You are on page 1of 30

Algorithmic management of work performance

I. Introduction

Emerging technological innovation is transforming the way businesses are managing

their employee, in an ever-changing way. In fact, data collected through workforce

analytics procedures is employed as fuel for algorithmic management solutions

which might determine choices that affect the workforce. The current revolution of

technological growth is driven by the expansion of cyber-physical infrastructure and

integrated systems, which offer innovative monitoring activities. Although using

technology can result in increased labor productivity, the use of technology is not

always risk-free. Such that, in the realm of human resource management, algorithms

have already shown to be far from unbiased decision makers. This issue is frequently

worsened by the lack of transparency that characterizes a substantial portion of

automated decision-making systems. As a result, workforce analytics and

algorithmic management methods may imply an unprecedented extension of

managerial prerogatives. However, only recently have International authorities

began to consider how to modernize present regulatory framework in order to

address the intricate and far-reaching challenges presented by pervasive tech

devices, notably Artificial Intelligence, a general-purpose tool capable of simulating

adaptive and predictive "functions associated with human intelligence"1. These

robust and diverse technologies may be seen as an optimal mix of big data analytics

1
Aida Ponce Del Castillo, A law on robotics and artificial intelligence in the EU? (2017 ETUI FORESIGHT BRIEFS
1).
and algorithmic governance in the context of an employment relationship2.

Technological advancements guarantee that information exchanges are interactive,

accessible, and efficient. More importantly, given that hyperconnected equipment is

answerable for a significant alteration of how work is rendered, both at the individual

and collective levels, these devices are expected to dramatically affect the

relationships between employers and employees (or between customers and

workers)3. Simultaneously, the employer's day-to-day decisions might be more

informed, property rights can be safeguarded, productivity can be enhanced, loss of

company assets can be prevented, and waste production may be decreased4. The core

of the current renovation is the extremely blurred line between personal and

professional lives, which "creates serious difficulties to privacy and data security”5.

These concerns raise a number of issues that must be addressed in order to avoid

abuses in the relationship between companies and workers (to the detriment of the

latter) and to ensure that the use of forms of remote control of workers is oriented

not only to generate profits but also to generate long-term competitive advantage6.

Furthermore, the issue of monitoring and managing activities and processes within

companies is a crucial element in order for the organizations to succeed. Improving

processes entails first measuring and monitoring them so that individual operations

2
Mirela Ivanova et al., The App as a Boss? Control and Autonomy in Application-Based Management, 2
interdisziplinärer arbeitsforschung (2018).
3
Id.
4
Massimo D’Antona, “Informatizzazione dei processi decisionali e diritto del lavoro: algoritmi, poteri datoriali e
responsabilità del prestatore nell’era dell’intelligenza artificiale”, Centre for the study of European Labour Law.
(2021).
5
Mirela Ivanova et al., The App as a Boss? Control and Autonomy in Application-Based Management, 2
interdisziplinärer arbeitsforschung (2018).
6
Id.
may be coordinated and create overall outcomes, allowing management to make

informed decisions. However, whereas technological advances allow for enhancing

such relationships and boosting worker autonomy and engagement, they may also

be utilized in a punitive fashion and with a function of oversight rather than

supervision. Instead of supporting an emancipating new environment, invasive

technology has the potential to strengthen hierarchy and control over work

performance7.

Ultimately, algorithmic management has altered company paradigms and frustrated

conventional economic, legal, and regulatory systems. These changes require new

techniques and ways of thinking about the power dynamics that are reshaping

workplaces. Focus will be given to creative ways of organizing the workplace,

especially inside Amazon’s warehouses, which includes these new systems

employed by more companies and economic sectors nowadays – providing

innovative approaches to workforce management - however, posing several ethical,

privacy, and safety and health issues.

II. Algorithmic management of work performance

Algorithmic management refers to the application of computer-programmed

methods to coordinate labor input in an organization8. Algorithmic management is a

revolutionary combination of existing (and easily accessible) technology that has the

7
ILO, “The Algorithmic Management of work and its implications in different contexts”, The Algorithmic
Management of work and its implications in different contexts- Background Paper Series of the Joint EU-ILO Project
“Building Partnerships on the Future of Work"
8
Id.
potential to disrupt current economic practices, which encompasses a wide range of

technology tools and strategies. In particular, is associated with analytics of big data,

machine learning, geolocation, integrated mobile devices, wearables, and other

technologies9. It should be understood as a specific method of merging and utilizing

such technologies in order to automate or at the very least support some of the duties

traditionally performed by human management for work coordination and

management10. It can be described as the ongoing monitoring and assessment of

employee performance and behavior, as well as the automatic application of

algorithmic decisions11. It is becoming increasingly common for employers to

utilize algorithms that are aimed at predicting or assessing workplace productivity

and performance. Many algorithmic management aspects were established by

enterprises in the "sharing" or "gig" economy, as Uber, Lyft, or multi-service

platforms like TaskRabbit12. These technologies, however, are not restricted to the

