Chapter 4

Responsible and accountable

algorithmization
How to generate citizen trust in governmental
usage of algorithms

Albert Meijer and Stephan Grimmelikhuijsen

Introduction
Algorithms are increasingly popular in the public sector in countries all around the
world: they are used to provide services (Pencheva, Esteve & Mikhaylov, 2018) but
also to, for instance, support decision-making (Van der Voort et al., 2019) and pre­
dict recidivism (Kleinberg et al., 2017). Furthermore, the police use algorithms to
predict crime patterns (Meijer & Wessels, 2019), tax departments use algorithms to
detect tax fraud (Zouridis, Van Eck & Bovens, 2020) and local governments use
algorithms to make garbage collection more efficient (Ramalho, Rossetti & Cacho,
2017). The current wave of algorithms uses relatively new techniques, such as
machine-learning and deep learning, to transform organizational processes
(Brynjolfsson & McAfee, 2014; Burrell, 2016).
The ‘magic’ of these new technologies is appealing to governments since they
promise to bring us more effective processes, better informed decisions and more
insights in complex realities through informative and seamless interfaces. At the
same time, the ‘magic’ of these new technologies is also risky since the use of algo­
rithms can produce bias and even discriminatory practices; it can result in errors in
the implementation of policies and it can also hamper the interactions between
governments and citizens. Therefore, various authors stress that we need to step
back and reflect on how algorithms can be applied to realize desirable outcomes
(O’Neil, 2016; Eubanks, 2018; Gerards, 2019). For instance, various municipal gov­
ernments in the Netherlands used a machine-learning algorithm (Systeem
Risicoindicatie, SyRI) to detect welfare fraud amongst citizens. While at the begin­
ning there was widespread support for this system among government officials, in
early 2020 the system was judged ‘discriminatory’ and ‘not transparent’, not only
by civil rights activists, but also by the District Court. Vulnerable citizens, often
those from a migrant background, were unfairly profiled to be a suspect of welfare
fraud.
The example of SyRI in the Netherlands illustrates an issue of wider signifi­
cance: there is an urgent societal need to not only focus on issues of effectiveness
and efficiency but also identify how governments can avoid negative unintended
consequences of the use of algorithms – such as bias and problems of fairness – to
maintain the trust of citizens (Hoffman, 2019). How can we expect citizens to trust
54 Albert Meijer and Stephan Grimmelikhuijsen

government if this system works with an algorithm that is both opaque and dis­
criminatory? Various concerns have been raised regarding the impacts of these
systems on privacy but also on discrimination of different groups in society and its
neglect of human contact. It is still unclear how government organizations will
deal with issues such as interpreting big data in a non-discriminatory manner, han­
dling privacy fairly, using means proportional to the objectives and ensuring human
contact (Dencik et al., 2019).
We argue that these concerns go beyond the mere implementation of algo­
rithms; they also relate to how organizations transform and change to enable the
use of algorithms. In this chapter, we will label this process algorithmization: an
organization that transforms its working routines around the use of algorithms for
its actions and decisions. We highlight that an analysis of algorithmization requires
a focus not only on the technology but also on its organizational implementation
in terms of the expertise of employees, information resources, organizational struc­
ture, organizational policy and monitoring & evaluation to understand why the use
of algorithms does or does not produce citizen trust.
In this chapter we link algorithmization to a crucial concept in contemporary
governance: citizen trust in government. Trust in government is regarded as an essen­
tial element in developed societies. It has been found, for example, that if govern­
ment institutions are not trusted by the citizens they serve, they are unable to
function properly (Fukuyama, 1995; Inglehart, 1999; Levi & Stoker, 2000). Given
the rapid algorithmization of government, we need to identify desirable forms of
algorithmization in the public sector. Two preconditions for maintaining citizen
trust have been proposed in the literature: (1) incorporating values in the design of
algorithms as a precondition for organizational responsibility (e.g. Friedman, Kahn
& Borning, 2008; Van den Hoven, 2013) and (2) demonstrating the correct usage
of algorithms to the public as a precondition for accountability (e.g. Diakopoulos,
2016). In this chapter we argue that algorithmization in the public sector can only
sustain citizen trust when it is based on both preconditions.
To this end, we first discuss what we understand by citizen trust and we outline
why trust is so important in the public sector.We then offer a discussion of algorith­
mization as an organizational process and we stress that this organizational process,
rather than the technology in itself, demands our attention if we want to strengthen
citizen trust. The next sections discuss how two preconditions – responsible and
accountable algorithmization – can contribute to citizen trust. We end this chapter
by presenting a model for responsible and accountable algorithmization in the pub­
lic sector.

