Publications and other outputs
This page consolidates five years' research by the iManage team, led by Professor Jeremias Adams-Prassl. You can read, watch, and listen to our research below. Many of the publications are available in an open-access special issue of the European Labour Law Journal. Our research has caught the attention of media outlets such as Wired and has been shared in forums such as the Institute for the Future of Work, LSE Law School, and the Institute for the Ethics of AI.
Publications
By Jeremias Adams-Prassl, Halefom Abraha, Aislinn Kelly-Lyth, Michael ‘Six’ Silberman, Sangh Rakshita
European Labour Law Journal
The promise—and perils—of algorithmic management are increasingly recognised in the literature. How should regulators respond to the automation of the full range of traditional employer functions, from hiring workers through to firing them? This article identifies two key regulatory gaps—an exacerbation of privacy harms and information asymmetries, and a loss of human agency—and sets out a series of policy options designed to address these novel harms. Redlines (prohibitions), purpose limitations, and individual as well as collective information rights are designed to protect against harmfully invasive data practices; provisions for human involvement ‘in the loop’ (banning fully automated terminations), ‘after the loop’ (a right to meaningful review), ‘before the loop’ (information and consultation rights) and ‘above the loop’ (impact assessments) aim to restore human agency in the deployment and governance of algorithmic management systems.
By Jeremias Adams-Prassl
European Labour Law Journal
This article scrutinises the potential of the existing regulatory apparatus in Union law to tackle the social, technical, and legal challenges inherent in deploying automated systems in high-risk settings such as the workplace, with a view to setting out key lessons for the proposed EU Artificial Intelligence Act. Surveying data protection and discrimination rules as well as the social acquis, it highlights key areas for further development, from coherence between different regulatory regimes to the role of social partnership in shaping key standards and monitoring their implementation.
By Jeremias Adams-Prassl, Reuben Binns, and Aislinn Kelly-Lyth
Modern Law Review
Discriminatory bias in algorithmic systems is widely documented. How should the law respond? A broad consensus suggests approaching the issue principally through the lens of indirect discrimination, focusing on algorithmic systems’ impact. In this article, we set out to challenge this analysis, arguing that while indirect discrimination law has an important role to play, a narrow focus on this regime in the context of machine learning algorithms is both normatively undesirable and legally flawed. We illustrate how certain forms of algorithmic bias in frequently deployed algorithms might constitute direct discrimination, and explore the ramifications—both in practical terms, and the broader challenges automated decision-making systems pose to the conceptual apparatus of anti-discrimination law.
By Aislinn Kelly-Lyth
European Labour Law Journal
The potential for algorithms to discriminate is now well-documented, and algorithmic management tools are no exception. Scholars have been quick to point to gaps in the equality law framework, but existing European law is remarkably robust. Where gaps do exist, they largely predate algorithmic decision-making. Careful judicial reasoning can resolve what appear to be novel legal issues; and policymakers should seek to reinforce European equality law, rather than reform it. This article disentangles some of the knottiest questions on the application of the prohibition on direct and indirect discrimination to algorithmic management, from how the law should deal with arguments that algorithms are ‘more accurate’ or ‘less biased’ than human decision-makers, to the attribution of liability in the employment context. By identifying possible routes for judicial resolution, the article demonstrates the adaptable nature of existing legal obligations.
By Halefom Abraha
European Labour Law Journal
The regulation of algorithmic management falls under the purview of multiple legal domains including but not limited to labour law, nondiscrimination law and data protection law. While labour law does not have explicit provisions to adequately protect workers from algorithmic harms, existing non-discrimination and data protection laws can address some aspects of these harms. This article examines the extent to which the GDPR offers the necessary tools to protect workers from harm stemming from algorithmic management. It argues that while the provisions tailored to automated decision-making (ADM) and the rest of the GDPR provide workers with some limited protections, significant gaps remain. It then suggests some policy options on how the existing protections under the GDPR can be further complemented, particularised, and strengthened through a combination of legislative and non-legislative measures.
By Aislinn Kelly-Lyth and Anna Thomas
European Labour Law Journal
Algorithmic outputs are increasingly shaping the employee experience, presenting a host of risks and impacts with far-reaching consequences. This contribution considers how algorithmic impact assessments should complement, as well as inform, an overarching ‘top-down’ framework for the governance of algorithmic management systems. The criteria for an effective impact assessment obligation in the algorithmic management context are identified, including the appropriate stages, actors, and procedure. The Good Work Charter, which operates as a synthesis of legal principles, rights, and obligations, as well as ethical principles as they apply to the workplace, is proposed as an assessment framework. Finally, the article compares the proposed model with the existing obligation to carry out data protection impact assessments for high-risk data processing.