"gig" economy. They are prevalent across several areas of employment and have

evolved among other business applications of computing13. Systems, for instance,

such as Netflix, Spotify, and Amazon are the driving force behind these platforms. As

this essay will illustrate, certain features of algorithmic management are increasingly

commonplace in workplaces ranging from transportation and logistics to retail and

9
Id.
10
M.H.Jarrahi, G.Newlands, M.K.Lee; C. T. Woolf, E. Kinder, W. Sutherland, “Algorithmic management in a work
context”, (2021)
11
Mareike Möhlmann and Lior Zalmanson “Hands on the Wheel: Navigating Algorithmic Management and Uber
Drivers’ Autonomy” (paper presented to the International Conference on Information Systems, December 2017) at 4–
5.)
12
Id.
service sectors. Companies are using these tools to a greater extent not only to

oversee the performance of employees in an optimized manner at a large scale, but

also because they play a crucial role in enhancing the quality of managerial decisions

adding value to them in terms of productivity benefits14. Algorithmic management

elements, whether paired with or utilized to replace existing processes, broaden the

scope, size, and purpose of surveillance and data collecting15. Therefore, it can be

affirmed that this type of technology is transforming how companies handle their

workforce, throughout a process were decisions, that once were taken directly by

humans, are now delegated to digital managers. Observing how the platforms

operates in this setting, it becomes clear that the employer is granting access to many

typical organizational functions, which transcend the traditional employer and

employee relationship creating an entirely new one, that inevitably led to several

issues for the workers subject to them.

Having said that, the constant tracking of workers’ behavior is a precondition for

effective algorithmic management. While human managers have typically been able

to establish deep, trust-based, and long-term relationships with their employees, this

is impractical while supervising thousands of people via digital platform. Unlike

traditional management settings, algorithmic management is based on a continuous

collection of data conducted by digital devices that connects the worker to the

platform16. Different technologies are used for monitoring and to collect a significant

14
A. Ingrao, “Il controllo a distanza sui lavoratori e la nuova disciplina privacy: una lettura integrata”, (2018).
15
A.Aloisi, E.Gramano, “Artificial Intelligence is watching you at work: digital surveillance, employee monitoring, and
regulatory issues in the EU context.
16
Id.
amount of data in a fast way, including both basic cameras and wearable sensors

worn by the employees themselves, that allow the company to rapidly determine

whether they are productive or not. In other words, the employer has the confidence

in the platform’s ability, whose role and responsibilities must be outlined, to achieve

its business objective17.

Another feature of the algorithmic management is the continuous performance

evaluation of workers, which is also possible through data tracking. Although

performance assessment is often employed in traditional work contexts to reward or

penalize employees, algorithmic management methods involve constant and real-

time performance review using a variety of micro-tasks, culminating in a significant

number of performance evaluations every day18. For instance, an algorithm may

utilize these data to forecast a worker’s performance at different time of day or when

matched with other individuals. Alternatively, the algorithm may simply

synchronize these variables to get a productivity/performance evaluation or "score"

for a specific worker. Thus, these technologies have made it possible to collect data

with more granularity while using subtle techniques on a larger range of aspects19.

Digital technologies are not restricted on monitoring workers performance; they can

also be used to monitor behaviors and personal characteristics, such as the usage of

information for reasons other than the original ones intended20.

17
Id.
18
Id.
19
Voza, R., “Il lavoro e le piattaforme digitali: the same old story?” 2017.
20
Id.
An additional feature of this kind of management is automated decision. In fact,

algorithmic management is distinguished by little or no human interaction.

Algorithms compute and make judgments that are typically carried out

automatically. Companies can use automated implementation to speed up operations

and adapt quickly to regularly changing circumstances21.

In this fashion, duties are thus determined through an automated process: If X is

exceeded, Y will happen. This could imply, for instance, that a letter of dismissal is

automatically sent by email if the employee's activity decreases by a certain number

of hours22. In other words, the company establishes a condition and a consequence;

upon the occurrence of that condition the algorithm automatically activates the

consequence, hence eliminating the emotional component that influenced such

interactions before the development of these technologies. Thus, this does not allow

for discussion or revision of choices based on unusual conditions that are not fully

captured by the data. For instance, workers may be kicked out for what the algorithm

believes to be a violation, even if it was a system error23.

In this circumstance, it is possible to for the company to increase the number of tasks

that needs to be completed during the shift while increasing the intensity of work

required to be done. This prospect for algorithmic decision-making has led to many

concerns about the welfare of the workers who are subject to these decisions.

Additionally, the results of the examinations carried out by the European Agency for

21
Aloisi, E.Gramano, “Artificial Intelligence is watching you at work: digital surveillance, employee monitoring, and
regulatory issues in the EU context.
22
Angeli, F., “La quarta rivoluzione industriale”, Milano 2016.
23
Id.
Safety and Health at Work demonstrate that the primary causes of stress are related

to working hours and workload24.

