Algorithmic transparency and accountability

in the world of work

A mapping study into the activities of trade unions

Conducted on behalf of the

International Trade Union Confederation (ITUC)

Analysis, February 2023

 Algorithmic transparency
and accountability
in the world of work

/ TRANSPARENCY AND REGULATION

OF ALGORITHMS
IN THE WORLD OF WORK
The rapid and pervasive impact of digital technology in the world of work poses
many challenges for workers’ rights, health and safety, skills and learning. In many
instances, it is transforming work faster than at any time in history. It also brings
huge opportunities and opens new possibilities to build better and more equal
economies and societies, and to tackle existential issues including climate change
and public health.

Technological change has historically been a net generator of jobs, however this is
not necessarily guaranteed today. It will take the right policy settings, backed up by
investment and engagement with the people whose working lives are impacted, to
ensure the historical trend of net employment creation can continue. Digitalisation
and other technological advances are also affecting the character of jobs, in many
cases dramatically.

Central to embedding the potential of technology to enrich people’s lives and to

build equity and sustainability is the need for humans to be in control. Equally
important is ensuring that technology is not simply imposed on working people,
but that they have a say in the process. That is an issue of justice and autonomy
and it also helps ensure that the full benefits of innovation can be guaranteed.
Freedom of association, the right to collective bargaining and social dialogue are
fundamental to maximizing the benefits and reducing negative consequences.

While technological advances take many forms in different work settings, a com-
mon characteristic in a great many cases, affecting hundreds of millions of workers
and the public at large, is the development and deployment of algorithms. Often
seen as a modern invention, algorithms trace their origins to 9th century Persian
mathematician Al-Khwarizmi (Algoritmi). Recent developments have led to an
explosion in their use and ever-greater complexity.

For many, and not just those working for “platform” businesses or in sectors such
as transport and logistics, life at work now means acting under the instruction of

Page 2
 Algorithmic transparency
and accountability
in the world of work

algorithms that are designed to exercise often detailed control of workers in order
to extract maximum profits. This is often dehumanising work and imposes a feudal
employment relationship. Whole business models are now built on this.

Trade unions on both sides of the digital divide are focused on the transforma-
tion of work, engaging with their membership, campaigning for digital rights, doing
research, setting policies, making demands on governments and including clauses
in collective agreements that give workers greater control.

This report, prepared by AlgorithmWatch for the ITUC, looks at examples from
around the world of how unions are tackling the crucial issues of transparency
and regulation in relation to algorithms. It reveals that while there is a substantial
and growing body of trade union knowledge and policy, there is a need to build
that into practical information, advice and guidance for union representatives and
negotiators. The report includes examples of where unions are doing this, and puts
the case for such efforts to be extended and deepened.

The report should be read in the context of many governments failing to properly
implement ILO Convention 98 on the right to organise and collective bargaining.
That Convention, ratified by 168 countries, requires governments to promote col-
lective bargaining, which is very often not the case. As the ITUC Global Rights Index
shows, the fundamental rights to freedom of association and collective bargaining
are under attack, either by governments or with their compliance and support.

This report is a valuable addition to the growing body of research and literature on
the future of work and the actions of trade unions in particular. We aim to continue
this line of work and support the report’s key observation on the need to accelerate
the transformation of policy into action.

We would like to express our thanks to the lead author, Anne Mollen, and all those
whose contributions are noted at the end of the report.

Owen Tudor Matthias Spielkamp

Deputy General Secretary, Executive Director,
International Trade Union Confederation AW AlgorithmWatch gGmbH

Page 3
 Algorithmic transparency
and accountability
in the world of work

/ CONTENTS

/ EXECUTIVE SUMMARY...................................... 5 / RECOMMENDATIONS FOR

STRATEGIC ACTION ......................................... 18
/ THE RESPONSE OF TRADE UNIONS TO
AUTOMATION IN THE WORLD OF WORK..... 6 / COUNTRY DESCRIPTIONS.............................. 19
1) A
 LGORITHMIC TRANSPARENCY / AFRICA................................................................ 19
AND ACCOUNTABILITY..................................... 6
SOUTH AFRICA.................................................... 19
2) A
 UTOMATION IS PERVASIVE,
KENYA.................................................................. 19
ANSWERS LESS SO............................................ 7

3) S
 TATUS: ANALYSIS AND / THE AMERICAS.................................................. 20
CONTEMPLATION............................................. 7 ARGENTINA.......................................................... 20

4) O
 UTLOOK: FROM PRINCIPLE BRAZIL.................................................................. 20
TO ACTION........................................................ 8 CANADA............................................................... 20
/ RESEARCH AND METHODOLOGY................... 8 CHILE.................................................................... 21
UNITED STATES OF AMERICA............................ 21
/ WORK-RELATED FOCUS: WHAT ISSUES URUGUAY............................................................ 22
ARE TRADE UNIONS ADRESSING?.................. 9
/ ASIA..................................................................... 22
1) AUTOMATION IN GENERAL............................. 9
CHINA................................................................... 22
2) FUTURE OF WORK.......................................... 10
HONG KONG....................................................... 22
3) JOB LOSS AND DE-SKILLING.......................... 10
PHILIPPINES......................................................... 23
4) PLATFORM WORK........................................... 10 SINGAPORE.......................................................... 23
5) ALGORITHMIC MANAGEMENT...................... 10 THAILAND............................................................ 23
6) W
 ORKER DATA RIGHTS AND
/ EUROPE............................................................... 24
WORKPLACE SURVEILLANCE......................... 11
CZECH REPUBLIC................................................ 24
7) C
 OLLECTIVE BARGAINING AND
UNION WORK.................................................. 12 ESTONIA............................................................... 24
GERMANY............................................................ 24
/ SCOPE AND ACTORS: WHO IS TALKING HUNGARY............................................................ 25
ABOUT ALGORITHMIC TRANSPARENCY ITALY..................................................................... 25
AND ACCOUNTABILITY?................................. 12 NORWAY.............................................................. 25
/ CONTRIBUTIONS: WHAT ARE POLAND............................................................... 26
TRADE UNIONS ACHIEVING?......................... 12 RUSSIA................................................................. 26
SPAIN.................................................................... 26
1) AWARENESS AND OUTLOOK......................... 13
SWEDEN............................................................... 27
2) ANALYSIS AND DIAGNOSIS............................ 14
SWITZERLAND..................................................... 27
3) STRATEGIES AND PRINCIPLES....................... 14 UNITED KINGDOM.............................................. 28
4) P
 OLICY POSITIONS AND
REGULATORY DEMANDS............................... 14 / OCEANIA....................................................... 28
5) E
 MPOWERMENT AND ­ AUSTRALIA........................................................... 28
CAPACITY-BUILDING...................................... 15
/ RESEARCHERS.............................................. 29
6) U
 NION ORGANIZATION, CAMPAIGNS
AND COLLECTIVE ACTION.............................. 16 / INTERNATIONAL TRADE UNION
CONFEDERATION (ITUC)............................. 29
/ TRENDS: WHAT TO TAKE FROM
THE ANALYSIS?................................................. 17 / ALGORITHMWATCH.................................... 29

Page 4
 Algorithmic transparency
and accountability
in the world of work

the time to focus more heavily on the implementa-

tion of such guidelines.
Photo: Brett Jordan @unsplash

These findings demonstrate that trade unions need

to shift focus to a more practice-oriented approach
on algorithmic transparency and accountability. As dis-
cussions on platform work have been more specific,
there might be material here to learn from. But also,
it will be essential for trade unions to debunk some of
the myths about workplace automation. Even though

/ EXECUTIVE SUMMARY providers of such systems make far-reaching promises

about efficiency gains through seemingly simple cheap
technological solutions, trade unions need to ensure
Worldwide, workers and their representatives are that this must not come at the costs of employees’
being confronted with automation in the workplace interests. Workers’ representatives need to be able
– through the introduction of specific automated to make their voices heard. Automation in the work-
procedures to manage the workforce leading to new place affects conditions of employment extensively.
forms of workplace surveillance. This report looks at In accordance with the ILO Right to Organise and Col-
what impact algorithms and automation are having lective Bargaining Convention (98)1 efforts need to be
on working life, how it relates to the future of work made to undertake collective bargaining within the
and what responses trade unions are developing in frame of social dialogue. Governments should be
response to protect workers’ interests. It presents the supportive of such efforts.
findings of a global mapping exercise aimed at identi-
fying and sketching responses of trade unions – spe- This research systematically maps trade unions
cifically concerning transparency and accountability of and their activities concerning algorithmic transpar-
algorithms in the world of work. ency and accountability in work-related contexts. It
provides an overview of ongoing discussions, analyzes
Trade unions are coming together to address trends and scenarios for trade unions, and presents
algorithmic transparency and accountability and examples. These insights stimulate recommendations
are starting to respond collectively to strengthen on how to respond to challenges arising from automa-
the rights of workers. However, as the results of tion and algorithms in the world of work from the per-
this report demonstrate, more needs to be done spective of trade unions. This report contains a range
to engage in more hands-on support on the shop of findings:
floor level. Workers need broader support in chal-
lenging new inequalities and power imbalances that —— A summary of the work-related focus that trade
algorithmic management brings with it. The discus- unions currently put on algorithmic transparency
sion on and how to respond to algorithmic manage- and accountability in the world of work,
ment from a worker perspective is still in its infancy.
Algorithmic transparency and accountability are —— The scope and actors involved in the discussions
not yet a stand-alone topic on a wide scale, but are and a typology of different actions relating to
often addressed as side issues by trade unions. Even algorithmic transparency and accountability,
though trade unions acknowledge that transparency
and accountability could solve problems arising from —— An analysis of the maturity of discussions and
automation in the workplace, starting to create spe- trends across countries and trade unions, and
cific hands-on tools would go a long way to support
worker representatives. First hands-on guides and —— Country-specific descriptions of the state of
legislative advances exist. But mostly the activities of ­discussion.
trade unions still center around formulating ethical
guidelines and principles and learning about the 1 https://www.ilo.org/dyn/normlex/en/f?p=NORMLEXPUB:121
impact of automation in the world of work. It is now 00:0::NO::P12100_INSTRUMENT_ID:312243

Page 5
 Algorithmic transparency
and accountability
in the world of work

ity law passing into legislation almost a year earlier,

riders could no longer rely on the support of unions.
Under the new law, the few unions in Hong Kong had
been facing tougher conditions, and the Hong Kong
Confederation of Trade Unions and its affiliates were
forced to cease operations altogether. Riders mostly
had to take things into their own hands. Using mes-
senger services, they discussed the everyday cruelty
of how their work was controlled by algorithms. Then,
Photo: Olga Serjantu @unsplash

they got organized.