By Michael Veale, Michael ‘Six’ Silberman, Reuben Binns
European Labour Law Journal
The European Commission proposed a Directive on Platform Work at the end of 2021. While much attention has been placed on its effort to address misclassification of the employed as self-employed, it also contains ambitious provisions for the regulation of the algorithmic management prevalent on these platforms. Overall, these provisions are well-drafted, yet they require extra scrutiny in light of the fierce lobbying and resistance they will likely encounter in the legislative process, in implementation and in enforcement. In this article, we place the proposal in its sociotechnical context, drawing upon wide cross-disciplinary scholarship to identify a range of tensions, potential misinterpretations, and perversions that should be pre-empted and guarded against at the earliest possible stage.
By Halefom Abraha
Modern Law Review
On 30 March 2023, the European Court of Justice ruled for the first time on the interpreta-tion of Article 88 GDPR, which gives Member States the power to provide for more specificrules on employee data processing. In response to a request from the Administrative Court ofWiesbaden (Hauptpersonalrat der Lehrerinnen) concerning the live streaming of video classes, theCJEU found the German law regulating personal data processing in the employment contextto be incompatible with the GDPR. This note examines the far-reaching implications of theruling, not least given that similar provisions have been enacted at the state and federal levels inGermany as well as in several other Member States. It further identifies guidelines for the appro-priate use of Article 88. Notably, national laws making use of Article 88 GDPR must providenormative content that is distinct from, but compatible with, the GDPR.
By Jeremias Adams-Prassl
Industrial Law Journal
Uber drivers, the UK Supreme Court ruled in early 2021, are workers—and they are working whenever they are ‘“on duty” logged onto the Uber app in London available to accept a trip request’. Uber, a global transportation services provider founded just over a decade ago, is one of the archetypical platforms at the heart of the gig economy, where digital labour intermediaries use sophisticated platforms to provide consumer services and control large, on-demand workforces. One core element of the vast majority of platforms’ business model is the claim that workers are merely independent contractors, falling outside the scope of employment law. While the rise of the gig economy has led to extensive scrutiny in Parliament, Government and the Employment Tribunals, the Uber drivers’ claim marks the first time that the employment status of digitally intermediated workers has come before the UK Supreme Court. This note sets out the central facts of the case and recaps the decision’s litigation history, spanning a period of over five years, before focussing on the Supreme Court’s unanimous decision. The conclusion then turns to Uber’s broader implications in two dimensions: doctrinally, in terms of the (eventual) scope of the Court’s clear commitment to purposive statutory interpretation; practically, in terms of the decision’s aftermath for Uber’s London-based drivers and the limitations of employment law enforcement
By Halefom Abraha
International Data Privacy Law
The distinct challenges of data processing at work have led to long-standing calls for sector specific regulation. This leaves the European legislature with a dilemma. While the distinct features of employee data processing give rise to novel issues that cannot adequately be addressed by an omnibus data protection regime, a combination of legal, political, and constitutional factors have hindered efforts towards adopting harmonized employment-specific legislation at the EU level. The ‘opening clause’ in Article 88 General Data Protection Regulation (GDPR) aims to square this circle. It aims to ensure adequate and consistent protection of employees while also promoting regulatory diversity, respecting national peculiarities, and protecting Member State autonomy. This article examines whether the opening clause has delivered on its promises. It argues that while the compromise has delivered on some of its promises in promoting diverse and innovative regulatory approaches, it also runs counter to the fundamental objectives of the GDPR itself by creating further fragmentation, legal uncertainty, and inconsistent implementation, interpretation, and enforcement of data protection rules.
By Aislinn Kelly-Lyth
Comparative Labour Law & Policy Journal
In April 2018, just before the GDPR came into force, the European Commission adopted a Communication on Artificial Intelligence (AI). In it, the Commission declared that the EU could “lead the way in developing and using AI for good and for all.” Three years later, the Commission published a proposal for a Regulation laying down rules on artificial intelligence (the proposed ‘AI Act’). The proposal is now progressing through the early stages of legislative consideration. This article outlines the relevance of the Act for the employment context, including its objects, scope, and possible impacts. It also highlights some key areas of concern: the lack of external supervision over compliance assessments; the inadequacy of the proposal’s transparency provisions; and the potential deregulatory effects of the Act at the domestic level.