The worker may be required to raise its working pace at the predetermined rates as

soon as the algorithm distributes the jobs and establishes the execution times. The

pressure placed on the employee in this way may also cause irritation and

discouragement. This is due to the fact that an algorithm, by definition, determines

the most effective way to complete a task and arrange a job in order to maximize

corporate productivity and optimize job management, by putting pressure on

employee's workforce and removing all chances for him to make independent

decisions25.

Therefore, the centralized knowledge can be translated into a centralization control

over the work process (and productivity). Such extensive monitoring and

surveillance lead to information asymmetries in which knowledge, and hence control

over production, is concentrated at the management level. Management gains

exclusive knowledge about the tasks to be performed, how to perform them, and

how to enhance such performance, especially when the data and techniques utilized

are ambiguous26.

As a result, since algorithmic management centralizes knowledge and control,

information gaps between management and employees can result in power

24
X. Irastorza, M. Milczarek, W. Cockburn, EU- OSHA, Second European survey of enterprises on new and emerging
risks (ESENER-2). Overview report: managingsafetyandhealth atwork,ISSN1830-5946,ISSN1831-9343,2013]
25
A.Aloisi, E.Gramano, “Artificial Intelligence is watching you at work: digital surveillance, employee monitoring, and
regulatory issues in the EU context.
26
Id.
imbalances, which tend to worsen if openness about algorithms and data is restricted

or constrained27. This is how algorithmic management may result in a change in the

power dynamics between employees and management inside an organization, which

could eventually have a negative impact on employment relations.

III. The Amazon experience

Algorithmic management is prevalent and developed mostly in digital employment

platforms, which provides various managerial duties by automatically distributing

tasks to employees via wearable devices, cellphones, and computers. In addition,

this new digital labor platforms have very high and invasive degrees of algorithmic

control for a multitude of work-related aspects, such as the working techniques, the

timing, and the amount of required effort. Meanwhile, the enormous amount of

information generated by the digitalization of economic processes is simply

impossible to manage without some level of algorithmic control, and once

digitalization has reached a certain level, it will probably be economically

advantageous to automate some management tasks for the coordination of labor

input28. For instance, centralized data servers often oversee the collecting of real-

time and geolocated data on goods and processes in a company. These servers

employ automated (algorithmic) techniques for processing the data and turning it

into useful intelligence29. Even if such information isn't specifically designed to

27
Id.
28 28
Acemoglu, D. (2002). “Technical change, inequality, and the labor market”, Journal of Economic Literature, vol.
40, no. 1, pp. 7–72.
29
Voza, R., “Il lavoro e le piattaforme digitali: the same old story?” 2017.
evaluate employee performance, managers are likely to utilize it for that purpose

because it will be extremely useful in that regard.

In this fashion, focus is given to Amazon's internal warehouse processes, which to

some extent emphasized the aforementioned methods. However, this represent only

the most striking example given that Amazon is considered one of the largest private

employers in the world, though is significant to highlight the system utilized by the

company. While algorithmic direction is central to many types of platform work, it

has also been identified in conventional employment settings, such as warehouses.

The warehouse management practices utilized by Jeff Bezos' company are indicative

of a collapsing world, and it is no coincidence that it has been labeled "Bezosism” 30.

In fact, the company implements one of the highly intrusive and aggressive systems

in the world to exert pressure on employees to work faster, so that they can monitor

their adherence towards the company culture or spy on labor activists protecting the

workplace. This highlights how workers are dependent on algorithmic

instructions sent to them via their handheld devices, which primarily serves as

an interface (surveillance) between managers and employees.

The way Amazon’s algorithmic management is implemented is primarily via

control systems, such as portable devices with integrated barcode scanners,

motion, and location tracking. In this manner employees adhere to a few

fundamental guidelines inside the warehouse, providing Amazon's software

30
(1) Christopher Mims,“The Way Amazon Uses Tech to Squeeze Performance Out of Workers Deserves Its Own
Name: Bezosism.
system the ability to precisely identify the position of each product and

automatically allocate the most productive item to employees by considering

both their relative placements and the locations of other workers who could

otherwise get in the way31. Once a worker has been assigned an item, they are

informed via text and images presented on their handheld device. Additionally,

the device makes real-time plans, taking into consideration the location of

goods and other employees, the path to be taken, communicating each step to

the worker, as well as the item’s location on the shelves32.

Such innovative method of organization demonstrates how warehouse’s

employee, as well as package delivery employees - are subjected to employment

conditions that are entirely managed and governed by algorithms, placing a

strain on productivity; for instance, the algorithm may require to pick up 100

products each hour from the warehouse shelves, or transport 30 packages per

hour to the streets of Rome or Paris. In other words, devices, such as the scanner,

assign tasks, read data from bar codes on products that Amazon sells, and use the

data to monitor every movement of operators within the warehouse.

To describe the issue, only a few of the company's systems will be discussed.