The fundamental role of trust in government

Although many scholars have emphasized the importance of trust, we start off with
a note that trust in government as such is not strictly necessary. People can ‘accept’
and obey an oppressive government, not because they trust it but because they fear
the consequences of disobedience. According to political scientist Russell Hardin
(1999, 2002), trust in government is ‘only’ needed under relatively benign
 Responsible and accountable algorithmization 55

circumstances, such as can be found in democratic regimes. Hardin further argues

that government functions well as long as it not actively distrusted by people.
A certain degree of trust makes governing much easier and more benign. Many
scholars argue that if government is perceived to be trustworthy, citizens tend to
comply more often with its demands, laws and regulations without coercion (Tyler,
2006). For instance,Tom Tyler and Peter Degoey (1996) found that people’s evalu­
ations of the trustworthiness of organizational authorities shape their willingness to
accept the decisions of authorities and influence their feelings of obligation to fol­
low organizational rules and laws. Indeed, trust can be viewed as an important
component of government legitimacy (Tyler, 2006).
Furthermore, Marc Hetherington (1998) highlights the relevance of political
trust. Political trust concerns citizens’ trust in their political leaders, which trans­
lates into more support for politicians and political institutions. This gives leaders
more leeway to govern effectively and offers institutions more support and legiti­
macy. Furthermore, without public support for solutions, problems tend to linger
and become more acute; if not resolved, this becomes the foundation for
discontent.
Citizen trust is not a necessity, but it is still regarded as highly important for
government. But what exactly is trust in government? To better understand citizen
trust in government we have to turn to a variety of scholarly disciplines.
Understanding why and how people trust has been a central focus of research by
psychologists, sociologists, political scientists, economists and organizational scien­
tists (Grimmelikhuijsen, 2012). Across and even within disciplines, countless defi­
nitions, concepts and operationalizations are being used. In an attempt to find
cross-disciplinary agreement about the concept of trust, Denise Rousseau, Sim
Sitkin, Ronald Burt, and Colin Camerer (1998) developed a definition that is fre­
quently cited in the social sciences. According to them, trust is ‘a psychological
state comprising the intention to accept vulnerability based upon positive expecta­
tions of the intentions or behaviour of another’.
According to Rousseau and colleagues, all definitions of trust assume the pres­
ence of some form of positive expectation regarding the intentions and behaviour
of the object of trust.The object of trust in this chapter is government. An element
of this definition that requires more elaboration is ‘positive expectations of the
intentions or behaviour of another’. Of what, in the context of government, are
these positive expectations comprised?
Trustworthiness concerns the characteristics of the object of trust as perceived by
an individual (Mayer, Davis & Schoorman, 1995). A large body of literature has
attempted to identify specific elements that might influence an individual’s percep­
tions of trustworthy behaviours and intentions (see, for an overview, McKnight,
Choudhury & Kacmar, 2002; McEvily & Tortoriello, 2011; Grimmelikhuijsen &
Knies, 2017). Generally, the importance of the various elements tends to differ accord­
ing to the discipline in question, yet some degree of commonality can be found.
The three most commonly cited elements are perceived competence, perceived
benevolence and perceived integrity (also sometimes called honesty)
(Grimmelikhuijsen, 2012). Perceived competence is the extent to which a citizen
56 Albert Meijer and Stephan Grimmelikhuijsen

perceives government to be capable, effective, skillful, and professional; perceived

benevolence refers to the extent to which a citizen perceives government to care
about the welfare of the public and to be motivated to act in the public interest;
perceived integrity is whether a citizen perceives government to be sincere, truth­
ful, and to fulfil its promises. It should be noted that benevolence and integrity are
of a different nature than competence, as they reflect ethical traits rather than some
kind of capability. Benevolence reflects the trustee’s motives and is based on altru­
ism. In contrast, competence is a utilitarian dimension of trust, as it refers to the
actual functioning of government organizations.
This section discussed the fundamental role of trust in government and how it
can be conceptualized.We highlighted that trust should be understood as a multi­
dimensional concept that consists of citizens’ perceptions of government compe­
tence, benevolence and integrity. In the next section, we outline what we mean by
algorithmization of government.