As they could not rely on an established trade union

structure, they used the aforementioned messenger
group to coordinate protests and strikes. This slowly
progressed until a core group of riders set out a list
of demands and negotiated them with management
at Foodpanda. This eventually forced the company
to agree to some commitments. In this report, we

/ THE RESPONSE OF present our systematic research into how trade

unions and other initiatives have started address-
TRADE UNIONS TO ing such issues around algorithmic transparency
and accountability in the world of work. As did the
AUTOMATION IN THE delivery riders at Foodpanda: They demanded more

WORLD OF WORK transparency on how the algorithms calculate dis-

tances, determine service fees, and how the company
automatically suspend riders’ accounts.

1) ALGORITHMIC TRANSPARENCY Generally speaking, demanding transparency and

AND ACCOUNTABILITY accountability of algorithms is often the first response
of workers when they are confronted with a negative
In November 2021, riders for Foodpanda, a delivery impact from an automated decision-making (ADM)
platform in Hong Kong, were fed up.2 What started system. This appears to be a reaction to the opaque-
as a group on the Telegram messaging app to let off ness that often comes with ADM systems – and it
steam about deteriorating work conditions, automatic reflects the power imbalance between those affected
account suspensions, reduced earnings, inaccurate by ADM systems and those implementing them.
journey distance predictions, and ghost orders soon
developed into a fully-fledged strike. Riders were frus- To strengthen, synergize and strategize the manifold
trated with the way algorithmic management was used activities in the field, this report builds upon system-
by Foodpanda and with how the company responded atic research on the activities of trade unions related
to complaints about it. With the Hong Kong secur- to algorithmic transparency and accountability in the
world of work. Our analysis shows that the implic-
2 https://www.scmp.com/news/hong-kong/hong-kong- ations of automation are apparent across con-
economy/article/3156590/foodpanda-hong-kong-couriers-
tinents, countries, industries, sectors, and at all
end-labour-dispute?module=inline&pgtype=article
employment levels; however, trade unions must
https://www.scmp.com/news/hong-kong/hong-kong-
economy/article/3157373/foodpanda-drivers-strike-look- move more beyond discussion and analysis of the
behind-scenes-2-day problem to concrete advocacy relating to ADM sys-

https://www.scmp.com/news/hong-kong/hong-kong- tems in the workplace.
economy/article/3157373/foodpanda-drivers-strike-look-
behind-scenes-2-day
 ttps://lausan.hk/2021/foodpanda-strike-how-do-workers-
h
In this report, we describe what trade unions have so
organize/ far been doing about algorithmic transparency and

Page 6
 Algorithmic transparency
and accountability
in the world of work

accountability. Furthermore, we analyze the common- mation in single-pilot flights, respectively. While these
alities and possible further activities in the responses examples illustrate specific concerns about auto-
of trade unions, and we propose ways for trade uni- mation, what needs to be conducted are analyses
ons to strengthen the rights of workers in the face of on how to tackle them – for instance, by address-
ADM systems in the world of work. ing algorithmic transparency and accountabil-
ity instead of simply rejecting the technology
entirely.
2) AUTOMATION IS PERVASIVE,
ANSWERS LESS SO Be it healthcare, education, logistics, retail – automa-
tion is impacting almost every sector and it affects
Trade unions started addressing algorithmic transpar- workers in a plethora of situations. Based on our
ency and accountability from 2017/2018 onwards and analysis, we can state that understanding is growing,
there was a big increase in interest between 2020 and and more nuanced problem definitions of the con-
2022. During that time, the discussion has moved from sequences of ADM systems for workers are emer-
abstract fears about automation and concerns about ging. However, algorithmic transparency and
jobs, to more detailed contemplations. Trade unions accountability have not yet been addressed in an
now need to transfer these mostly theoretical dis- implementation-oriented way on a large-scale.
cussions into concrete tools for workers to imple- But some first initiatives by trade unions are
ment whenever they are confronted with concerns promising. Hands-on practical guides, for instance
on automation in the world of work. Prospect’s (UK) “Digital technology – Guide for union
representatives”3, the Trades Union Congress’ (UK)
For example, the aforementioned platform work- “People-Powered Technology”4 publication on collect-
ers at Foodpanda in Hong Kong came up with spe- ive agreements and digital management systems or
cific responses to their immediate concerns. They Comisiones Obreras’ (Spain) guide on “Negociación
launched a Facebook group to document their injur- colectiva y digitalización”5, can be considered a starting
ies. This acted as a monitoring system that generated point for directly addressing worker representatives
knowledge on the situation of gig workers in Hong and how they can ensure that workers’ interests are
Kong to use in negotiations. Other platform workers, safeguarded when ADM systems become implemen-
mostly in Latin America, started the “Mi Jefe es Una ted in their respective work environment. Such initi-
App” campaign to support legislative proposals aimed atives can build the basis for giving workers and
to strengthen the rights of platform workers. Here, we their representatives guidance on how to address
see trade unions and similar collectives moving for- algorithmic accountability and transparency – as
ward and making an impact. well as orientation for regulatory approaches that
are equally needed. More such hands-on tools are
While the discussion on platform workers is compar- needed.
ably quite advanced, debates in other sectors related
to algorithms and the world of work now need to
move forward. On a general scale with concerns 3) STATUS: ANALYSIS AND
about automation emerging, these discussions have CONTEMPLATION
yet to dwell on the specifics of algorithmic transpar-
ency and accountability. For example, the Screen But as our study shows, the current status is still
Actors Guild – American Federation of Television and largely marked by analyses and contemplations. Trade
Radio Artists (SAG-AFTRA) has been concerned about unions are currently focusing on raising awareness,
deep fake technology as it raises ethical questions gathering knowledge and sketching scenarios about
and it could have a severe impact on entertainers,
actors, and musicians. Trade unions in the shipping 3 https://prospect.org.uk/about/digital-technology-guide-for-
union-reps/
and aviation industries (e. g. The International Long-
4  ttps://www.tuc.org.uk/sites/default/files/2022-08/People-
h
shoremen’s Association or The International Air Line Powered_Technology_2022_Report_AW.pdf
Pilots Association) are concerned about safety issues 5  ttps://www.ccoo.es/cms/cli/000001/o/af/
h
related to autonomous container cargo ships or auto- af6e35ab004a61334480e3b2bcae0e93000001.pdf

Page 7
 Algorithmic transparency
and accountability
in the world of work

what automation might bring for their specific sector step for trade unions. There is an urgent need to
– both in countries where the discussion is only start- move from principles and theoretical discussions
ing as well as in countries, where it has been going on to implementing these principles in practice. Trade
for a number of years. More advanced discussions unions should now demonstrate how workers can be
focus on formulating roadmaps and conceptualizing empowered to articulate their interests with regard to
strategies on how to react toward automation in the specific ADM systems and how they find opportunit-
world of work through transparency and accountab- ies to advocate for these. An inter-union exchange
ility or envisioning different steps to get there. What of effective strategies could also be an import-
is lacking on a large-scale are policy recommend- ant tool that until now has not been fully taken
ations, capacity-building initiatives, campaigns or advantage of.
even concrete bargaining results. Broadly speaking,
the discussion is currently characterized by research Further, the Right to Organize and Collective Bargain-
and exchange of information. It now needs to develop ing Convention (No. 98) states explicitly in Article 4
into advocating for workers’ interests. that all ratifying countries need to take measures in
accordance with national conditions in order to pro-
mote negotiations between employer’s and workers’
4) OUTLOOK: FROM PRINCIPLE organizations “with a view to the regulation of terms
TO ACTION and conditions of employment by means of collective
agreements”9. Algorithmic management touches
To summarize, the general discussion still remains in its core the terms and conditions of employ-
quite broad, and abstract fears about the con- ment. While a recent study finds that currently one
sequences of automation in the workplace tend to in three workers in Europe does not even know if
dominate. Trade unions should now formulate algorithmic management systems are being used
arguments on how best to address algorithmic on them10, it is high time that also governments set
transparency and accountability to empower the right frames to support social dialogue and col-
workers when they are confronted with specific lective bargaining on automation in the workplace.
ADM systems. What is needed are specific solu- In the following, we present details of our study.
tions that achieve algorithmic transparency and
accountability on a broad scale.

The most important contribution of trade unions, and

/ RESEARCH AND
a direct outcome of the many past and ongoing con-
templations and analyses, are principles or guidelines
METHODOLOGY
to be followed when implementing ADM systems
in the world of work. These guidelines often include This mapping study involved the following three steps:
such principles as transparency, fairness, privacy and
data protection, a right to be informed, human over- 1. Systematic inductive identification of activities
sight, correspondence with fundamental rights, high- that trade unions engage in regarding algorithmic
risk regulations, explainability, etc. (see for instance transparency and accountability in the world of
Syndicom – Switzerland6, the Trades Union Congress work.
(TUC) manifesto – UK7, AIITEU – India8 etc.).
2. Detailed descriptions of activities, including cat-
How these principles are implemented is the most rel- egorizations of their work-related focus, scope,
evant question to address, and it should be the next and key contributions for trade unions’ advocacy
work.
6  ttps://syndicom.ch/fileadmin/user_upload/Web/Website/
h
Branchen/ICT/2019_11_01_d_ICT-Sektor_Resolution_KI-
Gebote_def.pdf
9 
https://www.ilo.org/dyn/normlex/en/f?p=NORMLEXPUB:121
7 https://www.tuc.org.uk/research-analysis/reports/dignity- 00:0::NO::P12100_INSTRUMENT_ID:312243
work-and-ai-revolution
10 
https://www.uni-europa.org/news/algorithmic-management-
8 https://www.aiiteu.org/letters/bpo-workers-demand-charter/ 1-in-3-workers-left-in-the-dark/

Page 8
 Algorithmic transparency
and accountability
in the world of work

3. Analysis of the collected data set – focusing on ation (e. g., alternative media, news reporting, blog
the state of discussion in each country and on posts) were also included in the analysis.
key contributions and trends related to the work
of trade unions on algorithmic transparency and
accountability.
/ WORK-RELATED
Our analysis looked at 26 countries across the globe,
FOCUS: WHAT ISSUES
and it identified over 250 activities by trade unions ARE TRADE UNIONS
ADRESSING?
related to automation in the world of work – includ-
ing specific references to algorithmic transparency
and accountability in some cases, while other refer-
ences remained vague. The research was carried out As the current and future impact of ADM systems and
between December 2021 and March 2022, with select- algorithms in the world of work is so diverse and mani-
ive additions to the data set until September 2022 fold, we differentiated in a first step what trade unions
when high-profile new publications on algorithmic focus on when engaging with the issues. It is import-
transparency and accountability by trade unions ant to note that, as of now, questions of transparency
emerged. Most activities that relate to algorithmic and accountability of algorithms in the world of work
transparency and accountability are fairly recent. The often appear as one element in a broader context.
earliest date back to 2017, but the majority of the res- Many activities of trade unions focus on more general
ults are from 2020, 2021 and 2022. questions, which also touch on algorithmic transpar-
ency and accountability. In that regard, we identified
The research was predominantly conducted online the following topics and themes that appear on the
and was carried out by researchers with local expert- agendas of trade unions.
ise and/or contacts, who scanned publications, events
and reports for relevant information on activities
of trade unions related to algorithmic transparency

Photo: Rens Dimmendaal / Better Images

and accountability in the world of work. This online
research was partly complemented by interviews with

of AI / Man reversed / CC-BY 4.0

local experts.

