See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/349812823

Improving the State-Tracking Ability of Corona Dashboards

Article in European Journal of Information Systems · March 2021

DOI: 10.1080/0960085X.2021.1907235

CITATIONS READS

14 1,194

1 author:

Jan Recker
University of Hamburg
352 PUBLICATIONS 13,431 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jan Recker on 10 March 2021.

The user has requested enhancement of the downloaded file.

 Improving the State-Tracking Ability of Corona Dashboards

Jan Recker
Faculty of Management, Economics, and Social Sciences
University of Cologne
e-mail: jan.recker@wiso.uni-koeln.de

Abstract:

Corona dashboards are interactive geospatial information systems used by billions of users to

help them understand the evolution of the COVID-19 pandemic. I use a representational lens

to explore how these systems can be made more useful. With this lens, the usefulness of these

systems to convey information about the pandemic fundamentally depends on whether these

systems are implemented as representation or state-tracking systems. I suggest that corona

dashboards presently focus disproportionally on representing socially constructed properties

(infection rates, deaths, levels of vaccination) of various things such as people, regions, or

countries. They would become more useful if they additionally focused on tracking events

(such as policy implementations) and changes in states (such as capacities of lockdown wards,

usage of face masks). By applying a methodology for design science research involving

design archaeology, I analyse the in situ implementation of Germany’s RKI COVID-19-

Dashboard, develop new design principles to extend the state-tracking abilities of corona

dashboards, and explore the importance, actability, and effectiveness of these design

principles through an empirical case study. The contributions this paper makes are new and

validated design principles for new feature implementations that can help making corona

dashboards more effective and useful.

Keywords: geospatial information systems, dashboard, representation, state-tracking,

design principles, design science research

 This article has been accepted for publication in the European Journal of Information Systems,
published by Taylor & Francis, after three rounds of peer review (acceptance date 5 March 2021).

BIOGRAPHY

Jan Recker is AIS fellow, Alexander-von-Humboldt fellow, chaired professor of Information

Systems and Systems Development at the University of Cologne, and adjunct professor at

Queensland University of Technology. His research focuses on systems analysis and design,

digital innovation and entrepreneurship, and digital solutions for sustainable development

goals.

ACKNOWLEDGMENTS

I am indebted to the senior editor, Pär Ågerfalk, and two anonymous reviewers, for

constructive and developmental feedback that helped improve the paper. I am thankful to Esri

Germany, in particular Dr. Gerd Buziek, for participating in the study and assisting with case

access. All faults remain my own.

INTRODUCTION
"We built this dashboard because we think it is important for the public to have an
understanding of the outbreak situation as it unfolds with transparent data sources."
Lauren Gardner
Co-director, Johns Hopkins Center for Systems Science and Engineering

After the local outbreak of COVID-19 in Wuhan (Hubei, China) in December 2019,

the virus spread throughout China and 27 other countries by February 2020. The first

European case was reported in France on 24 January 2020. On 30 January 2020, the World

Health Organization declared the COVID-19 outbreak a global public health emergency.

This trajectory was accompanied by the development of a new type of information

system (IS), online interactive dashboards that display the location and number of confirmed

COVID-19 cases, deaths, and recoveries for affected regions on a geospatial map (Dong et al.,

2020). One of the first such dashboards was made public by the Center for Systems Science

and Engineering at Johns Hopkins University (https://coronavirus.jhu.edu/map.html) on

January 22, 2020. Similar dashboards, such as Germany’s RKI COVID-19-Dashboard

(https://corona.rki.de), or the WHO Coronavirus Disease Dashboard

(https://covid19.who.int/), followed soon after. Since the onset of COVID-19, these

dashboards have been used not only by policy makers (Lazzerini & Putoto, 2020) and health

professionals (Reeves et al., 2020) but also by billions of citizens worldwide (Rogers, 2020;

Stafford, 2020) as they seek information about the outbreak of the virus and attempt to make

sense of new regulations and measures taken by governments that affect them.

These “corona dashboards” illustrate an important role that IS play in handling global

crisis situations: that of representation. Representation is fundamental to understanding a

crisis and managing it (Adam, 2020b). As the outbreak evolved, a manual reporting process

quickly became unsustainable. As the virus spread across Chinese regions, then Asian

countries, then continents, it became impossible to track cases through direct observation
without the help of IS that would collect, aggregate, and report digital data streams from a

variety of sources.

Of course, representation is not the only role of IS in global crisis situations. Since the

onset of the COVID-19 pandemic, we have come to witness how IS are designed and used to

enable contact tracing (Trang et al., 2020), telework (Carillo et al., 2021), virtual collaboration

(Waizenegger et al., 2020), crowd monitoring (Adam et al., 2020), contagion control

(Urbaczweski & Lee, 2020), or analytics (Pietz et al., 2020). At the same time, IS are also

involved in issues surrounding the pandemic, such as cybercrime (Naidoo, 2020),

cyberchondria (Laato et al., 2020), or fake news (O'Connor & Murphy, 2020).

While studies such as the above have dealt with these roles, positive and negative,

there is little work that has focused on the most basic but perhaps also most fundamental role

of how to represent information about the COVID-19 pandemic. This is puzzling because IS

research has dealt with fundamental questions of representing and modelling information

since the dawn of the discipline (e.g., Batra & Antony, 1994; Kent, 1978; Langefors, 1973;

Rogers, 1986; Stamper, 1971; Wand & Weber, 1990). Our field possesses an excellent

repertoire to reason about representation and how to make information systems more effective

(Burton-Jones & Grange, 2013; Burton-Jones et al., 2017).

My proposition is that the presently available corona dashboards are not effective as

they could be because they mainly provide functionality for representation1 but not state-

tracking2. The key insight here is that a pandemic, the worldwide spread of a virus (World

Health Organization, 2010), is neither an object nor an event but instead a process, a

progression of events and actions that unfolds over time (Abbott, 2016; Tsoukas & Chia,

1
Allowing human users to extract meaning about a real-world domain through symbols that convey
information about the things in that domain, their states, and properties (Weber, 1997).
2
Maintaining an accurate and complete representation of focal things in a real-world domain as events
occur in the real world that change the state of these things (Wand & Weber, 1995).
2002). To understand process, we must track changes in states and the events that cause these

(e.g., changes in infection rates, implementations of non-pharmaceutical interventions such as

lockdowns, or breakthroughs in the development of pharmaceutical interventions such as

medications or vaccines), instead of only representing things and their properties (e.g., the

number of infected people in different countries and their mortality rates), which are the

momentary consequences of such events and actions.