Algorithmization in the public sector

Many academic debates focus on the characteristics and features of algorithms as a
technological artifact.This perspective focuses on the design of these algorithms to
prevent biases and prejudices. We acknowledge the importance of value-sensitive
design of algorithms but also stress that it is not only about the design. We know
from decades of studies of information and communication technologies in the
public sector that we need to study organizational practices to understand the
effects of the use of algorithms. This is why we use the term ‘algorithmization’ as
an organizational process rather than algorithm as a technological artifact. Building
upon earlier work on informatization in government (Zuurmond, 1994: 42–48),
we distinguish the following components of algorithmization:

1 Technology. The process of algorithmization starts with the introduction of a

technology into organizational processes.The algorithm can be a stand-alone
decision-support system but also a system that is well integrated in the orga­
nization’s infrastructure.
2 Expertise.The use of algorithms in an organization requires a variety of exper­
tise. Experts that know how to work with the system are needed but also
experts that maintain the algorithm and ensure that it is properly installed in
the organization’s information environment.
3 Information relations. Algorithmic applications will generally build upon exist­
ing information in the organization but also produce new types of informa-
tion.This means that the algorithm has an effect on the information relations
within the organization and often also outside of the organization if informa­
tion from other actors is used.
4 Organizational structure. The use of algorithms often leads to new collabora­
tions between different departments. Algorithmic applications can also lead to
new forms of organizational control if they dictate the implementation of
processes.
 Responsible and accountable algorithmization 57

5 Organizational policy. Organizations develop policies for the use of an algo­

rithm in the organization. These policies touch upon issues such as the trans­
parency of the algorithm, responsibilities for usage, maintenance, etc.
6 Monitoring and evaluation. Organizations develop methods and systems for
monitoring and evaluating the foreseen and unforeseen outcomes of the use
of algorithms in terms of output and effects.

These six elements form the cornerstones of a conceptual understanding of

algorithmization and help us to sharpen our perspective when we study how algo­
rithms are used in the public sector. They also provide starting points for thinking
about strengthening citizen trust in the use of algorithms.This trust does not only
depend on the nature of the technology but also on the experts that guide the use
of the algorithm, the information that is used by the algorithm to provide its out­
put, the organizational control over the algorithm, and the policies that organiza­
tions have developed to guide the usage and maintenance of the algorithm.
In an empirical study of the Berlin Police, Lukas Lorenz (2019) used this orga­
nizational perspective on algorithms to rethink the nature of bureaucratic organi­
zations. Building upon Max Weber’s classic work on bureaucratic organizations and
Arre Zuurmond’s (1998) work on infocratic organizations, Lorenz sketches the
contours of a new type of organization: the algocracy. He conceptualizes the algo­
cracy as new ideal type of rational-legal authority that helps to understand and
explain how algorithmic systems shape public organizations. He characterizes the
algocracy as a further rationalized organizational configuration of the professional
bureaucracy rather than of the machine bureaucracy: the standardization of skills is
replaced by automated advice.
This discussion highlights that we need to think about the organizational pro­
cesses surrounding the introduction of new technologies in organizations – the
emerging algocratic organizations – to develop ways to strengthen citizen trust in
the public sector.To this end, we emphasize two classical approaches to strengthen­
ing trust in government organizations: responsibility and accountability.

Responsible algorithmization
A first route for strengthening citizen trust in algorithmization is provided by the
notion of responsibility. Responsibility is one of the key concepts in ethical theory
and its roots can be traced back to Aristotle. He emphasized that moral responsibil­
ity grows out of an ability to reason, an awareness of action and consequences, and
a willingness to act free from external compulsion (Roberts, 1989). Responsibility
refers to the idea that persons have moral obligations and duties to others, such as
their well-being and their health, and to larger ethical and moral traditions such as
freedom and empowerment, and that persons, thus, need to consider these obliga­
tions and duties in their individual decisions and actions.
More recently, the notion of responsibility has been applied to public organiza­
tions. Political scientist Herman van Gunsteren (1976) highlights that we should
strive to embed responsibility in (complex) organizations rather than emphasize
58 Albert Meijer and Stephan Grimmelikhuijsen