The idea of this mapping exercise was not to

quantify the activities of trade unions, but rather
to identify the broadest possible range of activities
on algorithmic transparency and accountability
in the world of work that trade unions engage in,
and then subject these to a systematic and trend
sketching analysis. The research does not claim to
be representative. It provides selective insights into 1) AUTOMATION IN GENERAL
the activities of trade unions related to algorithmic
transparency and accountability. Broadly speaking, trade unions across this sample
are grappling with digitization, automation, ADM,
With such a diverse sample of countries, trade unions algorithms and Artificial Intelligence (AI) in the world
and activities included, the data collection process of work at a very general level. Furthermore, they tend
needs to be adjusted for different settings. Online to discuss what role automation and algorithms are
research works best for countries where there is a already having on workers in an abstract sense and
high degree of union organizing with professional give broad examples to illustrate the issue. By explain-
structures and sufficient funding. In countries with ing the relevance of algorithms and ADM systems for
less institutionalized structures for trade unions, local workers, the workplace and workers’ rights advoc-
contacts, and other more informal sources of inform- ates, and by voicing concerns, these activities set the

Page 9
 Algorithmic transparency
and accountability
in the world of work

agenda and justify why trade unions should be deal- 4) PLATFORM WORK
ing with the issue. Fundamental concerns and ethical
questions regarding algorithms and ADM might be Platform work is a very prominent issue. The activit-
addressed. However, the majority of these activities ies in this sample pertaining to platform work mostly
voice concerns about the impact of algorithms in the concern gig work, but also address crowd work or
world of work. Occasionally, it is also suggested that cloud work. Platform work is intrinsically related to
workers might benefit from their implementation. algorithmic transparency and exemplifies a very con-
crete, current and palpable example of what impacts
algorithms and ADM systems can have on workers
2) FUTURE OF WORK and their rights. The activities in this study articu-
late specific platform work-related concerns. For
By sketching specific outcomes and scenarios of what instance, on the opacity of algorithmic management
algorithms and ADM systems hold in store for work- (e. g. algorithmic allocation of gigs, calculations of
ers, activities that focus on the future of work move routes, closure of accounts etc.). This study also looks
beyond arguing about the relevance of the issue. at activities that exemplify how workers represent
These activities are usually based on a more stra- their interests against platforms or sketch solutions
tegic and analytical approach to understand what for platform work-related concerns. Specific solutions
challenges and concerns workers and trade unions proposed by worker representatives that address
need to worry about now and in the future. As a con- transparency and accountability are represented
sequence, they are often the prerequisite for trade alongside hypothetical solutions.
unions to articulate more strategic responses. While
these activities operate at a general level, they can
provide the cornerstone for developing concrete solu- 5) ALGORITHMIC MANAGEMENT
tions for prospective problems.
Several activities that trade unions engage in concern
how algorithms and ADM systems are, or might, be
overtaking managerial decisions. Some of these activ-
ities are so far hypothetical, for example imagining
the replacement of human resources departments
with algorithmic decision-making. However, some are
Photo: Alex Kotliarskyi @unsplash

also concerned with specific current issues, such as

the uptake of ADM systems in recruiting processes
or in People Analytics procedures, where the lack of
transparency and worker involvement is a major con-
cern. Some concerns relate to bias and discrimination
in these automated decisions. In addition, trade uni-
ons also address legal implications and limitations –
for example, the dismissal of workers through solely
automated decisions.
3) JOB LOSS AND DE-SKILLING
In contrast to more general discussions on automa-
One specific scenario, that quite a few trade unions tion and the future of work, here specific – albeit
are sketching, focuses on job losses due to algorithms sometimes hypothetical – cases, are under discus-
and automation. Such activities often also relate to sion. Where problems have been identified, the lack
those focused on the “future of work”. But due to their of solutions is being addressed. These activities are an
prominence in the sample, they qualify as a separate important step in clearly outlining concerns to which
theme. Alongside postulating that people might lose specific solutions need to be found, for instance,
their jobs due to automation, issues of de-skilling and through regulatory proposals or co-determina-
the need to retrain workers for new tasks are also tion processes. In these activities, trade unions
being addressed. identify their analytical needs regarding algorithmic

Page 10
 Algorithmic transparency
and accountability
in the world of work

Algorithmic Management 37
Collective Bargaining / Co-Determination / Union work 42
Worker Data Rights 45
Future of Work 52
Job Loss / Dequalification / Re-Training 58

Automatation / Algorithms / ADM / AI

99
Platform work
101
Image 1: Distribution of work-related focus across entire sample in real numbers.
The visualization shows that general and abstract contemplations on automation in general are dominant in the
sample, as well as a very specific focus on platform work – an area where the discussion of algorithmic transpar-
ency and accountability has progressed furthest. Further general contemplations on “Future of Work” or fears about
“Job Loss/Dequalificiation/Re-Training” also appear in the discussion, in slightly higher numbers than topics such as
“Worker Data Rights” or “Algorithmic Management”.

transparency and accountability in order to pro- on workers are collected and used – and the lack of
duce strategies: How do automated systems take information on this – is potentially a very problematic
decisions? What are the most relevant decision-mak- area.
ing criteria? What models of learning are used? How
are databases obtained and used? How do these pro- Related activities of trade unions vary in how
cesses relate to labor law and protection of worker abstract they are. At one end of the spectrum,
rights? The diagnosis is often a lack of transparency in some trade unions tackle broad concerns, such as
algorithmic decision-making and the declaration that raising awareness about workplace surveillance,
more transparency is needed. Such questions and arguing that data on workers needs to be protected,
conclusions should be further built upon in order to or aiming to find out what levels of privacy exist in
establish specific solutions on how to reach transpar- the workplace. At the other end, we see trade uni-
ency and accountability. ons already contributing specific answers to privacy
threats in the workplace – for example, in the for-
mulation of workers‘ privacy rights, data protec-
6) WORKER DATA RIGHTS AND tion charters, data protection impact assessments
WORKPLACE SURVEILLANCE (DPIA)11 or data stewards12 as frameworks for organiz-
ing and collective bargaining. Such solutions need to
Opaque algorithmic management, worker privacy, gain more traction and be applied at scale.
data protection, and workplace surveillance are key
concerns related to bias and discrimination. Data are 11 
https://d28j9ucj9uj44t.cloudfront.net/uploads/2020/12/
the foundation on which algorithmic management prospect-dpia-workers-guide.pdf

decisions are taken. How databases with information 12 https://library.fes.de/pdf-files/bueros/singapur/17381.pdf

Page 11
 Algorithmic transparency
and accountability
in the world of work

7) COLLECTIVE BARGAINING AND ant actors that we identified in our study. That is why
UNION WORK we mostly base our analysis on their activities. There
can be slight differences between the activities that
Several activities of trade unions reflect on the impact trade unions and trade union confederations engage
that algorithms and ADM systems have on union in. In part, trade unions are more likely to be specific
work, the representation of worker interests and when voicing their concerns about transparency and
related advocacy work. This is done in a general sense. accountability regarding ADM systems and algorithms.
For example, by addressing what the future of work In comparison, trade union confederations usually
means for bargaining processes, what role unions can address the same issues in the workplace at a more
play in creating algorithmic accountability and what general level. They tend to identify and analyze under-
obstacles might lie ahead. Trade unions are also trying lying and systematic challenges to workers’ rights and
to produce knowledge on the status of co-determina- propose more encompassing and transferable solu-
tion and bargaining power for worker representatives tions.
in current implementations of ADM systems (e. g.,
through surveys among workers and representatives). Next to trade unions, new collectives are also enter-
ing the scene, often in response to platform work. As
More concrete tools that aid union representatives a result, union building is emerging as a trend. These
and negotiators in collective bargaining processes collectives and new unions, which are often related
come in the form of recommendations for social dia- to gig work, are gaining traction, even if they aren’t
logue, guiding principles as an orientation in nego- fully institutionalized. They are a direct response to
tiations or specific checklists for negotiations13. For challenges that arise for workers through algorithmic
example, detailing worker interests regarding privacy, management and platform work.
skills development, data security, etc. when workplace
agreements need to be developed in case ADM sys- Other entities with an interest in worker rights
tems are implemented in specific work settings. Next related to algorithmic transparency and accountabil-
to a check list, one US union even provided a model ity are foundations, non-governmental organizations,
scenario of how digital workers in Silicon Valley could research institutes, and groups or councils installed
start to unionize14. through government bodies. Where relevant, also
in some countries due to a lack of union organizing
or lack of union resources to engage with issues of

/ SCOPE AND ACTORS: algorithmic management, we have also included the

activities of such entities as part of our analysis.
WHO IS TALKING
ABOUT ALGORITHMIC / CONTRIBUTIONS:
TRANSPARENCY AND WHAT ARE TRADE
ACCOUNTABILITY? UNIONS ACHIEVING?
Different organizations, entities and collectives engage As can be seen in the chart below (see Image 2) trade
in activities on algorithmic transparency and account- unions so far to a large extent deal with ADM in a
ability in the world of work. Traditional labor unions reflective manner – with contemplations, principles
and union confederations are by far the most import- and strategies to, among other things, attain trans-
parency and accountability. While we now see more
13 h
 ttps://innovation-gute-arbeit.verdi.de/themen/digitale- implementation-oriented guidelines and practical
arbeit/++co++c152781e-5866-11ea-b06f-525400f67940 / tools emerging, most trade unions still need to take
https://www.fagforbundet.no/a/280738/for-tillitsvalgte/
omstilling/veiledere-i-omstillingsprosesser/veileder--digital- the next step and develop as well as execute ideas for
omstilling/ implementations of policies, collective bargaining, as
14 https://www.code-cwa.org/ well as for empowering workers and their represent-

Page 12
 Algorithmic transparency
and accountability
in the world of work

Empowerment / Training / Capacity Building

Union organizing / Campaigning / Collective action / Call to Action

Policy Position / Recommendations / Regulatory Demands

Strategy / Roadmap / Principles

Awareness / Scenario Sketching

Analysis / Diagnosis / Knowledge Gathering

0% 5% 10 % 15 % 20 % 25 %

Image 2: Percentage distribution of the contributions that trade unions achieve in their
activities on algorithmic transparency and accountability in work-related contexts,.
Reflective contributions dominate in this sample. The highest numbers in this sample can be identified as contribu-
tions on doing analyses, raising awareness, and formulating strategies and principles. More concrete and implementa-
tion-oriented contributions like formulating policy position, campaigning on algorithmic transparency, and accountab-
ility or empowering worker and worker representatives through specific capacity-building initiatives happen at a much
lesser scale. There is a need to promote the more hands-on implementations over the more reflective contributions.