As briefly outlined, performance of Amazon’s warehouse employees is regularly

tracked through the newly implemented application named “ADAPT”33. The app

31
ILO, “The Algorithmic Management of work and its implications in different contexts”, The Algorithmic
Management of work and its implications in different contexts- Background Paper Series of the Joint EU-ILO Project
“Building Partnerships on the Future of Work
32
Id.
33
Associate Development and Performance Tracker.
allows the employer to delegate the various tasks to its employee, monitor business

operations, and compelling its staff to work more diligently and quickly to uphold

its goal to a smooth and convenient delivery. To achieve this objective, the company

provides its employees with computers, or manual barcode scanners, which enable

them to easily log onto Amazon’s algorithmic system. Once connected, the system

starts to allocate various and specific tasks that must be executed during the shift

within a specific time frame34. These tools finally give the information gathered to

the ADAPT software, which tracks employee productivity and evaluates the speed

of assigned tasks including locating, scanning, and packaging. Since each user logs

in using a unique device, Amazon's management can confirm the productivity rate

of each employee35.

The company evaluates the output of its employees by measuring their performance

using the "rate" or "making the beat" metric36. This practice has been named as

“Amazon pace”, basically tracking how quickly assigned activities are performed,

the amount of work that must be completed in an hour, or the period the employee

is away from the device for lunch break or to use the restroom37. When an operator's

inactivity exceeds a predetermined threshold, the system generates a "scores tot" and

operators who accumulate too many points receive warnings38.

34
A. Delfanti, L. Radovac, T. Walker, “Il panopticon di Amazon”, p. 6.
35
Id.
36
Id.
37
Surveillance, Stress, and No Bathrooms: Life as an Amazon Drive, Jake Alimahomed-Wilson
38
Id.
Monitoring warehouses’ workers performance represent just one of the most

unpleasant aspects of these practices.

Indeed, couriers and truckers delivering Amazon packages are not exempt from

digital surveillance. In this regard, in 2015 Amazon launched the app “Flex” which

currently operates in over 50 US cities and is based on a network of independent

contractors that use their personal vehicles to deliver products using the company's

delivery app. Flex is thus mainly used for delivery packages to the customer’s home,

specifies the delivery location and the route details, which must be achieved in

blocks of 2–6 hours3940.

Along with Flex, drivers must also download the "Mentor" app41, which analyzes

and assesses their driving behavior 42. These two apps work in tandem to ensure that

employees are constantly under surveillance, during their shift, controlling their

location and the time spent on each handover, without truly "escaping" from the

employer's sight. An additional, and perhaps unpleasant feature of Mentor, is that it

even manages the employee's mobile phone usage, keeping track of all calls or

messages sent (even private flows)43.

As though there weren't enough applications that monitor real - time location and

speed, Amazon drivers are now visually observed. In this regard, the company plans

to install on the delivery vehicles the system called “Driveri”44.

39
Id.
41
Annie Palmer, “Amazon uses an app called Mentor to track and discipline delivery drivers”.
42
Ibidem.
43
Ibidem.
44
A. Delfanti, L. Radovac, T. Walker, “Il panopticon di Amazon”, p. 11
Driveri combines cameras and artificial intelligence to track the employee driving

habits and behavior – in real time - from the moment they start the vehicle until the

end of the shift45. These cameras record the road ahead, the driver, and both sides of

the vehicle, and instruct the worker on how to respond to safety infractions. For

instance, it will advise workers to keep a safe distance from other automobiles, to

slow down if they exceed the speed limit, to avoid unscheduled stops, and to take a

15-minute break if they are found yawning46.

In addition, the systems do not record the audio and does not have the “live view”

option, but the app can issue beeps accusing the driver of “driving distractedly” if

he holds the phone (even if he is using the app Flex or Mentor)47. The software then

provides a real-time score on the driver's driving behavior and transmit the data the

site manager48.

In practice, Amazon employees are exposed to a sophisticated digital surveillance

experiment, which is causing a negative impact on their health and quality of life,

and that resemble a modern-day Big brother. This dystopian reality of an ever-

vigilant system has real-world consequences thus increasing each employee's

vulnerability to surveillance. In fact, Amazon's warehouses have the highest accident

rate in the industry, due to a given status of stress and anxiety, originated by a

perfectly designed system intended to exert pressure on employees to reach another

form of high rate which lean toward the company's welfare: the productivity.

45
Menegus, Bryan. “Amazon’s Last Mile.” Gizmodo, November 16, 2017
46
(Palmer, 2021)
47
Ibidem
48
Ibidem.
Finally, the technology management methods implemented by Amazon has been

defined, inter alia, as a “panopticon”49 – which creates a condition on absolute

uncertainty as to when and if the individual is supervised - causing a limitless state

of self-control and an unconscious lack of personal liberty.

IV. Legal issues.

From this latest revolution it seems evident how the beginning of the millennium

represents an historical turning point for the global employment conditions and

given the wide-ranging implications of AI on societies and individuals, it is certain

that internationally protected human rights will be affected by developments in this

field.