the need to control the activities of all individual members of organizations. His
approach stresses the need to acknowledge the limitations of bureaucratic control
mechanisms such as hierarchy and formal rules and to place more emphasis on the
responsibility of individuals in the organization. According to Van Gunsteren, the
notion of responsibility is needed when rigid moral norms – embedded in organi­
zational standard procedures – cannot keep up with rapidly changing circum­
stances and more flexibility is required to apply sound moral judgement. At the
same time, this does not mean that ‘anything goes’: judgement needs to be made
based on ethically respectable values and reasonable perceptions of relevant facts
(Van Gunsteren, 2015: 318). ‘Responsibility forums’ where individual judgement
needs to be explained can play a key role in strengthening the responsibility of
individuals in organizations.
Furthermore, the notion of responsibility has been used to think about innova­
tion. In the context of the European Union, the notion of responsible innovation
has come to play an important role in funding for technology development. The
key idea is that innovation traditionally focuses only on gains in efficiency and
effectiveness. Other considerations are often regarded as barriers to the innovation
process. The notion of responsible innovation reconceptualizes these barriers and
stresses that other values need to play a role in the way the innovation process is
structured and implemented. In our understanding of responsible algorithmization,
we build upon the broader notion of responsible innovation, which Richard
Owen, Phil Macnaghten and Jack Stilgoe define as ‘a collective duty of care, first to
rethink what we want from innovation and then how we can make its pathways
responsive in the face of uncertainty’ (2012: 757–758).
These general notions about responsibility in organizations and responsible
innovation can be used as a basis for ideas about the responsible use of algorithms
in public organizations. Key elements in this conceptualization are (1) ethical
judgement, (2) based on values, (3) and perceptions of relevant facts, (4) to enact a
duty of care (5) through responsive pathways. Based on these elements, we formu­
late the following conceptualization of responsible algorithmization:

Responsible algorithmization refers to the adequate weighing of ethical

dilemmas involved in the organizational use of algorithms based on knowl­
edge about the (possible) impacts to ensure that the algorithmization respects
moral obligations to others and to moral traditions through methods that are
responsive to the various other actors involved.

Responsibility for algorithmization can be operationalized by building upon the

notion of value-sensitive design, which can be defined as ‘a theoretically grounded
approach to the design of technology that accounts for human values in a princi­
pled and comprehensive manner throughout the design process’ (Friedman,
Kahn & Borning, 2008: 69). For a value-sensitive algorithmization of government,
a comprehensive understanding of the values at stake in a specific process of algo­
rithmic decision-making and of how these values are incorporated in the organi­
zational use of the algorithm is needed (Mingers & Walsham, 2010). Thus, we
 Responsible and accountable algorithmization 59

apply value-sensitivity not only to the design of the technology but also to the
other elements of algorithmization (expertise, information resources, organiza­
tional structure, organizational policy).The questions that can help guide organiza­
tions towards responsible algorithmization are presented in Table 4.1.
How will responsible algorithmization affect citizen trust in government? In the
above, we identified value-sensitivity as a core component of responsible algorith­
mization. This relates directly to the benevolence and integrity dimension of trust
in government. Benevolence refers to the expectations by citizens that government
is acting benign and in the interest of citizens; integrity refers to expectations of
honesty and truthfulness (Mayer, Davis & Schoorman, 1995; Grimmelikhuijsen &
Knies, 2017). If algorithmization is done with consideration of relevant values –
e.g. algorithms are used in a fair and unbiased manner – this will likely positively
affect citizens’ perceptions of the honesty and benevolence of government.
Conversely, when citizens, for instance, feel a welfare fraud algorithm targets them
unfairly, this is likely to cause a decline in perceived honesty and benevolence.
The different organizational policy components – organizational policy and
monitoring & evaluation – are also likely to contribute to citizen trust in govern­
ment competence. The basic argument here is that well-considered choices in
terms of how the algorithm is to be used in the organization and a consistent
monitoring and evaluation of the desired and undesired outcomes contribute to

Table 4.1 Assessment questions for value-sensitivity as a precondition for

responsible algorithmization

Dimension of Assessment question

algorithmization

Technology How are the different values at stake identified and

embedded in the design of the algorithm?
Expertise Are the experts that develop, support and maintain
the algorithm aware of relevant ethical
considerations and can they make value-sensitive
choices?
Information relations Have the value choices in the datasets used by the
algorithm been analysed and are the values that
follow from new combinations of data
acknowledged?
Organizational Has the overall responsibility for a value-sensitive
structure use of the algorithm in organizational practices
been clearly allocated?
Organizational policy Does the organization have a policy for ensuring
value-sensitivity in the organizational use of the
algorithm?
Monitoring and Does the organization have a system for monitoring
evaluation and evaluating outcomes in terms of the various
values at stake in the use of the algorithm?
60 Albert Meijer and Stephan Grimmelikhuijsen

the perception that government is using these algorithms in a rational manner.

There is a reason for caution here, however. If citizens expect governments to per­
form much better than they actually do, this can also result in a decline of perceived
competence (Grimmelikhuijsen, 2012).
In sum, the first condition that we have identified for maintaining the trust of
citizens in an organization that uses algorithms is responsible algorithmization.We
have argued that value-sensitivity is the basis for responsible algorithmization and
we listed six questions that can help organizations assess their level of responsible
algorithmization.