atives. These efforts can be advanced by engaging —— Country: Argentina

more in empowerment, capacity-building, cam- —— Trade union: Friedrich-Ebert Stiftung
paigning, advocacy, and formulating regulatory —— Type of Activity: Podcast
demands. —— Title: ¿Sueñan los androides con obrerxs
electricxs?
—— Description: The Friedrich-Ebert Stiftung in
1) AWARENESS AND OUTLOOK Argentina publishes the regular podcast series
¿Sueñan los androides con obrerxs electricxs?. It
Trade unions engage in many activities to raise provides information about how the world of
awareness about automation in the world of work. work is changing with digitization, and it also
Questions about transparency and accountability ­covers questions about automation, transpar-
are included in these activities and often relate to ency and accountability. It aims to strengthen
the future of work. Also included are questions about the ability of trade unions to address digitization,
what will be expected with increasing automation in including automation, and help with capaci-
work-related contexts, the predicted challenges, what ty-building and increasing resources. The orga-
trade unions should focus on, etc. In that way, trade nization recognizes that there is a digital gap
unions are researching the situation on a very large in the unions that does not allow them, for the
scale. They organize conferences and workshops, moment, to embrace discussions on algorithmic
publish reports and engage with many different stake- transparency, so work is needed to strengthen
holders to explain what is at stake. To broaden the dis- the sector to achieve this.
cussion, they also reach out by writing blog posts and —— Contribution: Awareness/outlook
newsletters, making podcasts, and organizing inform- —— URL: https://argentina.fes.de/e/podcast-
ation campaigns. suenan-los-androides-con-obrerxs-electricxs/

Page 13
 Algorithmic transparency
and accountability
in the world of work

2) ANALYSIS AND DIAGNOSIS —— Country: Norway

—— Trade union: The Norwegian Society of Engin-
More specific research endeavors go beyond simply eers and Technologists and others
describing the situation. Under the umbrella of ‘ana- —— Type of Activity: Joint declaration
lysis and diagnosis’, trade unions gather knowledge —— Title: Erklæring om ansvarlig bruk av kunstig
and experience about automation from workers. intelligens i arbeidslivet
These analytical and diagnostic activities address —— Description: Several trade unions and enter-
experiences with automation in the world of work, prises addressed a joint declaration on the
bring actors with different expertise together, and responsible use of AI in the workplace at the Nor-
connect groups from fields such as academia, civil wegian Minister of Digitization, Nikolai Astrup.
society, governments, etc. Gathering knowledge and The declaration – signed by the trade union for
diagnosing where problems lie with algorithmic trans- private sector employees, Negotia, together with
parency and accountability are crucial steps in finding Telenor Group and Norwegian companies such
answers to specific concerns around automation in as IBM Norway and Schibsted Media Group –
work-related contexts. focused on data quality and the importance of
transparency and accountability in ADM systems.
—— Country: United States of America It states that ADM processes must be com-
—— Trade union: American Federation of Labor and prehensible and individuals retain the right to
Congress of Industrial Organizations challenge the accuracy of a decision. In addition,
—— Type of Activity: Launch of a research institute responsibilities must be defined, and the com-
—— Title: Technology Institute panies that design, sell or use AI are responsible
—— Description: In the US, the American Federation for the decision-making process and results of
of Labor and Congress of Industrial Organiza- those decisions. Relevant control mechanisms
tions met the need for analysis by launching a must be in place to offer a sufficient degree of
technology institute. The institute aims to invest- transparency, and the privacy of employees
igate the problems posed by technology (includ- must be safeguarded. Arrangements should also
ing automation) in the workplace. It wants to give be made for internal risk reporting, and routines
a voice to working people at the forefront of the implemented for measures and follow-ups.
conversation about the future of work. —— Contribution: Strategies/principles
—— Contribution: Analysis/diagnosis —— URL: https://www.nito.no/contentas-
—— URL: https://aflcio.org/press/releases/ sets/72800175dcc745d98e328f35eb72f1f2/
afl-cio-launches-technology-institute ansvarlig-bruk-av-kunstig-intelligens-i-arbeid-
slivet-erklaring.pdf

3) STRATEGIES AND PRINCIPLES

4) POLICY POSITIONS AND
Gaining knowledge and insight are often the pre- REGULATORY DEMANDS
requisites to planning strategic actions. Trade unions
focus on what needs doing at a general level and what Trade unions have recently started to put forward
general principles should guide future action. More more detailed policy recommendations and demands
often than not, ADM systems are simply rejected on how to react to the increasing use of automation in
and, in order to advocate for the interests of workers, the workplace. More such policy positions are urgently
trade unions commonly grant no space to negotiate needed and especially more advocacy by trade uni-
about these systems. As a consequence, transpar- ons to see regulatory frameworks to safeguard work-
ency and accountability are not even considered ers' interests succeeding. Still so far trade unions
options for pursuing worker interests. have mostly been calling quite broadly for the need
In other cases, we however see quite detailed prin- to regulate ADM systems in the workplace – without
ciples on how to deal with ADM systems in the world providing specific ideas – or they have been reacting
of work. Examples can be found in many different to legislative proposals put on the table by regulatory
countries and trade unions. bodies. In the process, trade unions are now starting

Page 14
 Algorithmic transparency
and accountability
in the world of work

to produce more detailed policy positions – specifically

geared towards addressing the challenges of automa-

Photo: Javier Allegue Barros @unsplash

tion in the workplace in broader frameworks to reg-
ulate Artificial Intelligence. For instance, in Europe,
several trade unions reacted to the European Union’s
proposed AI Act by focusing on labor issues and ques-
tions around transparency and accountability. And in
Latin America, we see reactions to legislative propos-
als regarding the rights of platform workers.

As trade unions react, further policy demands and

recommendations emerge in response to specific
problems. Also, as governments in many countries resources, there are also concrete technical tools that
work on so-called ‚national AI strategies‘, trade unions support workers in their advocacy work. All these
have started to address how these strategies impact empowerment and capacity-building initiatives help
employee rights. Unions should cease the opportun- support worker representatives when they are con-
ity to make specific demands as these policies are still fronted with a specific ADM-related process.
being rolled out to concretize their policy goals.
Some training courses run by trade unions could be
—— Country: Philippines more practical and informative instead of being the-
—— Trade union: Kapatiran ng Dalawang Gulong ory based. This would help workers advocate for their
(Trade Union) own interests in respect to ADM systems15. Other
—— Type of Activity: Position statement capacity-building measures assist workers and their
—— Title: Food delivery riders are employees, not representatives in articulating demands regarding
freelancers ADM systems at all stages of development and imple-
—— Description: Riders’ advocacy group Kapatiran mentation. For example, by spelling out which precise
ng Dalawang Gulong comments on how delivery transparency measures are needed so that workers
platform workers are confronted by algorithmic can gain some oversight on algorithms.
management systems that lack transparency.
Platform workers want to change their status —— Country: Spain
from independent contractors to employees. —— Trade union: Comisiones Obreras, Trade Union
This would support them in their demand for —— Type of Activity: Guide
more transparency on how algorithms manage —— Title: Guía negociación colectiva y digitalización
their gigs. The same demand can be seen in 2020
many other countries, and a few laws are now in —— Description: This guide helps collective bargain-
place (e. g., the Riders’ law in Spain). ing when dealing with the digital transformation
—— Contribution: Policy positions/regulatory of work. The guide calls for workers to participate
demands directly in the implementation of ADM systems. It
—— URL: https://opinion.inquirer.net/142783/ also calls for the regulation of data collection and
food-delivery-riders-are-employees-not-free- processing through collective bargaining and to
lancers give workers the “right to explanation” when AI is
used in HR departments. Workers should also be
able to access, manage, and control any data that
5) EMPOWERMENT AND ­ AI and algorithms generate about them.
CAPACITY-BUILDING —— Contribution: Empowerment and capacity-­
building
More implementation-oriented contributions include
educational resources to enable workers and their
15 https://anejudchile.cl/2020/10/13/escuela-sindical-luis-
representatives to pursue their interests when con- cerda-castro-de-anejud-invitan-a-webinar-teletrabajo-
fronted with ADM systems. Apart from educational automatizacion-e-i-a-en-el-poder-judicial/

Page 15
 Algorithmic transparency
and accountability
in the world of work

—— URL: https://www.ccoo.es/cms/cli/000001/o/af/
af6e35ab004a61334480e3b2bcae0e93000001.
pdf

Photo: Stephen Dawson @unsplash

The Foodpanda Facebook group previously mentioned,
where gig workers in Hong Kong collected information
on their working conditions, including work-related
injuries, demonstrates how technology can support
workers in representing their own interests. The Fair-
work initiative equally offers several tools and evalu-
ative frameworks to help assess the algorithmic work
landscape in specific local contexts. These tools are
supposed to successfully assist workers and trade uni- sides agreed that riders will be treated as employ-
ons by advocating for their interests16. ees. The agreement aims at ensuring that salaries,
safety and rights are better protected17. In other cases,
—— Country: Australia established trade unions have assisted self-organized
—— Trade union: United Workers Union worker groups from the platform economy in advoc-
—— Type of Activity: App ating for better transparency and accountability of
—— Title: Unions mobilize AI to turn the tables on ADM systems within their respective platforms – for
wage theft in hospitality instance in Hong Kong18, China19 or India20.
—— Description: Unions are using AI to mobilize
younger workers against wage theft. The union It remains important to acknowledge meta effects of
announced the development of an app that uses automation on the workforce – as can be seen in Chile,
AI to answer workers’ questions about wage where there have been ongoing strikes at Walmart’s
theft and working conditions among those who Líder supermarket company over the introduction of
work in hospitality. automated checkouts and robots. Unions are asking
—— Contribution: Empowerment and capacity-­ the company to be transparent and take responsibil-
building ity for massive layoffs and the effect of automation on
—— URL: https://unitedworkers.org.au/media-re- the labor market21.
lease/unions-mobilise-ai-to-turn-the-tables-on-
wage-theft-in-hospitality/ But there is quite a bit to learn from the platform eco-
nomy, where workers are already able to voice specific
demands. While there remain overarching red lines
6) UNION ORGANIZATION, regarding the use of automation, trade unions need
CAMPAIGNS AND COLLECTIVE to find specific answers. Automation is not only about
job loss and dequalification, but especially about shift-
ACTION ing power relations. Here, it becomes necessary to
As previously described, the most prominent strengthen worker negotiating power in algorithmic
examples of successful union organization, cam- decision-making processes.
paigns and calls to action so far stem from platform
workers. Across continents, we can see platform 17 h
 ttps://www.nidil.cgil.it/rider-accordo-tra-cgil-cisl-uil-e-just-
eat/
workers gain more transparent wage calculations by
18 h
 ttps://www.scmp.com/news/hong-kong/hong-kong-
algorithms, less arbitrary account suspensions and
economy/article/3157373/foodpanda-drivers-strike-look-
even a strengthened contractual position, by acknow- behind-scenes-2-day
ledging their status as employees and not freelancers. 19 h
 ttps://clb.org.hk/content/acftu-beijing-issues-guidelines-
For example, in Italy, in 2021, we saw an agreement unionizing-gig-workers-what-comes-next

between three main Italian trade unions and the 20 https://www.thenewsminute.com/article/transport-workers-

union-approaches-sc-seeks-social-security-benefits-gig-
online food delivery service JustEat, in which both workers-155624
21 h
 ttps://www.lemondediplomatique.cl/2019/08/la-otra-
16 https://fair.work/en/fw/engage/resources-and-tools/ batalla-de-los-trabajadores-del-gigante-del-retail.html