When it comes to “surveillance”, new technologies represent a true gamechanger,

contributing to a decisive shift away from a direct and physical control by the

employer or by middle management to a model based on various data

collected through remote scrutiny50. It is legitimate for an employer to surveil

employees’ performance. The employer can vet employees with a view to assess

their skills and productivity as discussed above, but also generally to assess interests

and competences. The modifications introduced by algorithmic management in

work organization may have a detrimental effect on numerous areas of job quality.

49
Christopher Mims,“The Way Amazon Uses Tech to Squeeze Performance Out of Workers Deserves Its Own Name:
Bezosism., The Wall Street Journal.
50
A.Aloisi, E.Gramano, “Artificial Intelligence is watching you at work: digital surveillance, employee monitoring, and
regulatory issues in the EU context.
Concomitantly, employers are responsible for compliance with health and safety

measures, and liable for any tort committed at work by any employee under their

direction. The protection of the employee’s personal life represents another limit to

managerial authority: controls must be carried out without prejudice to the human

dignity of the employees51. In addition, centralized algorithmic control may also

result in increased workload and worse working time quality, especially when the

number of tasks to accomplish rises and/or the time to complete these activities

reduces, the workload increases52.

As a result, the employees are under time pressure and must adjust their work pace

to meet the demands, which ultimately causes stress. There are, inter alia, new

threats to the psychological and physical well-being of employees, necessitating new

precautionary measures, which need to be considered in this context.

There is no doubt that health is a fundamental right in the hierarchy of fundamental

values, which means that the employer must take measures to prevent performance

(even if indirectly determined by algorithms) from causing harm to the worker's

health. Health concerns for employees are primarily caused by the digitization of

information and the use of databases, which require the management of a significant

amount of data and generate a dangerous increase in performance intensity levels53.

In addition, the emotional distress caused by this greater commitment, is further

51
Catherine Delbar, Marinette Mormont, & Marie Schots, New technology and respect for privacy at the workplace,
Institut des Sciences du Travail (2003).
52
ILO, “The Algorithmic Management of work and its implications in different contexts”, The Algorithmic
Management of work and its implications in different contexts- Background Paper Series of the Joint EU-ILO Project
“Building Partnerships on the Future of Work.
aggravated by the usage of devices that enable constant connectivity, interfering

with both personal and professional life54. This blurred boundaries between work

and non-work leads to a subtle promiscuity between the private and professional

spheres, which, on the other hand, should remain separate to ensure sufficient rest to

recover the psychophysical energies55. To these concerns, there is the risk of

“multitasking syndrome”, which is the ongoing tension generated by the repeated

use of digital technologies, and that limits the capacity to focus having a negative

impact on interpersonal interactions, inhibiting the workers to have the possibility to

exchange views and perceive empathy on their jobs56. This would entail a substantial

loss for the social setting since human communication is crucial for maintaining a

positive social environment, ensuring the workplace health and safety of employees,

accepting, and minimizing – as far as possible - the negative effects of algorithmic

technology in the workplace.

This is associated with performing the tasks in different places far from the

workspace, as has been described with Amazon couriers, but Uber and Lyft drivers

undoubtedly fall into this category. This is particularly true in the case of workers

employed under atypical employment contracts, which are generally characterized

by reduced training and experience, which leads inevitably to a higher risk of

injuries57. In this respect, in addition to health protection, mention should be made

54
Katharine Parkers, Work hours and well-being: The roles of work-time control and work family interference.
55
Ibidem
56
Rosa Di Meo, “Il diritto alla disconnessione nella prospettiva italiana e comparata”, (2017).
57
ILO, Non-standard employment around the world: Understanding challenges, shaping prospects, Genève 2016 pp.
200 ss.
to safety protection as well, which involve all categories of workers, especially those

who perform their off-site duties. For instance, the divers who are exposed to the

risks related to road traffic and to the stress generated by the constant tension

imposed by the respect of the delivery times58. Indeed, if, on the one hand, the worker

is entitled to use the space provided by digital devices for a better work-life balance,

on the other hand, they also run the risk of being forced to maintain an unlimited

level personal availability in line with the productivity and interaction requirements

with the enterprise, up to the point to be conditioned in the personal use of the own

time life59. From this point of view, as is the case in some European countries, the

individual agreement between the parties is of particular significance, with respect

to «rest periods» and technical and organizational measures necessary to ensure the

disconnection of the worker from the technological tools of work, thus attempting

to the worker's right to a disconnection as extremely essential60.

Furthermore, as already pointed out, these new digital technologies have permeated

the workplace, significantly extending the possibilities for employee monitoring and

surveillance, which lead to concerns pertaining the protection of workers from

surveillance, thus raising question on whether these systems are compatible with the

human right to privacy protection61.

58
EUROFOUND and ILO, Working anytime, anywhere: The effects on the world of work, Luxembourg e Geneva 2017.
59
Ibidem
60
Rosa Di Meo, “Il diritto alla disconnessione nella prospettiva italiana e comparata”, Università Politecnica delle
Marche, VOL. 3, no. 2, 2017.
61
Lindsey Anderse “Human Rights in the age of Artificial Intelligence”, (2018).
The boundaries of workers’ privacy are challenged when monitoring extends to the

tracking of non-work-related activities, especially in the case of technologies such

as GPS location tracking, computer-monitoring software, wearables, and remote

sensors. Such intrusive monitoring also raises concerns and issues about the worker's

capacity to consent to or opt out of the collection of personal information62.