To expand on this basic proposition, I follow guidelines for research conduct (Peffers

et al., 2007) and artefact analysis (Chandra Kruse et al., 2019) from the design science

research tradition (Hevner et al., 2004). I proceed as follows: I first introduce the research

context, corona dashboards. I then explain my research approach. Next, I analyse a situated

implementation of corona dashboards, Germany’s RKI COVID-19-Dashboard

(https://corona.rki.de), and then develop design principles that extend the state-tracking ability

of corona dashboards by drawing on ideas from representation theory (Weber, 1997). I then

report on findings from an empirical case study that has the aims to evaluate the design

principles suggested in terms of importance, actability, and effectiveness (Iivari et al., 2020)

and to explore why and how design principles are implemented in the in situ dashboard, or

not. I conclude with a discussion of findings, implications, and limitations. The main

contribution of this article is that it reports the first artefactual and empirical study (Ågerfalk

& Karlsson, 2020) of corona dashboards, the most frequently used type of IS during the

Covid-19 pandemic (Rogers, 2020).

RESEARCH CONTEXT: GEOSPATIAL INTERACTIVE DASHBOARDS TO

TRACK THE OUTBREAK OF THE COVID-19 VIRUS

Corona dashboards are geospatial interactive information systems that semi-

automatically collect and display information about the existence and spread of the Covid-19

pandemic. These dashboards were developed with the aim to “provide researchers, public
health authorities, and the general public with a user-friendly tool to track the outbreak as it

unfolds” (Dong et al., 2020, p. 533).

Corona dashboards exist in a variety of formats (Datta, 2020) but most of them,

including Germany’s RKI COVID-19-Dashboard, the COVID-19 Global Map in the US, or

the WHO Coronavirus Disease (COVID-19) Dashboard, are representationally similar. They

use a variety of attributes, such as cumulative cases, recoveries, new infections, or reported

deaths, to represent the pandemic. They present these attributes geospatially, for example, by

country, region, or city. Most popular corona dashboards rely on the same technology,

ArcGIS (Esri, 2020), a cloud-based mapping, analysis, and data storage system, that can be

used to create, share, and manage maps, scenes, layers, apps, and other geographic content

(Scott & Janikas, 2010).

In what follows, my focus is specifically on a prominent implementation of corona

dashboards in Germany, the RKI COVID-19-Dashboard, which displays data provided by the

Robert-Koch-Institut (RKI), Germany’s main public health institute and one of the oldest

biomedical research institutes in the world. In Germany, the RKI plays a central role in the

national strategy for responding to the pandemic outbreak (Stafford, 2020). The RKI is

responsible for nationwide health monitoring and health reporting of the federal government.

Furthermore, it collects and interprets epidemiological data as a result of the Protection

against Infection Act (Infektionsschutzgesetz, IfSG).

Figure 1 presents an annotated screenshot of Germany’s RKI COVID-19-Dashboard.

The dashboard is divided into five main representation elements from top to bottom and left to

right. It reports three key figures, infections, deaths, and recoveries, in total and by day (top

element on right hand side of Figure 1). It also offers numerical (left hand side element of

Figure 1) and visual (middle element in Figure 1) representations of these figures by state and

population density, through drill-down functionality. It also decomposes the figures by

demographic attributes such as age group and gender (middle element on right hand side of

Figure 1). Because of disputes over the timeliness of reported data in various pandemic IS

(e.g., Cornish et al., 2020), the dashboard also represents the date of infection separate from

the date of record generation (bottom element on right hand side of Figure 1).

Infections, deaths, and

estimated recoveries,
in total and by day

Infections by age group

(bars) and gender
(color)

Infections and deaths, Date of infection (blue)

by German states and date of record
generation (yellow)

Infections relative to
population by German
states (color coded)

Figure 1: Screenshot of Germany’s RKI COVID-19-Dashboard (https://corona.rki.de, from 16

June 2020), with annotations

Corona dashboards are likely the most intensively used IS during the global pandemic.

For example, the COVID-19 dashboard at Johns Hopkins University reportedly hosts three to

five billion interactions every day (Rogers, 2020). But the use of dashboards is not stable, it

varies as the pandemic progresses. Figure 2 shows visits to the German RKI COVID-19-

Dashboard homepage and the number of new monthly infections with COVID-19 between

January and October 2020. Around the time the Covid-19 outbreak entered Europe in January

2020 and led to what is now known as the “first wave of the epidemic” in Europe, roughly

from February to May 2020 (Flaxman et al., 2020), total visits to Germany’s RKI COVID-19

web page increased from 630,000 (December 2019) and 1,700,000 (January 2020) to

6,200,000 (February 2020) and 66,600,000 visitors in March 2020. As the so-called “second
wave” started in Europe around October 2020 (Looi, 2020), visits to the dashboard homepage

again started to mirror the rise in infections.

70,000,000 350,000

60,000,000 300,000

50,000,000 250,000

40,000,000 200,000

30,000,000 150,000

20,000,000 100,000

10,000,000 50,000

,,0 ,0
Jan‐20 Feb‐20 Mar‐20 Apr‐20 May‐20 Jun‐20 Jul‐20 Aug‐20 Sep‐20 Oct‐20

Visits to rki.de web page (left‐hand y‐axis)

New monthly Covid‐19 infections in Germany (right‐hand y‐axis)

Figure 2: Visits to the RKI COVID-19 homepage and monthly COVID-19 infections in
Germany from January to October 2020. Data from (SimilarWeb, 2020) and (Robert Koch
Institut, 2020).

The number of reported interactions with corona dashboards might be seen as an

indicator that these systems are perceived as useful. However, the verdict about them is not

unequivocal, in part because different corona dashboards vary in accuracy, timeliness, focus,

and scope of information conveyed. In fact, guidelines have been developed to help users

evaluate different dashboards (Datta, 2020). In the political discourse and public media (e.g.,

Carthaus, 2020; Minji, 2020), several questions have been posed:3 How accurate is the

information represented? How is the data represented? Why do time lags exist in the

representation (e.g., every weekend we see dropping numbers of new infections, mainly

because of closed healthcare institutions)? Which representations, graphical or otherwise,

3
Some points of debate are not representational, they are informational. For example, some countries report
only selected data volumes, perhaps with an intent to “look more favorably” in comparison to others
(Pundir, 2020). Misinformation is an important issue (Laato et al., 2020) but not the topic of this paper.
appropriately convey the information that users seek? What is not represented accurately (e.g.,

number of negative tests) and to what extent is technical infrastructure to blame?