Page 16
 Algorithmic transparency
and accountability
in the world of work

—— Country: Spain Transparency is still the problem, not the solution

—— Trade union: Asociacion Española de la Banca, Generally speaking, the focus so far is on problems
CCOO, UGT. and not as much on solutions. Across this sample, it
—— Type of Activity: Collective bargaining agree- is apparent that transparency of ADM in the world of
ment work, or the lack of it, is a major problem. That is why
—— Title: Convenio de Banca trade unions need to move forward by providing spe-
—— Description: This latest revision of the collective cific ideas on how transparency could be improved in
bargaining agreement in the Spanish banking a hands-on way.
sector includes the right of workers not to be
subject to decisions solely based on algorithms Threshold for campaigning not passed
and not to be discriminated against based on Except in the case of platform workers, we largely see
algorithmic decisions. Companies are obliged to that the threshold for campaigning does not seem
inform workers’ representatives about the data to have been reached. There are many abstract and
and algorithmic systems used by “digital models almost theoretical concerns about what automa-
with no human intervention”. Reportedly, this is tion might mean for workers, and transparency and
the first time such rights have been written into accountability measures might be called for in order
a collective bargaining agreement in Spain. The to address these. However, there is a lack of shop
agreement also recognizes workers’ rights to floor experience of ADM systems that would support
digital disconnection and privacy. broader campaigns.
—— Contribution: Union organizing, campaigning
and collective action New initiatives have taken the lead in the form of
—— URL: https://www.federacionfine.es/recursos/ groups created by the platform worker community,
secciones/FINE/documentos/20210422_RESU- organized from the bottom-up. Their grievances come
MEN_CONVENIO_BANCA.pdf from their lived experiences with algorithmic manage-
ment and the lack of transparency and accountability
Across this sample we could not identify many specific of ADM systems. Including unions could bring struc-
successes, such as collective bargaining agreements ture and experience to these new passionate move-
that explicitly reference algorithmic transparency ments and help them achieve their goals.
and accountability. However, as the topic increasingly
appears on the agenda, more examples might follow Differing attributions of relevance
soon. Raising awareness is an important activity that trade
unions should be engaged in. There seems to be a dif-
ference in attributing relevance to the issue of auto-

/ TRENDS: WHAT mation in the world of work between union leaders

and workers22. This is possibly due to a lack of direct
TO TAKE FROM THE experience of and knowledge about automation on
the shop floor. Thus, information and awareness cam-
ANALYSIS? paigns are key.

Algorithmic transparency and accountability are However, feedback from trade union representatives
not yet stand-alone topics in our study suggests that – even for trade unions –
Algorithmic transparency and accountability are so far automation often needs to be de-prioritized as other,
addressed by trade unions largely under the umbrella presumably more urgent, matters have to be dealt
of broader discussions on automation and digitiz- with. While the general importance of automation for
ation. Analysis and advocacy could be improved by worker rights is recognized, there is in some cases
addressing transparency and accountability as indi- simply a lack of resources on the side of trade uni-
vidual key issues, which would lead to more detailed ons to deal with it. This is supported by the finding
and nuanced solutions and implementation-oriented
approaches. 22 h
 ttps://aflcio.org/2017/10/24/highlights-day-three-afl-cio-
2017-convention

Page 17
 Algorithmic transparency
and accountability
in the world of work

that where trade unions are very active on the topic, Focusing on implementations
there are key individuals who are pushing the agenda Along with this need to re-focus the analytical per-
– and who have the necessary resources at their spective, trade unions should concentrate more on
disposal. the implementation level, and better address specific
ADM systems and processes in terms of their actual
Self-organizing and union building planning, development, implementation and use.
Where we see workers organizing around spe-
cific issues related to algorithmic transparency and Translating principles and ethical guidelines for the
accountability, it is often outside of traditional trade implementation of ADM systems into concrete tools
union structures. For example, the platform workers that can be applied to individual cases will generate
regularly self-organize and form their own union and case-specific knowledge. This knowledge can then
own structures rather than making use of existing be used to build the basis for broader capacity-build-
structures. The situation is similar in the US where ing measures and more in-depth knowledge on
so-called ‘digital employees’ (people in the tech algorithmic transparency and accountability in specific
industry and specifically Silicon Valley) organized uni- work-related contexts.
ons and gained support from existing unions as they
did so. Platform and tech workers are possibly tar- Focusing on the implementation of ADM systems
geted by algorithmic management tools more than would also help identify what kind of transparency
employees in other industries. Their self-organization is more specifically needed to best represent the
can be seen a consequence of this. interests of workers. For example, what are the limits
of transparency? How can meaningful accountability
be achieved? It must also be recognized that there are

/ RECOMMENDATIONS
limits to addressing ADM systems at the technological
level. In addition, it is important to address the con-

FOR STRATEGIC ACTION text of an individual organization. This would enable

worker representatives to formulate demands and
claim positions in ADM-related internal processes.
Sketching practice-oriented analytical needs
It is crucial to gather and analyze knowledge on Moving to the shop floor level
algorithmic management and the extent to which Generating knowledge and a focusing on implement-
transparency and accountability can help safeguard ation should complement more awareness-raising
workers. Also, analyses of the national legislative activities on the shop floor. In general, it seems like
framework will help address challenges concerning workers place less relevance on automation than
ADM systems in the world of work – for example, trade unions and their representatives do. This is
regarding data collection and management practices, probably due to the complexity of the issue and a
the legal basis used when a decision is made automat- lack of experience of the injustices and problems that
ically without human intervention or oversight, the often come with these systems. The consequence is
right to information for employee and worker repres- that campaigning on automation in the workplace
entatives, etc. remains very difficult. Algorithms and automation at
work is so far a topic for experts, but this situation
There is an urgent need to move toward more prac- should not continue. More awareness-raising meas-
tice- and shop-level-oriented analysis – addressing the ures are therefore equally needed.
research gap between ethical principles and tools that
promote algorithmic transparency and accountability Learning from platform workers
and how they are integrated into processes of co-de- Platform workers have quite advanced demands
termination and labor representations. Such insights regarding algorithmic transparency and accountab-
will help reassess tools that support worker repres- ility, and worker representatives from other sectors
entatives at all stages of implementing ADM processes can learn from them. At the same time, one needs to
into organizations. be aware of the limits. The case of platform workers
cannot be compared to all other scenarios, because

Page 18
 Algorithmic transparency
and accountability
in the world of work

their employers (e. g. the ride-hailing platform) are an opportunity to make algorithmic transparency
usually the ones also developing the algorithms that in the world of work a part of the mainstream labor
manage platform workers. Those companies thus not agenda.
only have direct insight into the algorithms, they could
also quite easily change them. However, this is not
the case for most companies that use ADM systems
to manage their workforce. Usually, People Analytics
software is being bought by external providers, which

Photo: Arno Senoner @unsplash

means a human resource department has a lot less
insight into how the software really works or takes
decisions and even less power in changing its auto-
mated decision-making procedures.

Having said that, insights from platform work can still

help formulate demands regarding what information
should be available about ADM systems. It also raises
awareness of the role of a company as the purchaser
of ADM systems, and of third-party producers. In South Africa, some organisations exist outside uni-
ons. For example, the South African Informal Traders
Alliance (SITA) represents street traders. The parti-

/ COUNTRY
cipation of SITA, and other forms of informal worker
mobilization, within the National Economic Develop-

DESCRIPTIONS ment and Labour Council (NEDLAC) has contributed to

NEDLAC’s policy utterances on the role of algorithms
in the future of work. The recent nationwide 3-day
In the following we will provide exemplary insights e-hailing strike in South Africa, which took place from
into the kinds of activities that trade unions engage in 22 to 24 March 2022, demonstrated the capacity
on algorithmic transparency and accountability in the to organize workers across platforms, with drivers
world of work within select countries across the globe. across e-hailing platforms – Uber, Bolt, InDriver and
Didi – making demands for fair compensation and bet-
ter security under the umbrella of “Unity in Diversity”.
/ AFRICA Although nothing in their demands seems to have
addressed algorithmic transparency and accountabil-
SOUTH AFRICA ity directly, it is clear that such issues could become
actioned through their association.23
There has been some union activity related to
algorithmic transparency in South Africa. For example, KENYA
by establishing the Vulnerable Workers Task Team,
the Congress of South African Trade Unions demon- In Kenya, there are informal worker associations that
strated a desire to bring the concerns of vulnerable, attempt to advance transparency concerning work-re-
part-time and gig workers to the table. As transpar- lated algorithms. For instance, the Awesome Tran-
ency issues tend to emerge from digital workers who scribers in Kenya Facebook Group acts as a crowd-
are often part-time or self-employed, this could have sourced transparency initiative. Although perhaps
promoted useful mechanisms. However, the Task better understood as an informal form of trade union,
Team was seemingly derailed due to internal issues. (private) Facebook Groups such as this – self-organ-
At the same time, the Nation Union of Metal Workers ized by gig workers transcribing in Kenya – act as a
of South Africa – a significant South African union – mechanism for the voluntary sharing of information
has included the rights of temporary workers within
their mandate. These are early signs of a slow shift 23  ttps://algorithmwatch.org/en/algorithms-and-the-world-of-
h
in the “traditional” membership of unions and are work/?country=southafrica