Monitoring systems must be introduced and implemented in a way that protects

workers' autonomy while still complying to all data protection regulations63.

In this regard, the EU General Data Protection Regulation – which has been

implemented in national legislation in all member states, - has modernized the data

protection framework by improving individuals' rights and safeguards in terms of

their personal data, and this is applicable to the work environment. The Regulation

seems innovative because has made the employer accountable, having a duty of care

towards their employees, and requires transparency of their data collection

processes, which is essential especially in the context of new digital technology

introduced into the workplace. Considering these legislative amendments, the

employer may process employees' personal data if the treatment is essential for the

administration of the employment relationship, to accomplish specified

responsibilities or tasks from the business discipline, or for any other reason64. As a

result, data processing must be carried out in accordance with the principles of

privacy protection outlined in Article 5 of the GDPR. Employers must comply by

62
A. Ingrao, “Il controllo a distanza sui lavoratori e la nuova disciplina privacy: una lettura integrata”, (2018).
63
Rosa Di Meo, “Il diritto alla disconnessione nella prospettiva italiana e comparata”, Università Politecnica delle
Marche, VOL. 3, no. 2, 2017.
64
General Data Protection Regulation (GDPR) - (Articles. 6, par. 1, lett. c; 9, par. 2, lett. b and 4; art. 88)
national regulations, which "include suitable and specific measures to safeguard the

data subject’s human dignity, legitimate interests and fundamental rights, with

particular regard to the transparency of processing [...] and monitoring systems at

work place"65 in order for the processing of the relevant data to be legitimate66. As a

result, this concept requires the employer to implement a preventive organizational

structure capable of influencing device settings by design to preserve workers'

dignity. In consideration of employee data, it appears that the GDPR requirements

offer a variety of tools that may be able to mitigate some of the issues identified in

connection with the growing incidence of algorithmic management. As a result,

today's workplaces are inundated with more data on their employees than ever

before, including both professional and personal information. To be sure, an

employer's will to oversee and control an employee's activities, both within and

outside of working hours, is not something new. The concept of supervision was

most notably analyzed by Marx as a tool used by the capitalist employer to regulate

worker productivity and production processes. While companies' willingness to

monitor employee behavior is nothing new, AI technology has resulted in a

substantial increase in their capacity to do so.

This complicated reality has stimulated the interest of regulators across the world,

resulting in a variety of measures aimed at addressing one or more of the problems

to workers' rights in the AI reality. Generally, some solutions concentrate on

enhancing the existing legal framework, while others concentrate on advancing

65
GDPR., Art. 88, par. 2.
66
(Articles 6, par. 2, and 88, par. 2, GDPR)
technology to minimize these undesirable results as much as feasible. Indeed, AI is

already changing how the conventional workplace operates and, as discussed, could

have significant effects on both employee rights and society as a whole. To ensure

that AI actually promotes social well-being rather than harming it, and particularly

does not harm the most human rights within society, one approach that could be

taken is for policymakers to collaborate with representatives of workers and

businesses to develop solutions that will allow each of us to benefit from the

outcomes of AI technology.

V. A look into European and United States

International actors are nowadays trying to implement homogeneous strategies to

defend and regulate the implementation on Artificial Intelligence.

In this regard, on one side, the European Commission issued the Proposal for a

Regulation on the European approach to Artificial Intelligence on April 21, 202167,

as part of the European Strategy for AI, the first genuine European legislative

framework in this area. Concerning the Community project's principles, the proposal

is founded on two essential themes: the use of artificial intelligence while respecting

fundamental rights and European values, ensuring that the Regulation is not an

unnecessary obstacle to technological development, but rather a means of promoting

67
Proposal for a “REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL LAYING DOWN
HARMONISED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND
AMENDING CERTAIN UNION LEGISLATIVE ACTS.”COM (2021) 206 final.
it in an ethical and sustainable manner68. Placing the two objectives just mentioned

in the context of the economy’s great race for leadership means that the EU is

concretely aiming to increase citizens' and companies' trust in artificial intelligence

by simply placing its founding values at the top of the agenda, while also securing a

leading position alongside - or potentially ahead of - global competitors69.

The new compromise language addresses "high risk" AI systems as well as the

determination of the obligations and responsibilities that should be imposed on

providers of artificial intelligence systems. The fundamental feature of the

Regulation is the classification of products that use completely or partially AI

software-based on the risk of negative impact on fundamental rights, particularly

human dignity, freedom, equality, democracy, the right to non-discrimination, data

protection, health and safety70. The more the product is likely to endanger these

rights, the more stringent are the measures taken to eliminate or mitigate the negative

impact, up to the utmost measure of a total prohibition for complete incompatibility

with these rights71.