The proposition I develop is that corona dashboards do not effectively represent the

pandemic. They should not only represent attributes, such as infections and deaths, but also

track the evolution of these and other attributes over time in response to events that occur that

change the course of the evolution. The ability of a corona dashboard for state-tracking

becomes ever more important as the pandemic progresses because not only state variables

differ by regions but also events that occurred. In Europe, the period March to April was

characterized by a variety of governmental lockdown measures that differed across European

states (from stricter measures in Italy, Spain, or Germany, to less strict measures taken for

example in Sweden). Between May and June 2020, governmental lockdown restrictions in the

same regions were partially and stepwise released, but again differently across nations, states,

and even regions or cities (European Centre for Disease Prevention and Control, 2020). As

Europe has been dealing with a more severe second wave since fall 2020 (Looi, 2020) and

trying to anticipate if additional waves are yet to come (Flaxman et al., 2020), tracking both

events and states, and their relationships, will become ever more relevant.

RESEARCH APPROACH

I apply a staged research process informed by the methodology for design science

research by Peffers et al. (2007). It consists of three main steps (Figure 3). My entry point is

context-initiated (Peffers et al., 2007, p. 56) in that I start by observing the current practical

solution for Covid-19 tracking in the form of the existent implementation of a corona

dashboard in Germany, with the view to ascertaining the level of utility currently

accomplished by the artefact. This analysis is followed by developing new objectives for a

better artefact in the form of four new design principles for state-tracking, and then an

evaluation of these design principles with design practitioners. While presented in nominal
sequence, the three main steps were executed in iterative and logically connected fashion. For

example, the analysis of Germany’s RKI COVID-19-Dashboard proceeded logically

interwoven with the development and formalization of design principles for state-tracking.

Likewise, during the case study evaluation, insights were gained not only about the design

principles that specify the solution objectives but also about the decisions that led to the

artefact implementation as in-use today. In what follows, I explain key each main step briefly.

Iterative and logically connected process

1. Analysis of 2. Definition of 3. Evaluation

situated objectives of a
implementation solution

What does the What would a How important,

current artefact better artefact actable, and
accomplish? accomplish? effective are the
principles for
developing a
better artefact?

Context-
initiated
entry

Figure 3: Research process

Analysis of a situated artefact implementation. First, I begin by engaging in a type

of design archaeology (Chandra Kruse et al., 2019), that is, the analysis of an existing, in situ

corona dashboard artefact-in-use. My goal is to understand how and why present corona

dashboards that have been made available to the public operate the way to do and to evaluate

what they are able to accomplish representationally. I examine a current situated

implementation of corona dashboards, Germany’s RKI COVID-19-Dashboard. It is a

representative case of corona dashboards as it builds on the same technology (in particular

ArcGIS online) as most other dashboards. The basis of my analysis is what is called an

interpretation mapping (Wand & Weber, 1993), that is, a mapping of representational

elements featuring on the dashboards to a set of theoretical concepts to describe real-world

domains in terms of things that are of relevance, their properties, and the events that occur

that change the states of these things. A detailed description of these constructs is provided by

Weber (1997). Procedural guidelines for such a mapping are described by Rosemann et al.

(2009).

Definition of objectives of a solution. The analysis carried out in step one identifies a

lack of state-tracking ability in presently available corona dashboards. The dashboards only

offer a partial representation of the pandemic and its impacts, which could result in

misinformation, ineffective medical interventions, or lack of acceptance of non-

pharmaceutical public health interventions. The defined objective in step two therefore is to

develop and verify design principles that could extend the state-tracking ability of corona

dashboards. I specify these objectives by deriving design principles for state-tracking from

representation theory (Weber, 1997). Design principles are a form of prescriptive design

knowledge that capture a general solution in a class of artefacts, which can guide designing in

a wider range of problems and solutions and are thus valuable outcomes as theoretical

contributions for IS scholars (Baskerville et al., 2018) and as guidelines for practitioners to

design similar artefacts (Iivari et al., 2020). To make the design principles understandable and

useful, I specify them according to a schema provided by Gregor et al. (2020).

Evaluation. Third, to evaluate the extent to which the proposed design principles

could assist developing a new artefact that provides better state-tracking ability, I collected

empirical data about the design and operation of Germany’s RKI COVID-19-Dashboard.

Details are provided in the Appendix. Key case study informants are stakeholders involved in

design activities relevant to the dashboard (i.e., development, implementation, and operation),

such as solution architects, GIS developers, user interface designers, or geomatics specialists,

rather than end users of the dashboards (e.g., policy makers, health professionals, or members

of the general public) because the different designer roles are the main stakeholders that

would be concerned with implementing the proposed design principles (Iivari et al., 2020).
End users would only be able to report on consequences from these design choices (e.g., in

terms of changed perceptions of usefulness). I examine the relevance and applicability of the

design principles in terms of importance, actability, and effectiveness, following the

suggestions and template by Iivari et al. (2020).

ANALYSIS OF GERMANY’S RKI COVID-19-DASHBOARD

My starting point was an analysis of the present corona dashboard that is in-use in

Germany and maintained by the RKI. To evaluate the abilities of Germany’s RKI COVID-19-

Dashboard for representing information about the Covid-19 pandemic, I use a conceptual

lexicon that is grounded in representation theory (RT) (Wand & Weber, 1990, 1993, 1995;

Weber, 1997). Term definitions and reviews of the RT literature are provided by Burton-Jones

et al. (2017) and Recker et al. (2019).

The fundamental idea of RT is that IS are built and used because it can be more

efficient to learn about the world from computerized representations than by direct

observation (Burton-Jones & Grange, 2013, p. 636; Wand & Weber, 1993, p. 218). In

situations like the current global pandemic, one could argue that IS are the only feasible way

to learn about the pandemic – we cannot directly observe the spread of the pandemic as it

unfolds across the planet.