Page 19
 Algorithmic transparency
and accountability
in the world of work

among members to combat some of the risks of that initiatives in Argentina are geared predominantly
algorithmic opacity for gig workers, such as fee struc- toward algorithmic management. There is still much
tures and task allocation. to be developed in the debate on algorithmic trans-
parency. In 2018, the general strategic framework
While there are significant civil society activities con- concerning the use of AI in Argentina was defined
cerning algorithmic transparency and justice in Kenya, with the publication of its Digital Agenda for 2030. The
they largely relate to fintech and microlending and less agenda aims to take advantage of digital technologies
to the world of work. However, the significant amount for development and inclusion. It also established the
of research and activist work in this area might be an construction of the National Artificial Intelligence Plan
opportunity to align worker activism in-country. The that looks at the ethical aspects of AI and the impact of
Fairwork Project collaborates with civil society to pro- technology in the world of work.25
duce a Ratings Report on gig work conditions in Kenya
(and in South Africa). The report includes standards BRAZIL
on algorithmic fairness and transparency, aimed at
ensuring for example that if algorithms are used to In Brazil, the initiatives that look at algorithmic trans-
determine access to work or remuneration these do parency in the workplace are linked to publications
not result in inequitable outcomes for workers belong- and analysis of national regulations and laws. In
ing to disadvantaged groups. More broadly, and much addition, the Brazilian government has carried out
like in South Africa, there is generally a strong policy various consultations to build strategies on the eth-
focus on improving digital capacities in response to AI ics and impact of AI. These strategies target both the
and automation in Kenya.24 private and public sectors. Non-governmental organ-
izations take part in the consultations and come up
with recommendations on promoting algorithmic
/ THE AMERICAS transparency and accountability. In addition, other
actors seek to raise awareness and inform different
ARGENTINA audiences on the topic by creating and sharing spe-
cialized content in popular formats. Discussions on
In Argentina, initiatives that promote algorithmic algorithmic transparency in the world of work are
transparency and accountability are strongly linked to mostly centered on research and analysis of regu-
the “future of work”. Among the most common civil latory frameworks. There is still room to develop the
society, union and academic initiatives are aware- discussion further within civil society, particularly with
ness-raising, training, research/surveys, and analysis unions. The general outlook of AI regulation in Brazil is
of the future of work and the impact of digitization in marked by the Brazilian Artificial Intelligence Strategy
the world of work. Thanks to its editorial products – a and the national data protection law (Law N° 13,709
newsletter and podcast – and projects such as SinDi- / 2018). While they attempt to foster the use of AI by
gital, aimed at sharing perspectives, experiences and taking data protection and transparency into account,
capabilities on the impact of digitization in the world commercial secrets so far limit the transparency of
of work, the Friedrich Ebert Stiftung Argentina (FES algorithms in the private sector.26
Argentina) is a key actor in pushing the unions to
strengthen their focus on algorithmic management. CANADA
There are a variety of studies and investigations on
the implications of work on delivery and care plat- Trade unions in Canada address many concerns
forms. These are particularly concerned with labor around automation in the world of work. Instead of
rights and algorithm programming that promotes a focus on algorithmic transparency and accountab-
gender inequality. A first strike by platform workers ility, we can see many sector-specific theories about
led to the creation of the union “Asociación de Per-
sonal de Plataformas”. However, it should be noted
25 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=argentina
24  ttps://algorithmwatch.org/en/algorithms-and-the-world-of-
h 26 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=kenya work/?country=brazil

Page 20
 Algorithmic transparency
and accountability
in the world of work

what automation entails for a specific industry or a in the judicial system – for example, in the ‘Artificial
particular line of work (e. g., retail, education, aviation, intelligence and the judiciary. Chile and its pending
shipping industry). However, this does not lead to an challenges’ report written by researchers at the Pon-
in-depth contemplation about how to address auto- tificia Universidad Católica de Chile. Unions are rather
mation in the workplace. It often results in a call to working to stop the precarious labor situation related
reject and resist AI and automation. Discussions are to ADM systems in the retail and delivery sectors.
dominated by fear of job losses due to automation. As Broader discussions on AI and algorithmic transpar-
a result, a discourse around how algorithmic transpar- ency seem to be advancing quickly thanks to the pub-
ency and accountability might support workers and lication of a national strategy on AI by the Ministry
their representatives to mitigate risks associated with of Science in 2021. In civil society, the discussion has
automation and ADM should be explored. There are mainly been promoted in initiatives developed by uni-
occasional references to the need for up-skilling and versity academics and the Chilean Transparency Coun-
re-training in reaction to the risks posed by ADM sys- cil. But so far, discussions on algorithmic transparency
tems in the world of work. have focused heavily on the use of algorithms by the
state.28
The Canadian Union of Public Employees (CUPE) is
particularly concerned about automation in the public UNITED STATES OF AMERICA
sector, highlighting issues around workplace surveil-
lance. In a submission to Canada’s Office of the Privacy Questions of transparency and accountability do not
Commissioner, the union raised concerns about the figure very prominently in the discussion on automa-
expansion of AI in the public sector workplace. The tion in the world of work in the US. Instead, the focus
concerns particularly focused on the capacity of AI lies much more on risks associated with automation
to assess or monitor workers, both in the workplace within specific industries (e. g., in retail, aviation, the
itself, and in HR/recruitment activities. In reaction entertainment industry), on automation as a threat to
the CUPE introduced a “New Technology Committee” jobs in general or as a major challenge for trade uni-
to generate more union participation and insight in ons in representing employee interests. The discus-
decision-making on digitization and automation. Gen- sion in the US still seems to be very much centered
erally, trade unions are concerned with the possibil- around trying to understand the impact of automation
ities of collective bargaining and action in relation to in work-related contexts. This is reflected in the cre-
ADM systems in the world of work and theorize about ation of a research institute by the American Federa-
the role of unions in the future struggles that are tion of Labor and Congress of Industrial Organizations
foreseen.27 (AFL-CIO) to investigate problems posed by techno-
logy in the workplace, including algorithmic bias and
CHILE data privacy. The challenges of automation are widely
addressed by other organizations, including the UC
Several academic publications about the gig economy Berkeley Labor Center Report that has developed a
already exist in Chile. These include descriptive stud- framework for “Worker Technology Rights”, which
ies and evaluations of the experiences of delivery includes a framework for collective organizing and
and ride-hailing app workers that look at the opacity bargaining.
of their algorithms. There are also associations, or
groups of digital platform workers, such as Sindic- Specific support for strengthening worker rights
ato Cornershop Chile, Riders Unidos Ya Chile, Man- comes from the Communication Workers of Amer-
comunal de Repartidores de App de Chile (MAREA), ica (CWA) and their Campaign to Organize Digital
and Agrupación de Repartidores Penquistas. Bills Employees (CODE-CWA). They have developed a
seeking to regulate on-demand delivery apps are model through which tech, game, and digital workers
advancing through the National Congress. Activities in Silicon Valley are supported in unionizing. This way,
are also observed with regard to the use of algorithms workers who are subject to automation and AI, pos-

27 https://algorithmwatch.org/en/algorithms-and-the-world-of- 28 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=canada work/?country=chile

Page 21
 Algorithmic transparency
and accountability
in the world of work

sibly to a larger extent than in other industries, have tion of China in 2021, seek to regulate “algorithm-em-
the tools and knowledge to organize themselves.29 powered recommendation activities on the internet”.
These rules include provisions on transparency and
URUGUAY the use of AI for recommendations, data protection,
and independent audits. Separately, the government
While Uruguay is home to several technology-oriented has also initiated protections for gig workers, such
civil society organizations – such as, DATA, CAINFO as ensuring that unions are consulted before making
Uruguay and Datysoc – these have predominantly changes to an algorithm, assigning tasks, and fixing
focused on promoting open data, e-government wages. However, this power will largely be devolved
and access to public information, and other issues, to the All-China Federation of Trade Unions – a gov-
rather than on events or publications that address ernment-controlled union – that has been criticized
algorithmic transparency. The Friedrich Ebert Stiftung for not doing enough to protect the rights of work-
(FES) Uruguay has promoted activities and fostered ers. Furthermore, the protections in the draft rules
publications close to algorithmic transparency in the on algorithmic transparency that will apply for gig
labor field, and these mainly focus on the financial workers are not yet clear, since the rules are largely
sector and gig work. There are also associations and focused on protecting the interests of consumers and
organized groups of digital platform workers, such as users. This is also in the context of pervasive use of
the Asociación de Conductores Uruguayos de Aplica- electronic surveillance aimed at maintaining the abso-
ciones, Unión de Conductores Trabajadores de Aplic- lute control of the ruling party.31
aciones del Uruguay (UCTRADU) – grouped within the
PIT-CNT trade union center – and Sindicato Único de HONG KONG
Repartidores (SINUREP). More broadly, there are no
formal regulations on AI and algorithmic transpar- China’s imposition of the Hong Kong national security
ency in Uruguay. However, there is an incipient dis- law has stoked concerns over stifling civil society. The
cussion on algorithmic transparency. This is mainly law has direct consequences for gig workers trying
promoted by governmental entities, specifically by the to address their deteriorating working conditions by
Agency for the Development of e-Government and the demanding more algorithmic transparency. Several
Information and Knowledge Society (AGESIC) and its labor groups were forcibly disbanded since the law
AI strategy for digital government. This strategy incor- was introduced, most notably Hong Kong’s largest
porates transparency as one of its principles in the use trade union body – the Hong Kong Confederation of
of AI solutions in the public sector. The idea of trans- Trade Unions. Prior to the imposition of the security
parency in such principles includes the provision of law, freelancers in the gig economy had begun to uni-
information on algorithms and data used for training onize in the wake of rising exploitation mostly as a res-
and implementation, as well as tests and validations ult of algorithmic management in the world of work.
carried out, and the explicit visibility of all processes More recently, food delivery workers have achieved
that include AI in the state.30 success outside of traditional unions, by using digital
channels such as messaging platforms to take collect-
ive action with the support of experienced labor activ-
/ ASIA ist-volunteers. At least three strikes have been organ-
ized by Foodpanda delivery workers, demanding more
CHINA transparency of the algorithmic management systems
that determine their working conditions. Many of the
The Chinese government has been proactive in regu- concerns of food delivery workers relate to falling
lating AI, including its deployment in the world of work. income because of the automatic way gigs are alloc-
Draft rules, released by the Cyberspace Administra- ated. In addition, there is a lack of grievance redressal
to the way in which delivery times are automatically
calculated; in fact, algorithms use aerial maps (instead
29  ttps://algorithmwatch.org/en/algorithms-and-the-world-of-
h
work/?country=usa
30 
https://algorithmwatch.org/en/algorithms-and-the-world-of- 31 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=uruguay work/?country=china