That said, Title II of the proposed Regulation establishes a list of AI techniques that

will be considered forbidden, based on a risk-based approach. It distinguishes

between uses of AI that pose an intolerable risk, a high risk, and a low or minimum

risk. In particular, applications that constitute an unacceptable high risk, such as

68
Aida Ponce Del Castillo, A law on robotics and artificial intelligence in the EU?, 2017 ETUI FORESIGHT BRIEFS
1
69
Ibidem.
70
Leonardo Parona, “Prospettive Europee e internazionali di regolazione dell´intelligenza artificiale tra principi etici,
soft law e self-regulation”.
71
Ibidem.
those that are opposed to the Union's core values and principles, ought to be

recognized as invalid a priori given that the prohibitions covers practices that have

a significant potential to “manipulate persons through subliminal techniques beyond

their consciousness or exploit vulnerabilities of specific vulnerable groups such as

children or persons with disabilities in order to materially distort their behaviour in

a manner that is likely to cause them or another person psychological or physical

harm.” 72 As already discussed, the previously enacted EU General Data Protection

Regulation already carries implication for AI technology. Employees has the right,

even at the time and place of work, to respect for his/her privacy, which implies the

confidentiality of communication. No system of surveillance or data collection may

be installed without prior notice being given to employees and to employees’

representatives. One of the most critical challenges is the transparency of AI outputs,

or the necessity to maintain the clarity and comprehensibility of AI-based systems73.

This implies, first of all, that when an individual interacts with an AI-based

application, they must be informed about it, and they must also be able to understand

the decisions made with that tool74.

Thus, it can be said that the European Union is moving toward a kind of artificial

intelligence that is human-centric and based on respecting human rights to safeguard

human dignity and autonomy.

72
Title II, Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL, Brussels,
21.4.2021 COM(2021) 206 final 2021/0106 (COD)], pp. 12 and ss.
73
Leonardo Parona, “Prospettive Europee e internazionali di regolazione dell´intelligenza artificiale tra principi etici,
soft law e self-regulation”.
74
Ibidem.
Additionally, artificial intelligence is receiving more attention in the United States.

The United States is not just one of Europe's most important competitors in the AI

race, but it also has one of the most diversified regulatory frameworks.

In other words, whereas the European Union chose an approach based primarily on

the protection of fundamental rights by imposing stringent obligations, the United

States preferred to move toward a more appealing investment system and greater

flexibility so that the rules could easily adapt to future developments75. That is why,

in contrast to European rules, the American approach is built mostly on non-binding

principles and standards. Massive financial investments, previously decided from

the Obama administration and reiterated by the Trump administration, demonstrate

the US government's unwavering commitment to addressing the issues posed by

artificial intelligence. They are concerned, in fact, with outlining the principles that

US government authorities must consider when regulating the use of AI technology

in the private sector. These include system reliability, collective engagement in the

legislative process, data quality, and the adaptability of policies to the ongoing

advancement of technology76. Furthermore, government authorities are required to

prioritizing the production of guidelines, best practices, sectoral recommendations,

and non-binding standards to which companies can voluntarily adhere77.

75
Barbara Marchetti e Leonardo Parona, “La regolazione dell’intelligenza artificiale: Stati Uniti e Unione europea alla
ricerca di un possibile equilibrio.”, (2022).
76
Ibidem
77
Ibidem.
In addition, the US Congress enacted the National AI Initiative Act on January

202178 which provides the creation of numerous commissions to decide how state

agencies can use AI technology and to research AI's possible effects on the

workforce and consumers is a major focus of many of the rules that have been

passed. Common state initiatives, including the Algorithmic Accountability act of

202279, are still pending and would go a step further and govern AI systems'

transparency and accountability when they analyze and make choices based on

consumer data. Finally, President Joe Biden has launched the new AI Bill of rights,

a project that outlines a set of five principles and practices to be followed during the

development, use, and production of automated systems80. In comparison to the

European concept, however, the United States does not appear to be taking an

organic approach to AI, at least not yet. In the United States, there is a fragmented

approach by sector and issue on AI regulation. In fact, the Congress' initiatives seek

to strike a balance between the need to encourage the development of artificial

intelligence systems and the need to control the potential risks that their application

may entail, such as discrimination based on algorithm use, breaches of consumer

privacy, and data and information security81.Congress, on the other hand, is careful

in its missions, concerned, in the negative, with avoiding the passage of legislation

that might reduce innovation, and in the positive, with moving forward with the

78
National AI Initiative Act of 2020 (DIVISION E, SEC. 5001)
79
S. 3572 - 117th Congress (2021-2022): Algorithmic Accountability Act of 2022 | Congress.gov; “Intelligenza
artificiale e regolamentazione Internazionale”, Rapporto all’attenzione del Consiglio federale, Berna, 13 Aprile, 2022.
80
Leonardo Parona, “Prospettive Europee e internazionali di regolazione dell’intelligenza artificiale tra principi etici,
soft law e self-regulation”.
81
Ibidem.
development of a framework of rules that promotes the growth of artificial

intelligence82. Furthermore, except for a few restricted hypotheses, the US “hands-

off” approach appears to preclude both ex ante control instruments such as

authorization systems, accreditation and certification, and ex post control

mechanisms, which the European regulatory framework takes into account83.

