Regulatory responses to algorithmic management in the EU

What is algorithmic management? Why is it regulated?

Algorithmic management can be defined as 'the use of computer-programmed procedures for the coordination of labour input in an organisation'. 1Part of a broader process of the digitalisation of work, algorithmic management involves the collection, storage and processing of large amounts of workers' data and other organisational data with the purpose of achieving efficient and fast decision-making.

Algorithmic management therefore includes the potential automation of the full range of traditional employer functions: hiring and managing the workforce, allocating tasks, monitoring performance and deciding on promotions and dismissals. While algorithmic management systems currently deployed in companies do not fully automate decision-making, they are used to assist decision-making processes or to deal with specific situations in which parameters are predefined. Many of the algorithms currently used mainly automate internal processes without necessarily learning or providing the more advanced analytical assessments that are typically associated with artificial intelligence (AI).

The connection to platform work

Since algorithmic management practices were initially developed and deployed for managing platform workers, this also becamethe first area to be regulated by national legislation. In 2021, Spain developed the first piece of legislation (Rider's Law) which introduces a legal presumption of dependent employment for platform workers in the delivery sector and a set of transparency requirements for the use of algorithmic management systems applicable to all employers.

In March 2024, EU employment and social affairs ministers approved the provisional agreement made between the European Council and the European Parliament on the proposal for a directive on improving working conditions for platform workers. The approved text includes a chapter on algorithmic management that introduces a set of rights applicable to all people providing work through platforms. These include human oversight of automated decisions, the right of platform workers to obtain an explanation for any decisions taken or supported by an algorithm, and the right to contest decisions taken by an algorithm. The future directive also grants a better understanding of how prices are set and how tasks are monitored by the algorithms. As of April 2024, Belgium, Croatia, Greece, Malta, Portugal and Spain have moved to embed provisions of the directive on improving working conditions in platform work into their national legislation (see the figure below).

For example, the Maltese regulation is established through the Digital Platform Delivery Wages Council Wage Regulation Order of 2022. 2This explicitly addresses the rights of platform workers in relation to algorithmic management, including the right to know about how an automated system works and the extent to which this system has an algorithm with decision-making capacity. This is especially so if an automated decision-making system supports or impacts decisions on working conditions, working time, employment status, as well as access to work assignments and the worker's digital account.

It also highlights that the company must provide information about the data being collected and used by automated decision-making systems. An employer must also disclose to workers any information about automated monitoring or evaluation systems. Such information must be provided in writing and the employer is obliged to clarify the facts leading to such a decision. If a worker finds that insufficient information is provided by the employer, the platform is obliged to review the information and provide a response to the platform worker within one week of receiving the request (or two weeks for small and medium-sized enterprises). Moreover, employers must keep track of and evaluate the impact of automated monitoring and decision-making systems.

• ERM database: European Restructuring Monitor - Malta: Algorithmic management

Regulating personal data in the EU and the case of Germany

The key regulation governing the use of personal data, which is also relevant for addressing challenges raised by algorithmic management systems, is the General Data Protection Regulation (GDPR), in force since 2018. The relevance of the GDPR for the use of algorithmic management systems stems from its emphasis on personal rights. Under Article 22, the regulation prohibits decisions based solely on automated processing, including profiling. The GDPR also establishes a series of other rights which have implications in the context of algorithmic management systems, including rights related to access (Article 15), rectification (Article 16), restricted processing (Article 18), and data portability (Article 20). However, it does not specifically target the employment context. This is left up to Member States, which may, by law or by collective agreements, provide for more specific rules (Article 88).

More specific rules that expand on Article 88 exist in Germany. German legislation on algorithmic management is based on the Federal Data Protection Act (BDSG) of 30 June 2017, particularly in Section 26, which deals with data processing for employment-related purposes. The law details the purposes for processing employee data and sets strict rules on obtaining consent as well as conditions for processing sensitive employee data.

Recently, however, several court rulings on whether existing legislation sufficiently implements the GDPR in an employment context have led to an active discussion on the subject. On 7 February 2024, the German Federal Cabinet approved a draft law amending the BDSG. The proposal envisages, for instance, the inclusion of solo self-employed platform workers within the scope of the law, the strengthening of data subjects' rights, and the clear regulation of consent.

While the BDSG is concerned with individual rights, existing regulations in Germany also focus on collective rights. In this sense, the German Works Council Modernisation Act includes the right of works councils to be informed and consulted with respect to the use of AI in personnel selection and work process planning. This act can be understood in the context of the ample employee co-determination rights in this country as well as in the EU legislation, the directive establishing general principles on the minimum information and consultation rights for employees in companies based in the EU, discussed in the next paragraph.

• ERM database: European Restructuring Monitor - Germany: Algorithmic management

Application of the EU Directive on Transparent and Predicable Working Conditions in Italy

Another body of European legislation that can be used by Member States to regulate the use of algorithmic management systems in the workplace are the Information and Consultation Directive and the Directive on Transparent and Predictable Working Conditions. The Information and Consultation Directive (2002/14/EC) establishes a general right for information and consultation that can also be applicable when algorithmic management systems that introduce significant changes in work organisation are launched. 3Examples of such changes are instances in which algorithmic management systems allocate tasks, track performance and monitor employees.