Accountable algorithmization
A second route to strengthening citizen trust in algorithmization is provided by the
notion of accountability. This notion builds upon the concept of public account­
ability which political scientist Mark Bovens defines as ‘a relationship between an
actor and a forum, in which the actor has an obligation to explain and to justify his
or her conduct, the forum can pose questions and pass judgement, and the actor
may face consequences’ (2007: 450). Based on this general definition, we can pro­
vide a specific definition (see Wieringa, 2020, for an in-depth analysis) of algorith­
mic accountability:

Accountable algorithmization can be defined as the justification of the orga­

nizational usage of an algorithm and explanations for its outcomes to an
accountability forum that can ask questions, pass judgement and impose
consequences.

Transparency is an important condition for realizing public accountability,

although scholars have noted that transparency does not automatically lead to
more accountability (Hood, 2010). Albert Meijer (2014) indicates that transpar­
ency facilitates accountability if it presents an actual and significant increase in the
available information, if there are actors capable of processing the information, and
if exposure has a direct or indirect impact on the government or public agency.
Without transparency, accountability is difficult to realize since relevant facts that
need to be assessed are not available.
Lack of transparency is a key concern regarding the use of algorithms in the pub­
lic sector (e.g. Lepri et al., 2018). Machine-learning algorithms that use various
internal and external data sets are so complicated that the logic of decision-making –
and possible biases – are difficult to detect (Janssen & Van den Hoven, 2015). In addi­
tion, the lack of transparency may concern the responsibilities, procedures and
practices of algorithmic usage in the organization. In response to these concerns,
algorithmic transparency has been proposed as a key element of accountable algo­
rithms applications in the public sector (Diakopoulos, 2016; Lepri et al., 2018). We
extend this concept to transparency and apply it not only to the algorithm as a
technology but also to its organizational use.
 Responsible and accountable algorithmization 61

The basic idea is that algorithmic decision-making by government should be

accessible and explainable (Kroll et al., 2017). Accessibility implies providing clear
information about the input, throughput and output of a decision-making process:
which data has been used, which decision rules have been applied and what was the
outcome? Explainability concerns the substantive reasons for a decision: on what
grounds was the decision made and how does this relate to legislation and other
formal rules and policies? In short, accountable algorithmization means that algo­
rithmic decision-making needs to be accessible and explainable. Following these
considerations, we define a set of questions to guide organizations (see Table 4.2).
Procedural fairness theory offers valuable insights in how transparency – in terms
of accessibility and explicability – may affect trust. Procedural justice theory (e.g.
Tyler, 2006) posits that individuals can be satisfied with negative decisions as long as
they consider the decision procedure to be fair. Accordingly, accessible and explain­
able algorithmization helps foster fair procedures and eventually more trust in a
decision-maker (i.e. government) (Grootelaar & Van den Bos, 2018; Porumbescu &
Grimmelikhuijsen, 2018).

Table 4.2 Assessment questions for transparency as a precondition for

accountable algorithmization

Dimension of Assessment question for Assessment question for

algorithmization accessibility explainability

Technology Is there access to the code Does the algorithm provide

to scrutinize design substantive reasons for
choices? decisions or advice?
Expertise Are the function Are reasons provided for
characteristics of the the expertise involved in
experts involved in algorithmization?
algorithmization
transparent?
Information Is there access to the key Does the organization
relations features of the data sets explain which datasets are
used by the algorithm? used by the algorithm,
why and how?
Organizational Are organizational Are choices regarding
structure responsibilities for the organizational
algorithmization responsibilities for
transparent? algorithmization
explained?
Organizational Is the organizational policy Is the organizational policy
policy for algorithmization for algorithmization
accessible? explained?
Monitoring and Is there access to the results Are the foreseen and
evaluation of the algorithmization in unforeseen effects of
terms of foreseen and algorithmization
unforeseen effects? explained?
62 Albert Meijer and Stephan Grimmelikhuijsen

More specifically, explicability is a core component of procedural fairness.

Explaining citizens how an algorithm functions is expected to have a positive
impact on levels of trust (cf. Grimmelikhuijsen et al., 2019). A second component
of procedural fairness is the neutrality of the decision-maker (Van den Bos,Vermunt
& Wilke, 1997). Core elements of Table 4.2, such as the accessibility of datasets,
algorithms and organizational policies indicate that, in an optimistic scenario, deci­
sions based on algorithmization are neutral and unbiased. Eventually, similar to
value-sensitivity, this relates to how citizens perceive the benevolence and integrity
of government.
Transparency regarding the foreseen and unforeseen effects of algorithmization
may contribute to the perceived competence of government. If citizens see that
algorithmization actually leads to desired outcomes while the undesired outcomes
are limited, their trust in government’s ability to realize better outcomes may
increase. The monitoring and evaluation of algorithmization is also expected to
play a key role in contributing to perceived competence.
In sum, the second condition that we have identified for maintaining the trust
of citizens in an organization that uses algorithms is accountable algorithmization.
Using transparency as a proxy for accountable algorithmization, we have listed 12
questions that organizations can use to assess their level of accountable
algorithmization.