Page 22
 Algorithmic transparency
and accountability
in the world of work

of street maps) to calculate distances. Governmental tion thus mostly lies on employees working alongside
bodies have begun assessing the relevance of self-em- digital and automated tools, rather than addressing
ployed platform workers in the Hong Kong economy, concerns over automation in the world of work and
but the government has signaled that it will not bring its impacts on employee rights. In the field of platform
gig workers within labor protection mandates.32 work we see some reports about gig workers organ-
izing and unions looking to work with platforms to
PHILIPPINES achieve more transparency on their algorithmic man-
agement systems. Union and worker demands include
Various government bodies in the Philippines have fair compensation, insurance, income protections,
been active in regulating or commenting on AI and and transparency about incentive structures within
its impact on workers. Critically, the Data Privacy Act platforms.34
2012 provides a right to explanation. The National
Privacy Commission (NPC) that was set up under the THAILAND
Act also provides guidelines on the data privacy of
workers. For example, the NPC stated that employ- Like other countries in the region, Thailand has
ers are allowed to monitor employees using surveil- developed policy frameworks that enthusiastically
lance software in work-from-home setups. This was adopt Industry 4.0 and related technologies. The ‘Thai-
permitted providing such monitoring balances with land 4.0’ model pushes for the adoption of technolo-
the rights and freedoms of employees and adheres to gies such as automation and ‘the Internet of Things’
general data privacy principles – such as being trans- across many sectors, including agriculture, manufac-
parent about when employees are being monitored. turing, and services. Similarly, the ‘Thailand Digital
On protections for gig workers, a resolution called for Economy and Society Development Plan’ sees digitiz-
an investigation in aid of legislation into the provision ation and the digital economy as the panacea to solv-
of employment benefits and other social protections ing problems around leakages and corruption in state
for gig workers, which would also impact algorithmic institutions, income inequality, and lack of capacity/
allocation of wages and tasks to workers.33 productivity in agriculture, manufacturing, and ser-
vices.
SINGAPORE
These frameworks reflect an uncritical push to grow
The Government of Singapore is focused on pro- the digital economy. However, some researchers cri-
moting the growth of its digital economy. An integral ticize the approach due to the lack of focus on the
part of it is the trade agreements it stipulated with interests of workers. The gig economy has been a crit-
other countries, for example Australia and Chile. ical area of intervention for trade unions and collect-
These agreements do not focus on workers’ rights, ives, with public demonstrations and strikes gaining
and instead push for growth of the digital economy significant media attention. This is in conjunction with
through regulatory sandboxes. Generally, the sup- similar protests organized by Grab drivers in other
posedly positive role of digital technologies and auto- Asian countries.
mation seems to be promoted in current labor-related
discussions in Singapore. Digital tools and automation Algorithmic management is central to the demands of
are seen for their ability to improve efficiency and pro- workers, due to the impact it has on the amount of
ductivity for workers. It is acknowledged that these pay per task, allocation of work, and how it can sus-
might result in changes for the workforce – and these pend accounts automatically (due to factors such as
are addressed through re-skilling and training initiat- low acceptance rates). Other actors, including interna-
ives for employees to adjust to digitization and auto- tional and national think tanks and NGOs, have also
mation in a work context. A labor focus on automa- looked at the ride-hailing sector in the gig economy.
So far, only the gig economy has protested against
algorithms and automation in the workplace. Mean-
32 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=hongkong
33  ttps://algorithmwatch.org/en/algorithms-and-the-world-of-
h 34 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=philippines work/?country=singapore

Page 23
 Algorithmic transparency
and accountability
in the world of work

while, various think tanks and academic institutions of algorithms that supervise the work of Amazon
have produced reports that try to understand the employees in both the Czech Republic and Poland
macro impact of AI in the workplace. These stake- and the consequences for the rights of workers.36
holders have also provided recommendations and
developed principles for the governance of AI and ESTONIA
data rights, given their current and predicted impact
on workers.35 In Estonia, discussions on automation in the world of
work do not put much weight on algorithmic trans-
parency and accountability. Different trade unions
/ EUROPE reflect on how a decreasing workforce ultimately
forces organizations to use more automation. How-
CZECH REPUBLIC ever, no initiatives that focus specifically on how
automation and AI will affect work or how far trans-
The impact of algorithms on the labor market is parency and accountability might safeguard worker
extensively described in the Czech Republic’s 2019 rights were identified. Discussions tend to concen-
AI development strategy. However, the main focus of trate on the overall possibilities and problems of
the strategy is on the challenges of changing the mar- using AI and ADM in the world of work. When gig
ket structure and the need for re-training demanded work is addressed, the discussion tends to focus on
by ADM systems and less on algorithmic transpar- the legal status of gig workers. Some organizations
ency and accountability. The risk of discrimination is emphasize the need to develop human-centered
only mentioned incidentally and is not connected to AI-based on the protection of individual rights. How-
algorithmic transparency – it is instead seen as a res- ever, this discussion does not focus on issues affect-
ult of the country’s demographic structure and chal- ing the workplace – reinforcing the argument that
lenges for older people. Admittedly, the authors of increased automation and AI in the workplace has
the strategy claim to be monitoring the ethical implic- yet to raise sufficient concern.37
ations of AI, but these have not been linked directly
to employment relationships. GERMANY

Several studies have looked at the impact of In 2018, the Federal Ministry of Labour and Social
algorithms on the gig economy, the labor market and Affairs set up the “Denkfabrik Digitale Arbeitsgesell-
the economy. One study devoted to the transport schaft” (Think Tank Digital Labor Society). The aim of
services market used the example of the ride-hailing this think tank – and the associated “future lab” – is to
platform, Uber. Following the study, laws were intro- help the Ministry identify new fields of action as a res-
duced to force Uber to operate under the same prin- ult of digitization and related trends and to develop
ciples as other registered transport. Academics are new approaches to solutions for the working soci-
also looking at the impact of AI on the labor market, ety of the future. Automation is prominent among
e. g., in relation to re-training needs. While the trade these topics, therefore algorithmic transparency and
union movement could be more engaged with the accountability are being addressed. The think tank is
topic of algorithmic transparency and accountability, mostly engaged in project funding, knowledge distri-
the action of one of the unions present at Amazon‘s bution, and awareness-raising through public events
Czech branch is noteworthy. It started a Facebook and networking initiatives. Equally, researchers in the
page to inform members – including Amazon employ- trade union movement have looked at AI and auto-
ees in the Czech Republic – about collective agree- mation. They have focused on how increased trans-
ment negotiations, health and safety in the workplace, parency and accountability of algorithmic manage-
and issues around algorithmic transparency that ment can help protect the rights of workers. In addi-
manage the work of Amazon employees. The action
raises questions relating to the lack of transparency
36 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=czechia
35 https://algorithmwatch.org/en/algorithms-and-the-world-of- 37 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=thailand work/?country=estonia

Page 24
 Algorithmic transparency
and accountability
in the world of work

tion, trade unions have also looked at the broader ITALY

picture of automation in the workplace and the
fundamental principles associated with it. The Ger- In Italy, algorithmic transparency and accountability
man Trade Union Confederation (DGB) has conduc- in the world of work are increasingly under debate.
ted comprehensive work on this topic. For example, The “working group” on the algorithmic governance of
their “political work program for shaping good work work created within the Labor Ministry of the Draghi
in the digital world” introduced a process-oriented government was a signal of interest in providing clear
perspective on transparency which highlighted the rules for both the gig economy and automated/digit-
importance of involving workers and their repres- ized labor. The active role taken by the country’s data
entatives in processes concerning the planning, protection authority – which issued landmark fines
implementation, and use of AI systems in the work- to gig economy platforms – is a signal that clarity is
place. Their transparency requirements touch upon badly needed. Trade unions have actively shaped the
decision-making criteria, but also on data quality, discussion by providing training material and counsel
use of data, discrimination, and accountability pro- to gig workers, co-organizing events on digital labor
cedures. The United Services Union (ver.di) equally and microwork with progressive research centers
highlighted the importance of transparency in work- and foundations. Furthermore, the unions want to
place ADM systems by supporting employees in their give platform workers better tools and knowledge to
co-determination activities. The union issued prac- “negotiate the algorithm”.
tice-oriented guidelines intended to form the basis
of discussions between worker representatives and This approach has led to an important – if contested
employers and ADM system developers, program- – agreement between unions CGIL, CISL, UIL, and
mers, and decision-makers.38 JustEat. While this agreement makes riders equal
to employees, this has not prevented riders and gig
HUNGARY economy workers from going on strike from Turin to
Florence and Bologna, and frequently with the backing
Trade unions tend to focus on the challenges that of trade unions, including Nidil CGIL.
digital transformation and automation pose (e. g.,
regarding large-scale job losses) and on what action Platform work is still the main point of contention in
needs to be taken to tackle these. Heightened com- Italy. However, a broader debate on the dangers of
petitiveness is also a focus, due to the idea that a sig- the unregulated use of automation and how to build
nificant share of jobs can presumably be automated fairer alternative cooperative models is increasingly
– making Hungary vulnerable in the light of digital common. This move is thanks to a handful of inter-
transformation. Trade unions seek to protect the nationally renowned researchers who popularized
position of workers by calling for structural changes notions such as “microwork” (Antonio Casilli) or “inhu-
to make the workforce ready for the digital age. How- man work” (Valerio De Stefano and Antonio Aloisi)
ever, the unions lack the resources to collaborate with both in mainstream media and during high-level
the necessary stakeholders. One major obstacle is the policy debates.40
deterioration of democratic processes in the country
and the lack of social dialogue that would be inclusive NORWAY
and provide meaningful participation. Therefore, Hun-
garian trade unions value international and EU-level In January 2020, the Norwegian Ministry of Local Gov-
discussions, as these are the main platforms for ernment and Modernization – with input from several
meaningful exchange and collaboration.39 labor unions – published a national strategy for AI.
Regarding digitization in the workplace, the biggest
concerns were issues with privacy and integrity. Uni-
ons identified worries such as the threat to jobs from
automation, but it is thought that a combination of
38 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=germany
39 
https://algorithmwatch.org/en/algorithms-and-the-world-of- 40 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=hungary work/?country=italy

Page 25
 Algorithmic transparency
and accountability
in the world of work

education and the creation of new digital jobs will transparency. However, it is withholding its submis-
mitigate these concerns. However, due to algorithmic sion until the government starts work on amend-
bias, there is an awareness that caution must be ments to the Act.
applied when algorithms are built and then imple-
mented in the workplace. Protests by workers at companies that use algorithms
are also increasingly visible. Amazon employees and
One influential actor is the Norwegian Society of workers at the Glovo platform have been particu-
Engineers and Technologists (NITO), which issued a larly vocal in demanding more transparency of the
declaration on the responsible use of AI in working algorithmic management systems. Some tools have
life – highlighting transparency as one of the core been specifically created, from the bottom up, to
principles. It states that individuals must be able to inform workers about such issues and provide docu-
understand the decision-making process of an ADM mentation on the algorithmic systems that have been
system and must be able to challenge the accuracy implemented.42
of a decision. Furthermore, NITO recommended that
relevant control mechanisms must be in place and RUSSIA
offer sufficient transparency. In addition, arrange-
ments should be made for internal reporting of risks Researchers and trade unions in Russia pay little
and routines for measures and follow-ups should be attention to algorithms in work-related contexts.
implemented. However, the trade union of Moscow taxi drivers
have made intensive efforts to increase the trans-
Also, the Confederation of Norwegian Enterprise and parency of the algorithm developed by the author-
the Union for energy, electrical engineering, telecom- ities to determine how they work. In addition, the
munications, and IT workers have been at the fore- association of bicycle couriers and their support-
front by making AI a negotiating item in a collective ers conducted a campaign to draw attention to the
agreement. This resulted in the inclusion of a section non-transparent and discriminatory operations of
on safeguarding the privacy and integrity of employ- the algorithms of one of the largest internet deliv-
ees and protecting their rights when confronted with ery platforms. In 2019, Russia adopted a strategy to
discrimination or prejudice because of AI systems.41 develop AI that claims transparency as a core prin-
ciple and aims at protecting basic rights, including in
POLAND the workplace. It also guarantees a process to build
competency among workers to adapt to the gig eco-
At the end of 2020, Poland adopted a strategy to nomy. The strategy is seen as a way to disseminate
develop AI. It contained many references to the need knowledge on how AI-based products operate, rather
for transparency of algorithms and the development than letting workers know more about algorithms in
of ethical principles to reduce the risk of discrimina- their workplace.43
tion. However, it is difficult to find a direct reference
to this guarantee in the world of work. It is worth not- SPAIN
ing that a committee was created in the Polish Parlia-
ment to deal with algorithmic transparency, and one Even before Spain adopted its National AI Strategy
of the committee’s meetings was devoted entirely to in December 2020, the increasing use of AI and
the use of algorithms in the labor market. This topic is algorithmic systems in managing the workforce
also under discussion within some trade unions. For had been researched by trade unions, academia,
example, the All-Poland Alliance of Trade Unions often and some civil society actors. Within the trade
discusses the subject and it is mentioned in its 2018- union movement, the two main unions, Comisiones
2022 strategy document. Furthermore, in a package of Obreras (CCOO) and the Unión General de Traba-
proposals for amendments to the Trade Unions Act,
the alliance has advocated for inclusion of algorithmic
42 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=poland
41 
https://algorithmwatch.org/en/algorithms-and-the-world-of- 43 
https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=norway work/?country=russia