I use the RT lexicon in my artefact analysis because it distinguishes between

representation, that is, the ability of an IS to provide a faithful representation of focal real-

world phenomena (Wand & Weber, 1995, p. 207) and state-tracking, that is, the ability of an

IS to remain a faithful representation of the focal real-world phenomena as things in the real

world undergo change. Following RT, a useful IS for managing a pandemic is one that

faithfully represent all information users seek about the pandemic (the focal real-world

phenomena) and its effects on other real-world phenomena around us. This means, the IS

must feature representations for relevant things affected by the pandemic in the real world
(e.g., elderly people, health professionals, or children), systems comprised by them (e.g.,

families, schools, elderly care facilities), their properties (e.g., existing health conditions),

states (e.g., being infected, contagious, vaccinated, or recovered), and so forth.

It is important to note that alternative analysis lexica could just as well be used

(Ågerfalk & Eriksson, 2004). For example, Searle (1995) also recognizes the existence of an

objective world in which citizens could become infected, develop immunity, or die, from a

virus. Scholars working with this lexicon would speak of asserting “substantial properties” of

“concrete objects” (March & Allen, 2014, p. 1349) rather than representing “intrinsic

properties” of “things” (Weber, 1997, pp. 34-35). Yet, both the scholars working with

Searle’s lexicon and the scholars working with the RT lexicon recognize that the meaning of

data represented in information systems is socially constructed. For example, Wand and

Weber (1995, p. 206) note that information systems represent the meaning of some real-world

phenomenon as perceived by someone or some group. This is important because corona

dashboards display data that is not physically factual but socially constructed. For example,

the volume of Covid-19 fatalities they report depends on the definition of Covid-19 associated

fatalities agreed upon in that country (Sorci et al., 2020), which implies that the representation

of such social facts Searle (1995) could carry communicative meaning (Ågerfalk & Eriksson,

2004). However, a Searleian perspective would additionally recognize that representation (of

assertives) is only one of several functions of a language (Eriksson & Ågerfalk, 2021). Still,

for the class of systems that are Corona dashboards, the emphasis is clearly on assertives and

thus the fundamental representational question remains the same across the different schools

of thought: How can we model socially constructed information about the real world within

an IS such that the system can be effectively used?

Table 1 summarizes my analysis of the main representational elements used in

Germany’s RKI COVID-19-Dashboard to convey information about the pandemic (Figure 1)

 in terms of their usefulness for representation (the ability to observe the state of things, such

as a person or object) or state-tracking (the ability to follow a trail of changes over time).

Table 1. Analysis of the Main Representational Elements in Germany’s RKI COVID-19-Dashboard

Dashboard What is What information is What is the representation

element represented? conveyed? useful for?
Top element on Total infections, Numerical attributes convey For representation: they
right hand side deaths, and the values of properties in summarize the state of a collection
of Figure 1 recoveries (as general of the population of of people (Germany’s population)
numbers). Germany (i.e., a system of at some point in time (the time of
things that share non-binding visit).
mutual properties).
Middle element Total infections Numerical and visual attributes For representation: they
on right hand (as numbers), by convey the values of summarize the state of a subset of
side of Figure 1 age group and properties in general of some the collection of people
gender (categories subsets of the population of (Germany’s population) that are of
color coded). Germany. particular interest, because of
presumed risk of infection, at some
point in time (the time of visit).
Bottom Sum of daily Numerical and visual attributes For state-tracking: The inclusion
element on infections and convey the values of one of temporal event data (successive
right hand side daily reported property in general (infections) dates) allows following the change
of Figure 1 data, by date (as of the population of Germany in infection or reporting data over
numbers, the two by event (dates). time.
types of data are
separated visually
through color
cording).
Left hand side Total infections, by Numerical attributes convey For representation: they
element in state. the values of one property in summarize the state of the
Figure 1 general (infections) in different collection of people (Germany’s
subsets (states) of the population) decomposed into sub-
population of Germany. sets (by state), at some point in
time (the time of visit).
Middle element Total infections Visual attributes convey the For representation: they
in Figure 1 relative to values of one property in summarize the state of the
population general (infections) in two collection of people (Germany’s
(categorized different subsets of the population) decomposed into sub-
through color population (state and sets (by state and population
coding), by state population density) of density), at some point in time (the
(visual). Germany. time of visit).

The analysis suggests that the present corona dashboard implementation contains more

elements associated with representation than state-tracking. In terms of representation,

Germany’s RKI COVID-19-Dashboard allows drilling up and down to states, regions, or

municipalities (properties of geographical collectives) or by gender and age (properties of

demographic collectives) to examine state variables such as reported versus incurred

infections, population volume and density, recovered cases, or deaths. The state

representations are updated daily. However, no mechanism is discernible for selecting

additional or other epidemiological (e.g., reproduction number) or non-epidemiological (e.g.,

economic) attributes, such as job losses, GDP, or PMI (Haldane & Chowla, 2020).

In terms of state-tracking, the analysis suggests that the current implementation offers

only partial ability for state-tracking if any. Germany’s RKI COVID-19-Dashboard tracks

daily changes in three state attributes (reported infections, reported deaths, estimated

recoveries). Other lawful state changes (e.g., possibly recurring infections, actual length of

recoveries) are not tracked. Moreover, other possible lawful state transitions (i.e., changes that

could occur between attributes such as infections and death that are logically possible) that

could result in reported state attributes are not tracked either. For example, the dashboard does

not distinguish between deaths caused by COVID-19 versus different causes of death with

COVID-19 side effects. Moreover, Germany’s RKI COVID-19-Dashboard provides

timestamps for the state variables it displays but no representation is provided for any type of

external events that might be relevant to the evolution of the pandemic, such as the timing of

non-pharmaceutical interventions (Flaxman et al., 2020) in the form of lockdowns, border

closures, or contact restrictions, or – more recently – pharmaceutical interventions such as

vaccination rollouts (Limb, 2021).

DEFINITION OF SOLUTION OBJECTIVES: DERIVING DESIGN PRINCIPLES

FOR STATE-TRACKING

My analysis summarized in Table 1 identifies a lack of state-tracking ability in

presently available corona dashboards. To define a potential solution for a better artefact, I

again build on RT to develop four design principles that extend corona dashboards’ ability to

maintain an up-to-date, faithful representation as events occur that change the relevant real-
world phenomena. Table 2 presents four design principles for corona dashboards that I

derived deductively from four conditions for state-tracking suggested in RT (Weber, 1997, pp.