Conclusion
In this chapter, we have discussed how ‘algorithmization’ has become a potent
force, changing traditional bureaucracies into algocracies. The new generation of
algorithms that are now finding their way to governments across the globe are
more than a change of technology: they trigger a range of organizational changes
that eventually transform bureaucratic decision-making. In this context, machine-
learning algorithms are often portrayed as problematic for accountable and respon­
sible decision-making. We argue that both accountable and responsible
algorithmization are needed to sustain citizen trust in the use of these algorithms.
The argument we developed in this chapter is summarized in Table 4.3.
The two preconditions of value-sensitivity and transparency are a starting point
for realizing responsible and accountable algorithmization. However, this does not
resolve all issues. A first issue is that algorithmization is not limited to the use of
merely one technological system in an organization.There are entire ecosystems of
algorithms that use data from various sources and that are implemented in net­
works of organizations (Cicirelli et al., 2019).The ‘problem of many hands’ is com­
pounded by these developments and it is not easy to indicate who is responsible
and accountable. In that sense, further work on the assessment questions formu­
lated here is needed to test and re-develop them for ecosystems of algorithms.
A second issue that demands more attention is the fact that machine-learning
algorithms change over time. This raises questions about the dynamics of values:
value-sensitivity may be ensured at the start but a machine-learning algorithm can,
over time, develop patterns that conflict with key values. In addition, transparency
 Responsible and accountable algorithmization 63

Table 4.3 Potential relations between algorithmization and trust in government

Dimension of Value-sensitivity as a Transparency as a precondition

trust precondition for responsible for accountable algorithmization
algorithmization
Competence Value-sensitive Explicable and accessible
algorithmization may monitoring and evaluation of
strengthen perceived algorithmization provides
competence if the insights in the outcomes,
organization demonstrates which is expected to
that various values are increase perceived
measured and its policy competence.
focuses on realizing value.
Benevolence Value-sensitive Explicable and accessible
algorithmization ensures algorithmization ensures
that values important to that government works in
citizens (e.g. fairness) are the interest of citizens,
not overlooked, which is which is expected to
expected to increase increase perceived
perceived benevolence. benevolence.
Integrity Value-sensitive Explicable and accessible
algorithmization ensures algorithmization ensures
that decision-makers are that external stakeholders
more value-sensitive and have access, which
thus act more ethically; this contributes to more open
is expected to increase and truthful
perceived integrity. algorithmization.

may be ensured at the start, but after a process of machine-learning the algorithm
can have become opaque because its decision rules have changed following learn­
ing processes. This means that the questions formulated here may need to be
applied iteratively to ensure that responsibility and accountability persist over time.
Further research is needed to provide an understanding of these dynamics of
responsible and accountable algorithmization.
In sum, this chapter provides a basic understanding of the importance of respon­
sibility and accountability in producing citizen trust in algorithmization. The key
message is that organizations should not only look at designers of algorithms and
expect that they will bring the solution: public organizations need to take action
to organize responsible and accountable use of algorithms in their organizational
processes.

References
Bovens, M. 2007. Analysing and assessing accountability: A conceptual framework. European
Law Journal, 13 (4): 447–468. doi: 10.1111/j.1468-0386.2007.00378.x
Brynjolfsson, E. and A. McAfee. 2014. The second machine age:Work, progress, and prosperity in
a time of brilliant technologies. New York:WW Norton & Company.
64 Albert Meijer and Stephan Grimmelikhuijsen