By Six Silberman
LIMITS 2023 (Ninth Workshop on Computing within Limits)
This paper builds on the author’s 2015 paper ‘Information systems for the age of consequences.’ This paper reviews practical, policy, and research developments in the period 2015–2023. Building on the view that bottlenecks in addressing environmental problems are more ‘institutional’ than ‘technical,’ but that technology development can be understood as ‘politics by other means,’ the paper makes two main contributions to LIMITS. First, it offers a set of propositions about technology and institutions, from literatures including organizational psychology and software engineering, to help orient ‘limits-aware’ information technology practice. Second, it offers six pieces of advice. The paper is geared especially to students who will pursue technology careers outside academia.
By Reuben Binns, Jeremias Adams-Prassl, and Aislinn Kelly-Lyth
FAccT '23: Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency
Previous literature on ‘fair’ machine learning has appealed to legalframeworks of discrimination law to motivate a variety of discrimi-nation and fairness metrics and de-biasing measures. Such worktypically applies the US doctrine of disparate impact rather thanthe alternative of disparate treatment, and scholars of EU law havelargely followed along similar lines, addressing algorithmic bias asa form of indirect rather than direct discrimination. In recent work,we have argued that such focus is unduly narrow in the contextof European law: certain forms of algorithmic bias will constitutedirect discrimination [1]. In this paper, we explore the ramificationsof this argument for existing taxonomies of machine bias and al-gorithmic fairness, how existing fairness metrics might need to beadapted, and potentially new measures may need to be introduced.We outline how the mappings between fairness measures and dis-crimination definitions implied hitherto may need to be revisedand revisited.
By Hilde Weerts, Aislinn Kelly-Lyth, Reuben Binns, and Jeremias Adams-Prassl
FAccT '24: Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency
Emerging scholarship suggests that the EU legal concept of direct discrimination - where a person is given different treatment on grounds of a protected characteristic - may apply to various algorithmic decision-making contexts. This has important implications: unlike indirect discrimination, there is generally no ‘objective justification’ stage in the direct discrimination framework, which means that the deployment of directly discriminatory algorithms will usually be unlawful per se. In this paper, we focus on the most likely candidate for direct discrimination in the algorithmic context, termed inherent direct discrimination, where a proxy is inextricably linked to a protected characteristic. We draw on computer science literature to suggest that, in the algorithmic context, ‘treatment on the grounds of’ needs to be understood in terms of two steps:proxy capacity and proxy use. Only where both elements can be made out can direct discrimination be said to be ‘on grounds of’a protected characteristic. We analyse the legal conditions of our proposed proxy capacity and proxy use tests. Based on this analysis, we discuss technical approaches and metrics that could be developed or applied to identify inherent direct discrimination in algorithmic decision-making.
News articles
By Jeremias Adams-Prassl, Aislinn Kelly-Lyth, Six Silberman
Wired
Our researchers comment on the then-expected implementation of the Platform Work Directive and its influence on algorithmic management. The piece comments on the limitations of the legislation and advocates for a broader Directive on algorithmic management, which tackles privacy violations, information asymmetries, and the loss of managerial agency. The article introduces the policy options for regulating algorithmic management, which include prohibitions, requirements, rights and protections.
Blog posts
By Maria Ebenhöh, Anna-Elisabeth Hampel, Franziska Loschert, Six Silberman, Jeremias Adams-Prassl, Johannes Höller
Wolters Kluwer - Global Workplace Law & Policy
This blog post briefly examines one—little-discussed—piece of the Platform Work Directive: Article 20, which stipulates that platforms make available ‘[c]ommunication channels for persons performing platform work’. This creates a right for platform workers to communicate with one another—and their representatives—without surveillance. The blog post draws attention to four issues about how channels will be governed effectively with respect to (i) security and trustworthiness, (ii) content moderation, (iii) worker representatives, and (iv) linguistic and national differences.
By Jeremias Adams-Prassl and Aislinn Kelly-Lyth
Verfassungsblog
On 8 December 2021, the European Commission published its long-awaited draft of a Directive aimed at improving working conditions in the platform (or ‘gig’) economy. In response to the explosive growth of platform-mediated work, not least as a result of the Covid-19 pandemic, the proposal aims to ensure, first, ‘that people working through platforms have – or can obtain – the correct employment status’; second, ‘to ensure fairness, transparency and accountability in algorithmic management in the platform work context’; and third, ‘to enhance transparency, traceability and awareness of developments in platform work and improve enforcement of […] applicable rules’. In this post, we scrutinise key elements of the proposed Directive against these goals. Our tentative conclusion is positive: while there is some room for improvement during the legislative process, the framework laid down promises to tackle some of the most salient problems arising from platform work.