133-146). Each design principle is specified in terms of context, aims, mechanism, and

rationale, as per the schema suggested by Gregor et al. (2020). Because implementers

(providers of corona dashboards in organizations such as hospitals, public health institutions,

or governments), and users (policy makers, health professionals, local crisis response teams,

and members of the general public) are invariant across all four design principles, they are not

featured separately.
 Table 2. Design principles to extend the state-tracking ability of corona dashboards

Name Original state-tracking Relevance in the context of Aim, mechanism, and rationale of the design principle
condition (Weber, 1997) the Covid-19 pandemic

1. The Each state in the focal real- As the interest in the focal Aim: It must at all times be possible to observe relevant status indicators for
mapping world phenomena must map to phenomena associated with the particular collectives of people through the corona dashboard.
principle at least one state in the IS. Covid-19 pandemic changes
Mechanism: Ensure there is, and always remains, a 1:1 mapping between status
(e.g., from a purely
It must at all times be possible indicators that are considered relevant during a pandemic (e.g., total infections,
epidemiological interest to
to tell relevant states of the infections relative to population density, excess deaths, the reproduction number R,
include concerns about
focal real-world phenomena and so forth) and symbolic representations (e.g., text and graphics) available in the
economy, public psychic health,
based on a state of the IS. dashboard. Ensure the appropriate and automatic provision of relevant data for each
or social unrest), relevant states
identified status indicator.
that describe all relevant
phenomena must be mapped to Rationale: At different times over the course of the pandemic (e.g., before, during, or
an equivalent set of states after the first or second wave), and for different types of decisions, different
displayed in a corona parameters of the pandemic evolution are relevant: infections relative to population
dashboard. density might be relevant to traveling, reproduction might be relevant to lockdown
measures, total infections might be relevant to hospital infrastructure planning.

2. The State changes adhere to As knowledge about Aim: Corona dashboards must embody relevant transformation laws so that any
tracking transformation laws (e.g., transformation laws that pertain intermediate and resulting stable states in the focal real-world phenomena can be
principle resurrection is not lawful). to the Covid-19 pandemic faithfully represented by the IS. It should also be possible to project relevant future
States change (e.g., an infected changes through new states based on extant transformation laws.
person may recover, or die) discoveries (e.g., length of
Mechanism: Construct an algorithm that periodically, at reasonable and feasible time
because of events (e.g., incubation, contagion, or
intervals, evaluates the transformation laws in the IS against data about the state
hospitalization, care). recovery periods), it must be
changes in real-world things. For example, compare the algorithm that estimates the
possible to update the
When things in the focal real- number of recovered COVID-19 cases real-world data about the sequences of state
transformation laws in corona
world phenomena change changes that reported cases of patients underwent (He et al., 2020). If deviations
dashboards so that they
states as a result of events that occur, update the transformation laws in the IS that govern the estimation algorithm.
correctly update states (such as
are internal to the phenomena,
infections, vaccinations, or Rationale: Over time during the pandemic, more knowledge has been accrued about
the IS must change from a state
recoveries), and also project contagion periods, length of isolation periods required, or thresholds that indicate
that faithfully represents the
foreseeable future states (e.g., levels of relative infection by population density (in Germany: number of infection by
initial state of the thing to a
trends in infections, speed of 100,000 inhabitants). These laws are socially constructed and can change.
state that faithfully represents
recoveries, estimated deaths). Transformation laws in a corona dashboard must thus be made adaptable.
the subsequent state of the
thing.
3. The An external event in the focal For example, over the course of Aim: It must at all times be possible to use a corona dashboard to track external
external- real-world phenomena is a the pandemic, governments events (e.g., political interventions such as lockdowns, release of new technologies
event change of state that arises in implemented and stopped a such as vaccines, or change in season) and map changes in state variables (e.g.,
principle some thing in the phenomena variety of non-pharmaceutical infection rate, death rate, etc.) to the occurrence of those events.
by virtue of the action of some interventions (European Centre
Mechanism: Ensure that occurrences of relevant external event relevant to the focal
thing in the environment of the for Disease Prevention and
real-world phenomena are identified and can be reported to the corona dashboard.
phenomena. When an external Control, 2020) in response to the
Ensure symbolic representations (e.g., text and graphics) are available in the
event occurs in the focal real- trajectory of infections and
dashboard to convey the meaning of different events.4
world phenomena, an external deaths. Several countries
event that is a faithful including Russia and China have Rationale: The trajectory of a pandemic is not based on internal states (i.e.,
representation of the real-world started vaccination programs molecular and biological features) of the virus only, the distribution and frequency of
external event must occur (Cohen, 2020). Mass infections COVID-19 infections is an epidemiological process influenced by outside factors
within the IS. can occur through super- such as location (middle of Europe versus isolated island), season (spring versus
spreading events (Wong & fall), and the behaviours of social collectives and the design of social and technical
Collins, 2020). institutions that govern these behaviours. A corona dashboard must present changes
in the COVID-19 pandemic in coupling with those events that influence the epidemic.

4. The External events do not occur in A temporal sequence in the Aim: It must be possible to track the sequence of relevant events that occur. The
sequencing isolation but in a temporal and context of the Covid-19 sequence of external events in the focal real-world phenomena must match the
principle logical sequence. pandemic might be the stepwise sequence of external events in the IS.
Representations of external implementation of policy
Mechanism: Ensure that relevant event sequences are identified in the real-world
events in an IS must follow the interventions (e.g., Australia’s
and compare these to the sequences of state changes recorded in the IS. 5
same sequence as external three-step plan for removal of
events that occur in the focal lockdown measures). A logical Rationale: Pharmaceutical and non-pharmaceutical response interventions are not
real-world phenomena. sequence might the occurrence singular occurrences but instead interdependent temporal and logical sequences of
of collective action (e.g., public events. To understand their outcomes, it is important to track not only the events and
protests) in response to policy state changes but also their logical and temporal sequences.
implementations.

4
Such an algorithm could be executed manually, for example, by assigning new user roles who are made responsible for identifying and recording external events in the IS, or
semi-automatically, by scraping event information from digital chronologies (e.g., European Centre for Disease Prevention and Control, 2020).
5
Such an algorithm could be implemented automatically, for example, through event mining applications that identify event sequences from multimedia streams such as
political news (Xie et al., 2008) or from social networking sites such as twitter announcements (Aggarwal & Subbian, 2012).
 EVALUATION OF THE DESIGN PRINCIPLES IN A CASE STUDY

In the case study, I pursued two broad goals: first, to learn about how the design of

Germany’s RKI COVID-19 dashboard was completed (indicated in Figure 3 through an arrow

going back from evaluation to analysis), and second, to evaluate whether my design

propositions were viewed as important, feasible and relevant (indicated in Figure 3 through an

arrow going back from evaluation to definition). Design and procedures are explained in the

Appendix.

