill and training systems or ecosystems of firms that tie investment to a particular organization and location, or that increase the perceived value of worker skills as a complement to new technology investments. 4 Knotter, A. (2018). Transformations of trade unionism: comparative and transnational perspectives on workers organizing in Europe and the United States. Amsterdam University Press. Ross, P . (1970). Waterfront labor response to technological change: a tale of two unions. Labor Law Journal, 21(7), 397. 5 Berggren, C. (1993). Alternatives to lean production: Work organization in the Swedish auto industry. Ithaca: Cornell University Press. Kochan, T. A., Osterman, P . (1994). The Mutual Gains Enterprise. Cambridge: Harvard Business School Press. Dore, R. P . (2000). Stock market capitalism: Welfare capitalism: Japan and Germany versus the Anglo-Saxons. Oxford: Oxford University Press. Freeman, R. B.,

Shaw, K. L. (2009). International differences in the business practices and productivity of firms. Chicago: University of Chicago Press. 6 Doellgast, V. (2012). Disintegrating democracy at work: Labor unions and the future of good jobs in the service economy. Cornell University Press. Davis, G. F. (2016). The Vanishing American Corporation: Navigating the hazards of a new economy. Berrett-Koehler Publishers. Goldman, D. J. (2024). Disconnected: Call Center Workers Fight for Good Jobs in the Digital Age. University of Illinois Press. 7 Streeck, W. (1992). Productive constraints: on the institutional conditions of diversified quality production. Social institutions and economic performance: Studies of industrial relations in advanced capitalist economies, 1-40 Hall, P . A., Soskice, D. (eds.). (2001). Varieties of Capitalism: The Institutional Foundations of Comparative Advantage. Oxford: Oxford University Press. 8 Acemoglu, D. & P . Restrepo (2019) ‘The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand’ NBER Working Paper 25682, March 2019, https://www.nber.org/papers/w25682. 9 Doellgast, V. (2022). Exit, voice, and solidarity: Contesting precarity in the US and European telecommunications industries. Oxford University Press10 ILO Working Paper 144 Support for collective worker voice. The second condition is institutions and resources that support collective voice, or the ability of workers to have a say in and influence over practices and decisions in their workplace through democratic or representative structures. These include bargaining or participation rights, which encourage negotiation and consultation over management decisions at workplace, company, or industry level; worker rights to representation on company boards; or state support for tripartite social dialogue and social pacts over social, labor market, and training policies. In the context of AI, strong data protection rights can support worker voice, through providing additional transparency regarding management practices. Workers and their unions can also build strong support for collective voice at company and industry level - and inand training policies. In the context of AI, strong data protection rights can support worker voice, through providing additional transparency regarding management practices. Workers and their unions can also build strong support for collective voice at company and industry level - and institutional voice rights in their collective agreements - through organizing and mobilizing the workforce to increase countervailing power in consultations or negotiations. Strategies of inclusive solidarity. The third condition is labor strategies based on inclusive forms of solidarity, which seek to bridge divides in the labor movement and across the workforce. Solidarity is undermined by competition for jobs or investment across groups of workers, within countries or internationally. It is also weakened by historic divides within the labor movement, based on ideology, racism, or narrow construction of interests. Labor strategies that bridge these divides are needed to extend power from core, protected workers to more precarious ones across fissured supply chains and countries with stronger and weaker labor laws and protections. Many of the social dialogue examples we discuss in this report can be seen as attempts to establish or strengthen constraints on employer exit and support for collective worker voice, while deploying more inclusive strategies of solidarity, in the context of technology-enabled restructuring pressures. Figure 1 illustrates how this framework relates to the three ‘action fields’ we focus on in this report. These include social dialogue over: 1) the employment and skill impacts of AI – from labor replacing to complementing, 2) algorithmic management practices - from labor controlling to empowering, and 3) working conditions and rights in AI-enabled fissuring – from labor displacing to embedding. X Figure 1: Supporting social dialogue on AI through constraints on exit, support for voice, and strategies of solidarity This framework suggests that while the three action fields can involve all three factors (exit, voice, and solidarity), each one relies most heavily on two of the three.11 ILO Working Paper 144 First, social dialogue encouraging a shift from labor replacing to labor complementing uses of AI typically is most effective where unions are able to draw on existing institutions that constrain employer exit from employment contracts and social protections, such as job security agreeand solidarity), each one relies most heavily on two of the three.11 ILO Working Paper 144 First, social dialogue encouraging a shift from labor replacing to labor complementing uses of AI typically is most effective where unions are able to draw on existing institutio