Same Algorithm, Different Rights? Convention No. 193 and Human Review Beyond Platform Work

ILO Convention 193

Adopted in June 2026, the ILO Decent Work in the Platform Economy Convention, 2026 (No. 193) is the first global labour standard devoted to platform work. The Convention marks an important step in regulating the sector, while also addressing a broader concern: the use of algorithmic management in the world of work.

A 2025 OECD employer survey, covering 6,047 managers across six countries, found that 90 per cent of US managers and, on average, 79 per cent of managers in the European countries surveyed worked in firms using at least one algorithmic management tool. These tools allocate tasks and schedules, monitor working time, communications and many other decisions in the workplace.

Although they may make management faster and more consistent, they can also leave workers unable to understand how consequential decisions were reached or obtain meaningful review. According to the same survey, nearly two-thirds of managers using such tools reported at least one concern about their trustworthiness.

This expansion into traditional workplaces, examined by Aloisi, Joppe and Abraha as part of the wider “spillover effect” of algorithmic management, creates a regulatory mismatch. Workers may be exposed to comparable forms of automated control, yet receive different safeguards depending on whether they work through a digital platform or within a conventional organisation.

Convention No. 193 formally applies to digital labour platforms, and its binding effect depends on ratification and entry into force, while its practical effect will depend on national implementation. However, this article addresses a distinct and perhaps less explored question: can the Convention’s regulatory logic nevertheless serve as a useful benchmark for addressing comparable risks beyond platform work?

Convention No. 193 in the Evolving Regulatory Landscape

There is no question that platforms made algorithmic management more visible. Rating systems, automated task allocation and account deactivation are central to their business models, and workers often deal with the organisation almost entirely through an interface.

This justified urgent, sector-specific regulation. But it does not establish that the underlying risks are exclusive to that sector.

Convention No. 193, in Articles 13 to 15, creates an important architecture for automated management. Members must require platforms to provide information about systems used to monitor or evaluate work or generate work-related decisions and ensure the responsible use of those systems consistently with fundamental principles and rights at work. Where an automated system generates a decision, workers must be able to request a written explanation of significant adverse outcomes and, for matters such as non-payment, suspension, deactivation or termination, a review involving appropriate human involvement.

The European Union illustrates both the progress and fragmentation of the present framework. The Platform Work Directive contains detailed rules on algorithmic management, explanation and human oversight. The General Data Protection Regulation reaches beyond platforms, yet Article 22 focuses on decisions based solely on automated processing. Its application may therefore turn on whether human participation takes a process outside that category, even where an algorithmic recommendation exerted considerable influence.

The EU AI Act takes a broader risk-based approach and classifies several employment-related AI uses as high-risk. Article 86 provides a right to explanation for certain decisions based on their output. The Act does not, however, reproduce for every worker and algorithmic management tool Convention No. 193’s sequence of explanation, challenge and review.

Outside the EU, the legal picture is more uneven.

Some jurisdictions rely on general data-protection or employment-law principles, while others provide few specific safeguards. Existing data-protection laws often provide insufficient or inconsistent workplace protection, and some regimes exclude workers from parts of their coverage.

By contrast, an international labour Convention has a genuinely global institutional vocation. It was negotiated through the ILO’s tripartite standard-setting process, bringing governments, employers and workers into discussions designed to build broad agreement, and was formally adopted by the International Labour Conference. This institutional foundation gives the Convention the potential to help organise the next stage of regulation without stretching it beyond its formal reach.

Can a Sector-Specific Protection Become a Global Benchmark?

Earlier discussion on this Blog anticipated that the rules on algorithmic management in the Platform Work Directive could provide a blueprint for future legislation. Convention No. 193 takes that development one step further by placing rights to explanation, review and human involvement within the international standard-setting system of the ILO.

With legal scope and normative influence thus distinguished, the question is whether the Convention’s principles can guide responses to comparable risks beyond platforms.

Such influence is not unusual within international labour law. Nicolas Valticos distinguished obligations accepted through ratification from the wider function of international labour standards as guidance for national social policy. That authority is reinforced by the process through which ILO standards are negotiated and adopted by representatives of governments, employers and workers from different legal and economic systems.

As Xavier Beaudonnet’s study of domestic jurisprudence demonstrates, national courts may apply a ratified Convention directly where the domestic legal system permits, interpret national law in light of ILO instruments, or draw on those instruments when developing legal principles. The latter forms of use may include standards that are not directly binding in the case before the court.

There is a strong functional case for such influence here. The need for explanation and meaningful reconsideration does not arise from the use of a platform as such. It arises when an automated system materially shapes a decision affecting access to work, remuneration, discipline or continued employment, while the person concerned has little knowledge of the information or criteria used.

Despite being engaged through different organisational arrangements and subject to different legal regimes, workers may experience substantially the same procedural imbalance under algorithmic management.

The imbalance therefore arises not from the platform model itself, but from the combination of technological influence, concentrated decision-making power and an absence of effective means to challenge the outcome.

This functional reasoning is consistent with Guy Davidov’s purposive approach to labour law, according to which the scope and content of labour protections should be understood in light of the purposes they serve and the vulnerabilities they are intended to address. Davidov identifies dependency and democratic deficits (subordination broadly understood) as central justifications for labour regulation. Where an automated system exercises comparable power over a worker’s livelihood, the persistence of that vulnerability provides a principled basis for considering similar procedural safeguards.

Applying this reasoning would not extend Convention No. 193 beyond its formal scope. Instead, it would support the use of its framework of explanation and review as a benchmark where the underlying imbalance is functionally equivalent.

That reasoning should not, however, be taken too far. As David Cabrelli argues, labour law rests on a plural set of justifications rather than a single form of vulnerability. Functional similarity does not therefore establish that every platform-specific safeguard should be reproduced unchanged in conventional workplaces. Differences in the role assigned to the system, the seriousness of its influence, existing avenues of appeal and the surrounding institutional framework may justify different thresholds or remedies.

A purposive analysis therefore supports comparison without removing the need to justify the scope and content of each protection.

A useful parallel can be drawn with the Domestic Workers Convention, 2011 (No. 189), whose significance extends beyond the protections granted to a particular sector. It also shows how an international labour standard can expose wider regulatory gaps and develop solutions with broader relevance. By adapting general labour rights to the particular conditions of domestic work, Convention No. 189 anticipated concerns later addressed more broadly through the ILO frameworks on formalisation and violence and harassment, covering workers irrespective of their contractual status.

The Decent Work in the Platform Economy Convention may perform a similar function: although its legal obligations remain sector-based, it identifies a regulatory problem that extends beyond the sector and offers a framework for explanation and human review that may inform responses to comparable risks throughout the labour market.

Therefore, where an automated system materially shapes a decision carrying serious consequences for a worker’s livelihood, the shared vulnerability identified provides a persuasive reason to consider similar rights to explanation and genuinely empowered human review. The platform boundary can continue to define the Convention’s formal reach without also defining the limits of its normative contribution.

Seen in this light, Convention No. 193 is neither a universal code for artificial intelligence at work nor simply an isolated response to one sector. Its wider contribution may lie in providing an internationally negotiated reference point for a principle of growing importance: when an automated system plays a material role in a decision capable of threatening a person’s livelihood, that person should be able to understand the basis of the decision and obtain genuine review by a human being.

The Convention’s legal obligations remain sector-specific, but the human-centred logic underlying them need not.