The Design of Germany’s RKI COVID-19-Dashboard

In February 2020, when the COVID-19 pandemic had started to impact Europe, the

then-available Johns Hopkins COVID-19 Dashboard in the United States created political

pressure in Germany to provide a similar technology to inform the German public. The RKI

and the German Federal Ministry of Health (BMG) tried to develop a dashboard solution

themselves but quickly realized that an industry-strength scalable and loadable infrastructure

for usage by 80 million citizens could not be developed within a matter of days. They decided

to implement and further develop a prototype created by a German taskforce of Esri, an

international supplier of geographic information system software, web GIS and geodatabase

management applications. RKI and Esri officially presented their solution on 20 March 2020,

two days after a tweeted picture featured federal minster of health Jens Spahn at a BMG

office with the Esri dashboard prototype already on the wall.6

Three attributes characterize the design of Germany’s RKI COVID-19-Dashboard.

First, the dashboard displays only data provided and verified by the RKI in fixed time slices

of 24 hours, consistent with other public information released by the RKI. It does not convey

6
https://twitter.com/BMG_Bund/status/1238447752935325698.
estimations or predictions. The data is hosted on an open data hub, NPGeo7, a national

platform for geodata analysis, combination, and visualization. This data platform supplies

data to the Germany’s RKI COVID-19-Dashboard but can also supply specialized

dashboards, for example those used by states, municipalities, or special interest groups such

as the German Association for Fire Protection, a non-profit expert network that unites parties

involved in civil protection, rescue and security.8

Second, RKI data is displayed at agreed upon geographical levels such as states,

counties, districts, and municipalities. This was difficult to implement because Germany’s

health data reporting infrastructure is federated and relies on all health department (about 400

in total) to input their data in a comparable, automatable format but on basis of their own

infrastructure and resourcing. One of the executive business development managers from Esri

commented:

“Our administrative structure for health care is a matter for the federal states and in
this respect it is something that is regulated for the federal states and the federal states
then decide together with the state registration offices for infections, which are located
at the health ministries, how they inform the population.”
Third, sovereignty for information displayed on the dashboard lies entirely with the

RKI. A specialist team at the RKI decides, which data to provide and display. The Esri

Germany team consults but final decisions rest with the RKI. In the words of an Esri

executive:

“The sovereignty of information lies with the RKI. Because that's where the specialists
are, the ones who are familiar with the phenomenon, who have been researching it for
decades. And who, in principle, also have ownership of the data and data
infrastructures. So, we must not forget that. And they also have the information
sovereignty, whereas here we [Esri] are the technology specialists. And that means
that a group discusses optimization at the RKI, and we then come together to discuss
whether this can be done and, if so, how it can be done and with what effort. And then,
of course, we also contribute our cartographic knowledge and then talk about issues
of cartographic design, user ergonomics, user interface design and so on.”
7
https://npgeo-corona-npgeo-de.hub.arcgis.com/.
8
www.vfdb.de/coronaampel.
 Evaluation of the State-Tracking Design Principles

The primary objective of the case study was to examine the relevance and applicability

of the four state-tracking design principles in terms of importance, actability, and

effectiveness (Iivari et al., 2020). Figure 4 visually summarizes the insights gained through

the case study on each of the design principles. Each design principle is positioned in terms of

reported level of importance (from low to medium and high), and estimated effectiveness

(from low to medium and high). The bubble size summarizes the feedback received on

actability of each design principle, with a small circle suggesting low actability and the larger

circle indicating medium actability (no design principle was rated as highly actable). All

design principles were rated as being of medium or higher importance. The visual summary of

the qualitative evaluation suggests in particular that the extern design principle 3 (external

event representation) was deemed both importance and effective if implemented. One

manager remarked:

“From a research point of view, it's clear that you need to just let one thing correlate
with the other and you see how it goes.”
Similarly, design principle 4 (sequencing) was deemed important and of medium

effectiveness if implemented. For example, a business development manager at Esri

commented:

“I think it is absolutely necessary to go in this direction, that one somehow depicts the
complexity with all its interactions. You can do that. […] if several of these events,
such as a lockdown and the obligation to wear masks, come together because they are
very close in time, and then the infection rates fall, then of course every citizen draws
a different conclusion and says, yes, this measure worked, although you do not really
know which of the measures had what effect. Well, that [would be] actually quite nice
to see.”
The two design principles that are presently at least partially implemented (DP1 and

DP2) were rated of medium-to-high importance and low-to-medium effectiveness. Together,

this evaluation suggests that especially those design principles that are presently not

considered in corona dashboards were seen as relevant additions.

 Legend DP3 (External events)
high
(I: high, A: low, E: medium-high)
Actability
(low, medium, high)
Effectiveness

DP4 (Sequencing)
(I: high, A: medium, E: low-medium)

DP1 (Mapping)
(I: medium-high, A: low-medium, E: low-medium)

DP2 (Tracking)
low

(I: medium, A: medium, E: low)

low high
Importance

Figure 4: Qualitative evaluation of case study feedback on the design principles

Reported actability of all design principles was markedly low. None of the design

principles was rated high in terms of actability. Three main reasons became salient through

the data analysis for why such features were difficult to implement at present:

Distribution of information sources. Relevant information in Germany is provided

through a federated administration and reporting system in which several hundred local

authorities collect, store, and report data in a largely unstandardized manner. With German

data regulations in place, a centralized information infrastructure is not yet existent from

which relevant data could be sourced for representation through a dashboard even though

digital means are already available. An Esri executive stated:

“What does not exist […] is that all these measures are reported to a central location,
so that in which county you know what measures have actually been taken for what
period. They don't have that. And you could map that. For example, through the
Federal Office for Civil Protection and Disaster Relief and the apps KATWARN and
NINA [German mobile apps for emergency and disaster control], which are operated
from there. But then again, a decree or an injunction must be brought about that the
 crisis staff decide - and that should be one in every county - then also pass this data to
a central location.
Possibility of automated extraction. A second challenge is the transmission of data

into the NPGeo hub. Relevant data is sourced from local health administration offices. While

all these offices display local information in some format (e.g., via their homepages), not all

of them make the data directly available in a digital let alone open format. Updating

information in the corona dashboard therefore regularly involves substantial manual effort or

relies on makeshift automation tools such as web scraping. An Esri business development

manager explained:

“The disaster control apps, you know NINA and KATWARN for example, they are fed
with this information, so that you are then informed. But not everyone has these apps.
So, the question is, how do you find out which state registration offices or regional
and local health authorities - we have more than 400 of them in Germany - how do
they inform their local population about the measures taken in a timely manner? Once
you know that, you have to ask the question: is this source of information digital and
can this information be accessed automatically? And at the moment it is not
automatable because you don't get a CSV table or anything else, you have to actively
read HTML pages.”
The experienced difficulties with feeding all relevant data to the dashboard led to the

initiation of a common data infrastructure project, DEMIS (Diercke, 2017), which, since June

2020, is successively replacing the extant reporting system with an online framework that

gradually connects approximately 170 laboratories and 400 health authorities and provides

electronic reporting, cooperation between health authorities, and provision of data for

evaluation and analysis.