By Six Silberman
Verfassungsblog
On 9 December 2021, the European Commission announced its proposal for a Directive on improving working conditions in platform work—the ‘Platform Work Directive.’ The Directive’s main goals are to reduce false self-employment among persons performing platform work, to regulate algorithmic management on digital labour platforms, and to provide legal certainty for platforms. The Directive is now in interinstitutional negotiations (‘trilogues’). Others have observed that the Council’s position has weakened the Directive significantly (see the upcoming blogposts by Kocher and Barrio). This essay, however, focuses on an element of the proposed Directive that has gone relatively unremarked in the scholarly and policy debates so far: the definition of ‘digital labour platform.’ It appears from the preparatory materials—and from the press release announcing the proposal—that, in drafting it, the Commission mainly had in mind the most visible and widely-discussed type of platform work: on-location work, such as transportation and delivery. And it seems clear that the Directive’s definition includes platforms mediating this type of work. It is however not clear whether other important types of platform work are also included.
By Halefom Abraha, Jeremias Adams-Prassl, Aislinn Kelly-Lyth
Oxford Business Law Blog
The European Commission’s draft Directive on working conditions in the platform (or ‘gig’) economy is a promising instrument. It addresses key risks facing gig workers, including employment status misclassification, algorithmic management, and the enforcement of existing rules. The Directive’s focus on harms which stem from algorithmic management, from automatic pay-docking and shift scheduling to automated termination, is particularly welcome. As the Directive progresses through the legislative process, we suggest targeted amendments to improve the draft provisions on algorithmic management further and ensure comprehensive protection of workers’ fundamental rights.
By Halefom Abraha, Aislinn Kelly-Lyth, and Jeremias Adams-Prassl
Oxford Business Law Blog
Since 2006, the Council of Europe has celebrated 28 January as Data Protection Day—known globally as Privacy Day. Data processing now impacts on many spheres of life. One of these is employment: increases in the scale and ease of data processing have led to a sharp increase in employers’ use of algorithmic management tools, which facilitate hitherto inconceivable worker surveillance and analysis and enable the automation of functions that were once carried out by HR. To help raise awareness of data protection at work, the team at iManage—Halefom Abraha, Aislinn Kelly-Lyth, and Prof Jeremias Adams-Prassl—have brought together ten resources examining this new phenomenon through a data protection lens. As the GDPR enters its fourth year of application, these papers provide a great starting point for stakeholders to understand and address data protection issues around AI systems in the workplace.
By Halefom Abraha, Aislinn Kelly-Lyth, and Jeremias Adams-Prassl
Oxford Business Law Blog
Complex algorithmic management tools are increasingly used to make high-stakes decisions that have traditionally been made by human managers. These tools could significantly change working conditions and social relationships—and pose significant risks to the privacy, human dignity, health, equal treatment, and autonomy of workers. While the regulation of algorithmic management falls under the purview of multiple legal domains, data protection law has been the area of law that is most engaged owing to the vast troves of personal data processing underpinning most algorithmic management tools. As we celebrate the 17th global Data Protection Day, it is important to recognise the key role that data protection legislation plays in taming some of the most intrusive algorithmic management systems and practices. To help raise awareness of data protection at work, the team at iManage have brought together six resources examining this new phenomenon through a data protection lens. As the GDPR enters its fifth year of application, these papers provide a great starting point for stakeholders to understand and address data protection issues around AI systems in the workplace.
By Six Silberman, Halefom Abraha, Sangh Rakshita, and Jeremias Adams-Prassl
Oxford Business Law Blog
On 20 September 2023, the Financial Times reported that Uber had ‘warned’ that the proposed EU Platform Work Directive, if adopted in its current form, could ‘force’ the transport company to shut down its ride-hailing service in hundreds of EU cities—and to ‘raise prices by as much as 40 per cent.’ At the same time, the company sought to reassure investors that the proposed law would not impact Uber’s profitability in Europe. ‘This isn’t about Uber’s profits,’ the head of Uber’s European mobility division insisted. What lies behind these surprising—and seemingly contradictory—claims? In this post, we explain the background to the proposed Directive and the ‘rebuttable presumption’ of an employment relationship at its core. In a follow-up post this week, we will turn to an exploration of Uber’s concerns that it could lead to ‘a 50-70 per cent reduction in the number of work opportunities’ available to drivers through its app.
