Decoding Algorithmic Management through Workers' Calculative Power: Potential and Limitations of a Participatory Methodology for AI Accountability in Platform Labour

This paper discusses the potentials and limitations of a new participatory methodology combining interviews, workshops, and data science techniques for analysing algorithmic management in digital labour platforms. It was developed within Data4Workers, an international comparative study of food delivery and care sectors spanning Chile, Uruguay, Spain, and Switzerland. It examines to what extent workers' personal data, recovered from platforms, can be mobilised as an empirical entry point into the data infrastructures and algorithmic logics structuring work organisation and managerial control.

At the intersection of the sociology of work, digital labour studies, and labour and data protection law, and working alongside workers, an NGO, and trade unions, the project examines to what extent platform power can be countered by developing collective tools and practices allowing workers to reclaim, and analyse their data.

The methodology draws on the General Data Protection Regulation (GDPR) and equivalent regimes because the platforms studied are European-headquartered, GDPR Article 15 access rights formally apply to workers in Latin America as well as Europe. In practice, their exercise varies. In both regions, the right is underused due to lack of awareness and weak collective organisation; in Latin America, workers also face lack of institutional support and adverse political conditions. Platforms obstruct access through technical and procedural barriers, and in Uruguay, workers who exercised these rights faced retaliation. The variability is revealing: where one platform complied in Europea, workers mobilised the data to deepen understanding of their conditions, whereas in Latin America the same company responded with obstruction.

Outcomes also differed across sectors. The methodology proved more effective in food delivery, in Europe, where workers are more publicly visible and platforms have more established access procedures. In care, workers are even more isolated, and requests were more often unanswered or delayed, with platforms invoking lack of precedent.

Where requests succeeded, the returned data required extensive processing and analysis, which researchers carried out to equip unions and workers with usable evidence. These data were then discussed through individual interviews and collective workshops, enabling a co-construction of knowledge central to the methodology.

Systematic comparison across countries and sectors produced a second-order finding: the quantity, granularity, and structure of recovered data reveal the extent of platform data collection, the degree of automation in managerial decisions, and the data points potentially used as input for algorithmic management. These differ markedly between food delivery, where timestamped geolocation sustains working time and task control, and care, where logs of job-ad applications and messages structure economic dependency through distinct platform logics.

We contribute to debates on algorithmic management in both masculinised and feminised sectors of the gig economy, while assessing the potential and limitations of participatory research based on data access rights for holding AI-driven labour control accountable. These rights are increasingly important to labour rights yet unevenly enforced, routinely violated, and likely to be further fragilised by the AI Act and Digital Omnibus, which favour companies' interests. The experience exposes the dominant power of platforms over workers, unions, and states