Centralized information sovereignty. The third main reason for the lack of actability

on certain aspects of the design principles, such as additional state variables (e.g., economic

markers of the pandemic) or external events (e.g., political measures taken), resides in RKI

being the sole decision authority over the information conveyed through the corona

dashboard. As a public health institute, the RKI made the decision to restrict the scope of the

dashboard to selected pure epidemiological data, which renders the conveyance of additional,
non-epidemiological information (such as the number of requested or granted temporary

economic state aids) on Germany’s RKI COVID-19-Dashboard impossible even if deemed

implementable and relevant.

Tensions Influencing the State-Tracking Ability of Corona Dashboards

A second objective of the case study was to develop deeper insights why a particular

design principles was not deemed actable, and how a design principle, if implemented, was

designed into the in situ artefact. This analysis (Table A3) yielded three tensions:

A first tension concerns the existing federated legislation and administration versus

the need for a centralized information infrastructure. This tension played out primarily in

the recording and transmission of relevant data to be conveyed on a dashboard; however, it

also became evident in matters of IT resourcing. Depending on the size and funding of the

relevant local public and health authority, also the technological resourcing varies across the

country, in terms of availability of infrastructure, IT capabilities or support availability. One

of the regional district geo-managers, for example, explained:

“I know that in [a different district], for example, the dashboard is operated by a data
processing centre and not by the district administration. And the public health
department is included in the district administration. We don't have a data processing
centre, we only have an IT department, and we ourselves are more or less
independent, and we are also members of the crisis team. This means that we
ourselves can react quickly but do not have to access a data processing centre. […]
There is always the question: who maintains the data there?”
Moreover, not only is IT support for data collection and recording different across the

units of the federated system, also the decision-making about extent and format of disclosure

of such data varies by city, district, region, or state. Several districts, for example, chose not to

report COVID-19 deaths, others decided to report absolute but not relative numbers, or to

report (or not) statistics such as the reproduction number R (Adam, 2020a). In the words of

the business relations executive:

 “The problem at this point is again our administrative structure. Health care is a
matter for the federal states and in this respect it is something that is regulated for the
federal states and the federal states then decide together with the state registration
offices for infections, which are located at the health ministries, how they inform the
population.”
The main reason for the existence of this tension lies in German regulations that

prohibit interdepartmental responsibilities and data interoperability. Health data is collected in

federated departments, and no legal basis exists for hosting the data in a central repository.

Only in a declared disaster protection case is the Federal Office of Civil Protection and

Disaster Assistance entitled to combine and provide relevant tasks and information in a single

place.

A second tension emerged when discussing the inclusion of mechanisms for external

event and sequencing representation (DP3 and DP4) and addresses the balancing between

accurate historical data and uncertain forecast data. This tension becomes apparent

because through a potential inclusion of external event (and sequencing) representations, it

would be possible not only to display the unfolding of the pandemic in relation to those events

that occurred in the past but also in relation to predictions about how it could unfold if the

event would not occur (or some other event instead). Such prediction models are feasible and

also used for strategic planning (Flaxman et al., 2020) but case study participants expressed

concerns about how such prediction models, which are necessarily forecasted and with more

uncertainty than the presently reported historical and verified data, would be interpreted by

end users of the dashboards. A regional crisis management team coordinator stated:

“Of course, we also have this forecast data, but it has been shown, or was agreed, that
such data does not belong in a public dashboard. Only the pure facts should be
included, in order not to make oneself vulnerable to attack.”
This concern not only played out in the discussion about whether or not to include

forecasted, predicted data together with, or instead of, historical and verified data. It also

presented a general tension in terms of the confidence in data processing versus data

interpretation. A general tenet in the case study was palpable hesitance about providing
more, and more complex data, in fear of misinterpretation or misuse by the general public.

This was one noted reason why data about infection numbers was not correlated with

intervention events such as political measures. One of the geomatics specialists in a regional

public office commented:

“If the result now comes out, for example, "despite the use of a mask, the numbers are
going up again", that ultimately leads to people all saying "The mask is nonsense. I'll
leave it out". Whereby maybe the cause that it goes up was a completely different
one. […] Maybe there are reasons why the curve and the numbers go up anyway. That
would lead to misinterpretations among the citizens. In this respect, you have to weigh
more carefully what you are representing.”
A similar view was held by one of the GIS developers:
“In my opinion, two major aspects play a role here: The fear of politicians that the
data might show that a certain measure did not affect cases but only restricted daily
life. Secondly, the temporal coincidence of a change in state variables and an event
might lead to false claims about the impact of events because the effects of certain
events might be temporarily shifted.”
In this vain, often a conscious decision was taken not to report more complex

statistical information including forecasted data, uncertainty estimates, or in cases even

statistics used in the public and political debate (e.g., incidence thresholds or reproduction

numbers). One of the Esri executives stated:

“And so that means such uncertainty - we've discussed this, can one represent the

uncertainty of the numbers? It can be represented, but we know that the public in particular

cannot deal with it. […] The most important thing is official and really verified data, because

there is also enough fake news and whatever else is thrown around.”

DISCUSSION

Germany’s RKI COVID-19-Dashboard is presently more useful for representation

than for state-tracking. There are good reasons for this focus. During the early phase of the
pandemic outbreak in Europe, a pressing need was to understand the state of the pandemic

and the European continent. How many people are affected and where, and how many people

are dying from the virus and where, were important questions. Likewise, as a public institute

for nationwide health monitoring and reporting, the RKI understandably maintains their

decision to represent only epidemiological data but not political, economic, or social data

(e.g., incident level thresholds that are the basis of many political measures).