By Six Silberman, Halefom Abraha, Sangh Rakshita, and Jeremias Adams-Prassl
Oxford Business Law Blog
Despite Uber’s remarks to the Financial Times, the proposal would not force platforms to ‘reclassify’ persons performing platform work: it does not change the criteria used to determine employment status. The focus instead is on the determination procedure—namely, changing the ‘default assumption’ to employment rather than self-employment—in order to ensure the effective enforcement of existing rules. Nearly 7 years ago, the London Employment Tribunal found that Uber’s drivers were workers, rather than self-employed contractors. In an unusually pointed judgment (subsequently upheld by the Supreme Court), the Judge was
‘struck by the remarkable lengths to which Uber has gone in order to compel agreement with its (perhaps we should say its lawyers’) description of itself … we cannot help being reminded of Queen Gertrude’s most celebrated line: The lady doth protest too much, methinks.’
In this post, we look at some of the broader implications of employment (re-)classification under the Platform Work Directive, and suggest potential further improvements to an already impressive instrument.
By Six Silberman, Halefom Abraha, Sangh Rakshita, and Jeremias Adams-Prassl
Oxford Business Law Blog
World Data Protection Day, observed annually on January 28, commemorates efforts to raise awareness about the importance of safeguarding personal data in an increasingly digitized world. It serves as a global initiative to promote privacy rights, emphasizing the significance of responsible data processing and protection. Data processing is intimately involved with every aspect of daily life, including at work. The growing ubiquity and complexity of processing techniques, contexts, and scale, particularly the use of algorithmic management—and the increasing granularity and intimacy of the personal data collected—raises serious concerns for workplace environments. This necessitates heightened regulation and proactive measures to deal with violations of privacy as well as other related harms, such as economic risks (eg unfair dismissal and loss of within-job economic opportunities); physical risks (eg injury due to repetitive strain; illness due to stress); and psychosocial risks. Effective regulatory measures can also protect workers, managers and organisations from ‘bugs’ in the software systems they build or buy—the dramatic consequences of which have been most recently on display in the UK in the recent return to the headlines of the tragic Horizon/Post Office scandal. On World Data Protection Day, we bring back our annual selection of resources that deal with issues of data protection and privacy at work.
By Six Silberman, Halefom Abraha, and Jeremias Adams-Prassl
Oxford Human Rights Hub
On 8 December 2021, the European Commission published its long-awaited draft of a directive aiming to improve working conditions in the platform (or ‘gig’) economy. In this blog post, we briefly set out the background to the Directive, explain key elements in the proposal, and provide a few first thoughts.
By Six Silberman, Halefom Abraha, and Jeremias Adams-Prassl
Oxford Human Rights Hub
January 28, 2025 marks the 44th anniversary of the Council of Europe’s Convention 108 (now Convention 108+), which is celebrated globally as Data Protection/Privacy Day. This day commemorates the opening for signature of the Convention in 1981, highlighting its significance in establishing international standards for data protection and privacy rights in the digital age. The Convention has served as the foundation for the development of privacy and data protection legislation worldwide. Currently, approximately 82% of the global population resides in a jurisdiction with established data protection laws or regulations, with numerous additional legislative initiatives under consideration. This proliferation of legislation represents a remarkable evolution in privacy and data protection, warranting celebration. However, the remarkable achievements of the Convention and its progeny, such as the GDPR, must not overshadow what these frameworks have overlooked. Employees, who constitute one of society’s most vulnerable demographics both in terms of their relative position and the possible exploitation of their data in the face of increasing workplace digitalization and widespread use of algorithmic management systems, often receive inadequate protection under existing data protection laws or are even entirely excluded from such protection in some jurisdictions.
By Jeremias Adams-Prassl
Oxford Law Faculty Blog
Line manager, or line of code? When your boss is a computer program, questions of fairness, transparency and human agency become increasingly urgent. It may sound like dystopian science fiction, but algorithmic management – the use of automated systems to hire, supervise and even dismiss workers – is already transforming how organisations operate. What began in the gig economy (think ride-share drivers and food delivery couriers) has spread into a host of professional spaces. Professor Adams-Prassl has been examining these developments for over a decade. His early research into the gig economy – culminating in the award-winning book Humans as a Service – challenged the idea that companies such as Uber or Deliveroo represented something entirely new. “From a legal perspective, there was nothing novel there,” he explains. “Some employers have long tried to disguise their workforce as self-employed. What was new was the technology – the way apps were used to control workers in extremely tight ways.” That insight led to his European Research Council-funded project, iManage, which explores how employment law can adapt to algorithmic management. The project asks a deceptively simple question: when human decisions are replaced by algorithmic ones, how can the law keep up?
Videos
Podcasts
This site is part of a project that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 947806).
If you have any questions or feedback about this site or our research, please get in touch by emailing ai.work@law.ox.ac.uk.