Burrell, J. 2016. How the machine ‘thinks’: Understanding opacity in machine learning
algorithms. Big Data & Society, 3 (1): 1–12. doi: 10.1177/2053951715622512
Cicirelli, F., A. Guerrieri, C. Mastroianni, G. Spezzano and A.Vinci. (Eds.). 2019. The internet
of things for smart urban ecosystems. Cham: Springer.
Dencik, L., Redden, J., Hintz, A. and H. Warne. 2019. The ‘golden view’: Data-driven gov­
ernance in the scoring society. Internet Policy Review, 8 (2): 1–24.doi: 10.14763/2019.2.1413
Diakopoulos, N. 2016. Accountability in algorithmic decision making. Communications of the
ACM, 59 (2): 56–62. doi: 10.1145/2844110
Eubanks, V. 2018. Automating inequality: How high-tech tools profile, police, and punish the poor.
New York: St. Martin's Press.
Friedman, B., P.H. Kahn and A. Borning. 2008.Value sensitive design and information sys­
tems. In The handbook of information and computer ethics, edited by K. Himma and H.Travani,
69–102. New Jersey: John Wiley & Sons Inc.
Fukuyama, F. 1995. Trust:The social virtues and the creation of prosperity. New York: Simon and
Schuster.
Gerards, J. 2019. The fundamental rights challenges of algorithms. Netherlands Quarterly of
Human Rights, 37 (3): 205–209. doi: 10.1177/0924051919861773
Grimmelikhuijsen, S. 2012. Transparency and trust: An experimental study of online disclosure and
trust in government. Doctoral dissertation. Utrecht: University Utrecht.
Grimmelikhuijsen, S. and E. Knies. 2017.Validating a scale for citizen trust in government
organizations. International Review of Administrative Sciences, 83 (3): 583–601. doi:
10.1177/0020852315585950
Grimmelikhuijsen, S., F. Herkes, I. Leistikow, J.Verkroost, F. de Vries and W.G. Zijlstra. 2019.
Can decision transparency increase citizen trust in regulatory agencies? Evidence from a
representative survey experiment. Regulation & Governance. doi: 10.1111/rego.12278
Grootelaar, H.A. and K. van den Bos. 2018. How litigants in Dutch courtrooms come to
trust judges: the role of perceived procedural justice, outcome favorability, and other
sociolegal moderators. Law & Society Review, 52 (1): 234–268. doi: 10.1111/lasr.12315
Hardin, R. 1999. Do we want trust in government? In Democracy and trust, edited by
M.E. Warren, 22–41. Cambridge: Cambridge University Press.
Hardin, R. 2002. Trust and trustworthiness. New York: Russell Sage Foundation.
Hetherington, M.J. 1998. The political relevance of political trust. The American Political
Science Review, 92 (4): 791–808. doi: 10.2307/2586304
Hoffman, A.L. 2019.Where fairness fails: Data algorithms, and the limits of the antidiscrimi­
nation discourse. Information, Communication & Society, 22 (7): 900–915. doi: 10.1080/
1369118X.2019.1573912
Hood, C. 2010. Accountability and transparency: Siamese twins, matching parts, awkward
couple?. West European Politics, 33 (5): 989–1009. doi: 10.1080/01402382.2010.486122
Inglehart, R. 1999. Trust, well-being and democracy. In Democracy and trust, edited by
M.E. Warren, 88–120. Cambridge: Cambridge University Press.
Janssen, M. and J.Van den Hoven. 2015. Big and open linked data (BOLD) in government:
A challenge to transparency and privacy?. Government Information Quarterly, 32 (4):
363–368. doi: 10.1016/j.giq.2015.11.007
Kleinberg, J., H. Lakkaraju, J. Leskovec, J. Ludwig and S. Mullainathan. 2017. Human deci­
sions and machine predictions. The Quarterly Journal of Economics, 133 (1): 237–293. doi:
10.1093/qje/qjx032
Kroll, J.A., S. Barocas, E.W. Felten, J.R. Reidenberg, D.G. Robinson and H. Yu. 2017.
Accountable algorithms. University of Pennsylvania Law Review, 165 (3): 633–706.
 Responsible and accountable algorithmization 65