To conclude, the discipline of artificial intelligence appears to be rooted on two

distinct directions: prompt and accurate regulation comprised of conditions,

prohibitions, and explicit regulations on the one hand, and flexible guidelines and

rules approach on the other. In general, two objectives that are not necessarily

incompatible are addressed through two alternative approaches towards

advancement. Indeed, who, in fact, adopts a regulatory strategy analogous to that of

Europe prioritizes strengthening citizens' trust and ensuring that new technology

may become rooted in that population's culture as a way to enhance everyday life

rather than as a threat; on the other hand, more flexible and adjustable regulations,

as well as greater freedom for private individuals at the expense of governmental

supervision, describe situations that seek above all to increase the attraction for

investment84. The real challenge, however, seems to be to balancing and

harmonizing both approaches through regulations that are capable of not making

people feel "threatened" by technological development, but at the same time do not

hinder progress. It is still too early to predict who will succeed first or whether

82
Ibidem
83
Barbara Marchetti e Leonardo Parona, “La regolazione dell’intelligenza artificiale: Stati Uniti e Unione europea alla
ricerca di un possibile equilibrio.”, (2022).
84
Ibidem.
legislators around the world will be prepared for the upcoming developments, but it

is hoped that the competition for the title of “artificial intelligence country leader”

won't obscure the rights of those who will have to live with it in the future.

regulations on the one hand, and flexible guidelines and rules approach

VI. Conclusion

This transition, as we have seen, represent a radical change in global employment

conditions, which poses questions of safety, health, dignity, and respect for the

fundamental rights of individuals. It doesn't just concern Amazon experience (which

was used as an example), it also concerns the worldwide employment market.

We are witnessing, a real transition from a historical society to a hyper historical big

data society, which entails algorithmic management of some enterprises' growth

performance, mostly independent of human supervision. This perspective coexists

with present legal frameworks, which have been insufficient in terms of protecting

and promoting the fundamental principles of employment law.

Given the reality described the following question arises: can employees still be

sufficiently protected by their rights in the workplace in the digital age? The answer

to this question might seem challenging. We are confronted with a set of regulations

that are occasionally obsolete and do not seem to be adequate to ensure the effective

accomplishment of the protection of the individuals that are involved in the

workplace, to achieve a stable balance between the interests of the company and the

worker. Moreover, the great part of the existing legal system has been conceived

with reference to a model of organization of production that is now strongly


outdated, according to the paradigms of Fordism and, in part, post-Fordism.

Therefore, it appears inevitable that reforms must be implemented to enable the

regulatory framework to safeguard all types of workforces, in particular, those

emerging from the adoption of innovative technologies of the latest generation. The

widespread utilization of digital technologies has encouraged the expansion of

companies, especially in the logistics sector, and have changed the parameters

through which to conceive the working environment.

The concern is that the algorithmic tool, and, more broadly, artificial intelligence,

can become a prescriptive instrument with no control. As it has been described, such

algorithms primarily operate like systems to generate canons to consider the

standard to which to adapt in order to maximize performances. The same standards

are also employed by these systems to control, manage, and possibly sanction

employees. The management of the workforce is thus essentially entirely delegated

to computers as a consequence of the adoption of algorithms, which ensures that

procedures for task selection and management are more efficient from different

perspective. This has implications for worker health and safety, stemming from

workplace accidents to a lack of protection or safety procedures. As stated in the

introduction, the purpose of this research is not to place emphasis over one direction

or another, but rather to underline the issues and challenges of the phenomenon and

analyze what the legal responses at the international level are and might be. There is

no doubt that the automation of employment resulted from new technologies, a

phenomenon that has an impact on the global economy and transcends national
boundaries, needs to be regulated to safeguard employees by providing optimal

conditions in terms of rights and guarantees.

The challenge of the twenty-first century is to modernize existing laws to protect

the safety and dignity of workers, and specifically considering the time and place of

workplace for the risks associated with new tools and machinery, both with regard

to the potential psycho-social problems resulting from the stress generated by the

constant connection, the lack of social interaction, and, generally, the use of

information and communication technologies. Everything suggests that in the future

more and more aspects of our activities will be delegated to the machines. If this

were to be the scenario European, American, and international legislation would

have to promote the positive aspects while opposing the negative ones.

The feeling - so far - is that the use of algorithms increases the risk that workers are

treated as mere fungible goods at machines disposition rather than as human beings,

ending up to the point where individuals start acting like just another simple machine

throughout the production process.

In conclusion, the best and most practical approach to technology is neither overly

enthusiastic nor overly pessimistic; we must take an intermediate path that seeks the

human civilization in every context and on every innovation, looking at

technological advancement not to be charmed, but rather to understand how to use

it in an effective way.

Giovanna Proietti

Artificial Intelligence Seminar, 2022

You might also like