The Directive on Transparent and Predictable Working Conditions provides a set of employment rights that can be used to address challenges raised by the use of algorithms at work. These include, for example, the right to be informed early on essential aspects of work and the right to be informed within a reasonable period in advance when work will have to be done (Articles 4 and 5).

In transposing the Directive on Transparent and Predictable Working Conditions in 2022, the Italian law (Decreto Legislativo 104/2022) makes explicit reference to the use of algorithmic management systems and requires employers to declare the use of such systems in any aspect of the employment relationship (e.g. recruitment, work organisation, evaluation, dismissal). The law also requires employers to provide information on the type of data collected to train the systems and the cybersecurity measures implemented to protect collected data against potential breaches.

As several aspects of the original decree remained unclear, such as to whom this information is supposed to be provided (Palermo court dispute No. 14491 on 3 April 2023), amendments to the 2022 legislation were made through Article 26 of Law No. 85 of 3 July 2023. The updated decree requires that the information prepared pursuant to Decree No. 104/2022 must not only be provided to individual workers, but must also be communicated to trade unions. In the event of non-compliance, the employer risks being charged with anti-union conduct, as recently seen in the Palermo court. The amendment also limits the applicability of the decree only to fully automated systems. Moreover, the amendment clarifies that the obligation to provide information about the use of fully automated decision-making is not applicable to systems protected by industrial or commercial secrecy.

• ERM database: European Restructuring Monitor - Italy: Algorithmic management

Beyond legislation: The Dutch example

While the examples above focus on hard law approaches to addressing the issues raised by the use of algorithmic management systems, other countries such as the Netherlands have opted for a different approach. The Netherlands established an algorithm register on 21 December 2022. The register is freely accessible to the general public and, according to the Dutch government, aims to make the use of algorithms understandable and transparent. 4

All government bodies and agencies are encouraged to enter information about the algorithms they use into the register, with the goal of improving the transparency of such systems. For example, one register entry is the 'Automated document check and face comparator' algorithm used by a municipality. However, while this example shows how detailed the information about individual algorithms in the register can be, there are no publicly available rules on what information needs to be included in the description or in other sections. The current entries in the register therefore include different levels of detail.

In launching the register, the Dutch government consulted civil society organisations in the field of human rights and other experts in the field. As the register continues to develop, the goal is for it to provide full coverage of algorithms used by public bodies and to be up to date from 2025 onwards.

• ERM database: European Restructuring Monitor - Netherlands: Algorithmic management

Additional regulation implemented at EU level

In addition to existing regulatory levers, efforts are ongoing in the EU to address broader concerns around AI, which are also consequential for regulating the use of algorithmic management systems. In March 2024, the European Parliament approved the expected final text of the EU Artificial Intelligence Act, and is currently waiting for final confirmation from the Council. The act classifies AI systems into three levels of risk: prohibited AI systems, high-risk AI and low-risk AI. AI applications used in employment contexts are classified as high-risk AI and must comply with mandatory requirements. These include conducting risk assessments, ensuring the use of high-quality data, informing users about the purpose of the systems and ensuring human oversight. The act will be gradually phased in over the next three years, with the key obligations expected to be implemented in the next 24 months.

Conclusion

This article has highlighted existing national and EU initiatives addressing the use of algorithmic management systems in the workplace. The EU initiatives are currently the driving force behind national legislation around algorithmic management. EU legislation includes some provisions to address risks for workers, in particular by increasing the transparency of algorithms such as in the case of the GDPR, the Directive on Transparent and Predictable Working Conditions and the Information and Consultation Directive.

Nevertheless, there are important differences at national level across countries. Member States are at different stages of transposing EU regulations and are doing so in different ways. At the time of writing, the vast majority of European countries still do not have a national regulation in place and only some countries have embedded relevant provisions in existing legislation. Those who have, legislate the use of algorithmic management systems in different ways. While some countries prioritise the transparency of processes that automated systems use to inform decisions, others focus on the inputs or the data that are collected and used by such systems.

Image © Adene Sanchez/peopleimages.com/Adobe Stock

Footnotes

• 1

  Baiocco, S., Fernández-Macías, E., Rani, U. and Pesole, A. (2022), The algorithmic management of work and its implications in different contexts, European Commission, Seville, JRC129749.

• 2

  Department for Industrial and Employment Relations of Malta(2022), FAQs on the Digitals Platform Delivery Wages Council Wage Regulation Order of 2022(PDF).

• 3

  De Stefano, V. and Wouters, M. (2022), AI and digital tools in workplace management and evaluation: An assessment of the EU's legal framework, Osgoode Legal Studies Research Paper No. 4144899.

• 4

  Overheid.nl (n.d.), Het Algoritmeregister van de Nederlandse overheid.

Related series