Lepri, B., N. Oliver, E. Letouzé,A. Pentland and P.Vinck. 2018. Fair, transparent, and account­
able algorithmic decision-making processes. Philosophy & Technology, 31 (4): 611–627. doi:
10.1007/s13347-017-0279-x
Levi, M. and L. Stoker. 2000. Political trust and trustworthiness. Annual Review of Political
Science, 3: 375–507. doi: 10.1146/annurev.polisci.3.1.475
Lorenz, L. 2019. The algocracy: Understanding and explaining how public organizations are shaped
by algorithmic systems. MSc Thesis, Utrecht University.
Mayer, R., J.H. Davis and F.D. Schoorman. 1995. An integrative model of organizational
trust. Academy of Management Review, 20 (3): 709–734. doi: 10.5465/amr.1995.9508080335
McEvily, B. and M. Tortoriello. 2011. Measuring trust in organisational research: Review
and recommendations. Journal of Trust Research, 1 (1): 23–63. doi: 10.1080/
21515581.2011.552424
McKnight, D.H.,V. Choudhury and C. Kacmar. 2002. Developing and validating trust mea­
sures for e-commerce: An integrative typology. Information Systems Research, 13 (3):
334–359. doi: 10.1287/isre.13.3.334.81
Meijer, A. 2014. Transparency. In The Oxford handbook of public accountability, edited by
M. Bovens, R.E. Goodin and T. Schillemans, 507–524. Oxford: Oxford University Press.
Meijer, A. and M. Wessels. 2019. Predictive policing: Review of benefits and drawbacks.
International Journal of PublicAdministration,42 (12):1–9.doi:10.1080/01900692.2019.1575664
Mingers, J. and G. Walsham. 2010. Toward ethical information systems: the contribution of
discourse ethics. MIS Quarterly, 34 (4): 833–854. doi: 10.5555.2017496.2917505
O’Neil, C. 2016. Weapons of math destruction. How big data increases inequality and threatens
democracy. New York: Crown.
Owen, R., P. Macnaghten and J. Stilgoe. 2012. Responsible research and innovation: From
science in society to science for society, with society. Science and Public Policy, 39 (6),
751–760. doi: 10.1093/scipol/scs093
Pencheva, I., M. Esteve and S.J. Mikhaylov. 2018. Big data and AI: A transformational shift
for government: So, what next for research?. Public Policy and Administration, 35 (1): 24–44.
doi: 10.1177/0952076718780537
Porumbescu, G.A. and S. Grimmelikhuijsen. 2018. Linking decision-making procedures to
decision acceptance and citizen voice: Evidence from two studies. The American Review of
Public Administration, 48 (8): 902–914. doi: 10.1177/0275074017734642
Ramalho, M.A., R.J. Rossetti and N. Cacho. (2017). Towards an architecture for smart garbage
collection in urban settings. Presented at the 2017 International Smart Cities Conference (ISC2),
Wuxi, September 1–6.
Roberts, J. 1989. Aristotle on responsibility for action and character. Ancient Philosophy, 9 (1):
23–36. doi: 10.5840/ancientphil19899123
Rousseau, D.M., S.B. Sitkin, R.S. Burt and C. Camerer. 1998. Not so different after all: a
cross-discipline view of trust. Academy of Management Review, 23 (3): 393- 404. doi:
10.5465/amr.1998.926617
Tyler, T.R. 2006. Why people obey the law. Princeton: Princeton University Press.
Tyler,T.R. and P. Degoey. 1996.Trust in organizational authorities:The influence of motive
attributions on willingness to accept decisions. In Trust in organizations: Frontiers of theory
and research, edited by R. Kramer and T. Yler, 331–356. Thousand Oaks, CA: Sage
Publications.
Van den Bos, K., R.Vermunt and H.A.Wilke. 1997. Procedural and distributive justice:What
is fair depends more on what comes first than on what comes next. Journal of Personality
and Social Psychology, 72 (1), 95.
66 Albert Meijer and Stephan Grimmelikhuijsen

Van Gunsteren, H.R. 1976. The quest for control: A critique of the rational-central-rule approach in
public affairs. London: John Wiley & Sons.
Van Gunsteren, H.R. 2015.The ethical context of bureaucracy and performance analysis. In
The public sector: Challenge for coordination and learning, edited by F.-X. Kaufmann, chapter 15.
Berlin:Walter de Gruyter.
Van den Hoven, J. 2013. Value sensitive design and responsible innovation. In Responsible
innovation, edited by R. Owen, J. Bessant and M. Heintz, 75–84. Chichester, UK:Wiley.
Van der Voort, H.G., A.J. Klievink, M. Arnaboldi and A.J. Meijer. 2019. Rationality and poli­
tics of algorithms. Will the promise of big data survive the dynamics of public decision
making? Government Information Quarterly, 36 (1): 27–38. doi: 10.1016/j.giq.2018.10.011
Wieringa, M. 2020. What to account for when accounting for algorithms. A systematic literature
review on algorithmic accountability. Paper presented at ACM Conference on Fairness,
Accountability, and Transparency (ACM FAT*), Barcelona. ACM, New York, NY, USA,
January 27–30.
Zouridis, S., M. van Eck and M. Bovens. 2020. Automated discretion. In Discretion and the
quest for controlled freedom, edited by T. Evans and P. Hupe, chapter 20. London: Palgrave
Macmillan.
Zuurmond, A. 1994. De infocratie. Een theoretische en empirische heroriëntatie op Weber's ideaaltype
in het informatietijdperk. Den Haag: Phaedrus.
Zuurmond, A. 1998. From bureaucracy to infocracy: Are democratic institutions lagging
behind? In: Public administration in an information age:A handbook, edited by I.Th.M. Snellen
and W.B.H.J. van de Donk, chapter 16. Amsterdam: IOS Press.