Page 26
 Algorithmic transparency
and accountability
in the world of work

jadores, founded by the Socialist Party (UGT), both – specifically how to make the workplace more effi-
with around 900,000 members, have produced cient and safer. However, concerns have been raised
reports and guides to be used in collective bargaining about privacy, algorithmic bias, and a lack of trans-
and to educate and equip works councils and work- parency regarding ADM systems. There are active
ers’ representatives to deal with issues connected discussions about ADM systems in the workplace,
with algorithms and data protection. and most labor unions have touched upon the sub-
ject in articles, blog posts, and at conferences. Aka-
When a new collective bargaining agreement was demikerförbundet SSR, the union for professionals
adopted in the banking sector in 2021, it stated that or academics in social sciences, is a vocal critique of
workers had the right to be excluded from decisions ADM systems in the public sector. It has produced
taken solely by algorithmic systems – reportedly, this reports, articles, a podcast, and, most notably, repor-
is the first time such an agreement explicitly men- ted the town of Trelleborg’s automated social service
tions this issue. allocation system to the parliamentary ombudsman.
Their main criticism concerns the lack of transpar-
In civil society, the Eticas Foundation, a research and ency within ADM systems in the public sector and,
advocacy organization, is becoming very active. It col- among other things, the lack of involvement of public
laborated with CCOO and UGT to produce a report sector employees in the development process. These
on how trade unions can better protect the rights of reflections by the SSR resulted in specific demands,
workers in the face of digitalization and technological e. g., the need for a government framework ensuring
change, including the use of algorithms. In addition, transparency through a public register of ADM sys-
Eticas is also developing an external audit method- tems in the public sector, a red list of cases where
ology that can be used by trade unions and other ADM systems should not be used, and a higher level
entities to examine algorithms in the workplace. of data security.45

However, the key event in Spain regarding algorithms SWITZERLAND

and labor was the adoption of the so-called Riders’
Law in 2021. The main point of the law is to force In Switzerland, automation in the world of work comes
delivery companies to hire riders (who used to be up in general discussions on digitization and the labor
self-employed). However, the law also enforces trans- market, but it is often seen from a sector-specific
parency and explicability in the use of algorithms perspective (e. g., automation in the retail sector and
used in workplace and workforce management: this healthcare). Trade unions seek to understand how
is now an obligation extended to all companies oper- automation impacts the rights of employees in the job
ating in Spain. market and to share knowledge.

The Riders’ Law generated a lot of public debate Specific discussions on algorithmic transparency and
both before and after being passed, and the main- accountability are still on a small scale. For example,
stream and specialized media, as well as law and con- the Swiss Federal Railways (SBB) and railroad staff
sultancy firms, published articles explaining not only unions have jointly set up a digitalization fund to sup-
this particular law but how AI and algorithms have port studies that will develop the basis for a socially
been getting more and more common and pervas- responsible design of the digital transformation at
ive across all sectors in many companies when man- SBB.
aging employees.44
The need for instruments of social partnership in
SWEDEN the face of increasing automation in the workplace
is recognized. For example, the trade union repres-
In Sweden, trade unions discuss work and ADM enting private sector workers in Switzerland, Unia,
systems from an efficiency and safety point of view put forward the idea that employees should receive

44 https://algorithmwatch.org/en/algorithms-and-the-world-of- 45 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=spain work/?country=sweden

Page 27
 Algorithmic transparency
and accountability
in the world of work

training to adapt to changes caused by digitization shaped by the UK’s departure from the European
and automation and implement this in collective labor Union, which meant that areas of the law previously
agreements. covered by GDPR are now under the jurisdiction of
the UK government. Prospect Union campaigns to
Furthermore, Syndicom, a trade union in the telecom- ensure that worker’s rights are protected as the prin-
munication and IT sector, came up with nine principles ciples of GDPR are brought into UK law under the
to guide when implementing automated systems. Two Data Protection Act 2018.
of these principles include transparency and account-
ability and ask that ADM systems are transparent, The lack of documentary evidence at the grassroots
understandable, and recognizable and that results level suggests that union action in this area has not
should be reproducible as well as traceable. Organiza- yet reached a high level of importance; however, a
tions implementing ADM systems should analyze and critical mass, whereby union action becomes an
evaluate the systems and must account for them. In organized campaign that the union needs to build
addition, there should be regulated access for govern- support and solidarity, can still be reached. There
ment bodies and researchers to monitor AI systems are likely to be highly individualized cases in various
externally.46 workplaces that require workplace representation,
but these are not a matter of public record yet.47
UNITED KINGDOM

In the UK, AI and automation have gained the atten- / OCEANIA

tion of policy-makers, union-supporting think tanks,
and researchers in the trade union movement. Sig- AUSTRALIA
nificantly, the Trades Union Congress (TUC) launched
a manifesto that sets out plans for campaigns and a In Australia, trade unions focus on the impact of auto-
strategy in this area. This manifesto offers guidance mation in specific industries (e. g., education, retail,
to unions as they begin to encounter workplace dis- health). This leaves a gap, where algorithmic trans-
putes because of AI and automation. The manifesto parency and accountability could be considered on a
and also a soon-after published guide on collective general scale. Noteworthy is a submission by the Aus-
bargaining and digital management systems pushes tralian Services Union to the Australian Human Rights
for the promotion of workers’ voices through collect- Commission acknowledging that the world of work
ive bargaining, and argues for a duty to consult when has drastically changed as a result of technological
high-risk ADM systems are implemented in work-re- developments such as automation. The union states
lated contexts. Also, legislation should ensure access that principles formulated as part of a just transition
to information and a universal right to explainability toward technologically transformed labor markets in
regarding high-risk systems – to better ensure trans- the future must include social dialogue and demo-
parency and accountability. The TUC encourages cratic consultation of social partners and stakeholders.
member unions to pledge support for the manifesto. The union also emphasizes the need for research and
early assessment of social and employment impacts.
Prominent campaigns concerned with matters of Even though algorithmic transparency and accountab-
automation in the world of work focus on the legal ility are not mentioned directly, they are likely to align
aspects of data gathering in the workplace. For with the formulated principles.48
example, the Prospect Union’s work on digital tech-
nologies and data protection impact assessments
(DPIA) provide guidance for worker representatives
encountering automation in the workplace to mitig-
ate some of the potential problems for employees
in the datafied workplace. This legal approach was
47 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=uk
46 
https://algorithmwatch.org/en/algorithms-and-the-world-of- 48 https://algorithmwatch.org/en/algorithms-and-the-world-of-
work/?country=switzerland work/?country=australia

Page 28
 Algorithmic transparency
and accountability
in the world of work

/ RESEARCHERS / INTERNATIONAL
Publication: February 2023 TRADE UNION
Funding: This research was conducted on behalf of CONFEDERATION
the International Trade Union Confederation (ITUC).
(ITUC)
Authors/Researchers:

Content Lead: Dr. Anne Mollen, Senior Policy & The ITUC represents 200 million workers in 163 coun-
Advocacy Manager (AlgorithmWatch) tries and territories and has 332 national affiliates.
As the world’s main global confederation of national
Editorial Lead: Fabio Chiusi, Research Associate union centres, its’ mandate is focused on the protec-
(AlgorithmWatch) tion of workers’ rights, social and economic justice,
Initiated by Tim Noonan (ITUC) anti-discrimination, Just Transition to a zero-carbon
With support from Felix Barbieri (ITUC) future, development, peace and democracy. Shaping
the future of work is one of its key areas of activity.
Layout: Beate Autering
Cover Photo: Chantal Garnier @unsplash https://www.ituc-csi.org/

The country and region-specific research was produced

by the following individuals or organizations.
Argentina: Derechos Digitales
/ ALGORITHMWATCH
Australia: Cate Hopkins
Bangladesh: Aayush Rathi & Ambika Tandon AlgorithmWatch is a non-profit research and advocacy
Brazil: Derechos Digitales organization that is committed to watch, unpack
Canada: Cate Hopkins and, analyze algorithmic/automated decision-making
Chile: Derechos Digitales (ADM) systems and their impact on society. While the
China: Aayush Rathi & Ambika Tandon prudent use of ADM systems can benefit individu-
Czech Republic: Krzysztof Izdebski als and communities, they come with great risks. In
Estonia: Maris Männiste order to protect human autonomy and fundamental
Germany: Sarah Michot rights and maximize the public good, we consider it
Hong Kong: Aayush Rathi & Ambika Tandon crucial to hold ADM systems accountable to demo-
Hungary: Nikolett Aszódi cratic control. Use of ADM systems that significantly
India: Aayush Rathi & Ambika Tandon affect individuals’ and collective rights must not only
Indonesia: Aayush Rathi & Ambika Tandon be made public in clear and accessible ways, individu-
Italy: Fabio Chiusi als must also be able to understand how decisions are
Kenya: Gabriella Razzano reached and to contest them if necessary. Therefore,
Norway: Vian Tahir we enable citizens to better understand ADM systems
Philippines: Aayush Rathi & Ambika Tandon and develop ways to achieve democratic governance
Poland: Krzysztof Izdebski of these processes – with a mix of technologies, regu-
Russia: Krzysztof Izdebski lation, and suitable oversight institutions. With this, we
Singapore: Aayush Rathi & Ambika Tandon strive to contribute to a fair and inclusive society and
South Africa: Gabriella Razzano to maximize the benefit of ADM systems for society at
Spain: Jose Miguel Calatayud large.
Sweden: Vian Tahir
Switzerland: Sarah Michot https://algorithmwatch.org/en/
Thailand: Aayush Rathi & Ambika Tandon
United Kingdom: Cate Hopkins
Uruguay: Derechos Digitales
United States of America: Cate Hopkins

Page 29