However, as we are witnessing in the public and political debate, the pandemic is not

just an epidemiological process, it has wide-reaching consequences on many other facets of

everyday life, from health to economic and social issues. Representationally, the “focal

phenomenon” that is the COVID-19 pandemic has been expanding gradually since its onset.

Providing a dashboard with the aim to help people understand the situation and its

consequences must thus inevitably be expanded in reach and coverage.

Even though the Covid-19 virus is a factual object with certain properties, some of

which we know and many of which we only begin to learn about, we should not think about

the pandemic as a physical thing to represent. The pandemic resembles a socially constructed

process – a logical and temporal progression of events and activities that unfold over time.

Any IS whose purpose it is to help understanding of the outbreak situation and managing the

pandemic must thus be helpful in (re-) constructing the processual narrative that reveals the

mechanisms by which the events and actions play out over time.

A continued lack of ability for state-tracking will over time undermine the intended

purpose and usefulness of corona dashboards. Early signs are clear. While useful during the

early phase of the pandemic’s onset in Europe, the system has become less useful with time as

more events and changes occurred in the real-world (e.g., political measures were taken, the

epicentres of the pandemic shifted, the number of intensive care units became overloaded in

certain regions, and capacities were booted in others) that were neither represented nor
tracked, in the dashboard. We all have come to realize that we cannot understand the

pandemic, or how to deal with it, by reducing our information to the number of infections,

fatalities, and recoveries alone. To anticipate and undertake actions such as restarting the

economy, balancing safety and well-being of citizens, rolling out vaccination programs, or

simply making decisions about inter-state travel for holiday purposes, we need to understand

how the pandemic evolves in relation to a variety of events and actions that have occurred and

continue to occur and interfere with the progression of the pandemic.

One key insight from the analysis of the corona dashboards and the empirical data in

the case study is that corona dashboards are only ever as effective as the underlying

information system infrastructure including its technological, administrative, and legislative

components. Presently, Germany’s main corona dashboard is maintained by the RKI, which is

entitled by the Protection against Infection Act to act as a single integrated source of health

data. But to make the dashboard a more effective state-tracking system for the general public,

information sovereignty must extend beyond the RKI such that non-epidemiological data

(e.g., economic state aids, unemployment rates, capacity of hospital beds, etc.) and key events

and their sequences can also be covered by the dashboard, especially in terms of non-

pharmaceutical interventions (lockdowns, contact restrictions, or school closures) and social

events (e.g., demonstrations, or super spreading events). However, relevant interdepartmental

connections beyond the health authorities do not legally exist, and data interoperability

between different ministerial departments is low. Efforts towards a national information

infrastructure must continue to develop a legal and technological framework for data

interoperability and automated reporting.

Another key insight is that the effectiveness of many regional variants of the RKI

Corona Dashboard is presently limited by local authorities’ hesitance to display complex or

uncertain data in fear of user issues such as lack of acceptance, misinterpretation, or abuse.
These issues are certainly real; however, they are issues of professional information

communication, not provision through dashboards.

Finally, the combination of an open data infrastructure being available through

NPGeo, the federated system of legislation and reporting, and different preferences for

presentation of data have led to the emergence of a multitude of corona dashboards in

Germany (some examples are listed in Table A1). With the number of dashboards increasing,

the risk of inconsistencies rises, as does the risk of selective (mis-) interpretation because

users can basically “choose” a dashboard that provides only those socially constructed

information that match subjective preferences. This situation will not help public acceptance

of pandemic measures, which is immediately relevant, for example, to the current rollout of

vaccination programs (Limb, 2021). Technologically, it would be possible to develop one

central corona dashboard, as was intended originally, and equip it with features such as view

filters or role-based access control such that different user groups (e.g., members of the

general public versus political decision-makers or local crisis response teams) can access only

those information most relevant to their tasks, whilst maintaining centralized control over data

verification or information display.

Implications for Research

The issue of representation versus state-tracking is neither new nor unique to the

pandemic. However, the present pandemic vividly demonstrates how important state-tracking

is. Thus, while the main contribution of this paper primarily artefactual (Ågerfalk & Karlsson,

2020, p. 111) and hopefully helps improving corona dashboards, it carries important

theoretical and empirical implications. Modifying existing corona dashboards such that they

implement state-tracking ability allows reflecting, evaluating, and amending theoretical ideas

that were initially proposed by Wand and Weber (1995) and have since largely been

forgotten. I doubt Wand and Weber (1995) had a pandemic in mind as a process that would
change the lives of many when they formulated their ideas. But the question begs whether

earlier attention to called state-tracking would have led to more IS yielding such abilities. IS

scholars have so far missed this opportunity. Burton-Jones et al. (2017, p. 1321) write:

“Our review showed researchers have not been interested in the STM [state-tracking
model by Wand and Weber 1995]. We suspect one reason is that its predicted outcome
(maintenance of a faithful representation) appears overly academic rather than
practically useful.”

The practical relevance of state-tracking is here now. The opportunity is to help

institutions to develop more useful IS that can faithfully track relevant sequences of state and

event changes that bear relevance to understanding the pandemic and its future trajectory.

Making this move will help the world, but it will also inform the future development of our

theories of the IS artefacts and their design. By refuting, accepting or modifying the theorized

criteria for faithful state-tracking that are at the core of tracing ability we can learn much

about the success or failure of state-tracking. Both outcomes will be an advance.

The primary empirical implication of the ideas in this paper is that analyses similar to

the inspection of Germany’s RKI COVID-19-Dashboard could now be carried out to compare

the wide range of publicly available reports, maps, and other dashboards. Many such IS have

been built over recent months and not all of them are equally useful (Datta, 2020). Empirical,

comparative research using the design principles or language in this paper as evaluation

criteria will help policy makers to utilize the best available IS infrastructure in their decision-

making, and it will allow the general public to better understand advantages and limitations of

the wide variety of corona dashboards available.

Limitations and Future Research

The obvious limitation of the research in this paper is the focus on one representative

case of an artefact, Germany’s RKI COVID-19-Dashboard. However, as noted, it builds on

the same set of technology infrastructure as most other corona dashboards, so the insights

should have some generalizability to other dashboard implementations. However, some of the

insights about the underlying information infrastructure or federate regulation may be limited

to countries with a similar federated democratic system. Analysing other cases of dashboards,

their technological, regulatory, and informational infrastructures in comparison to Germany’s

RKI COVID-19-Dashboard will certainly deepen and broaden the insights.

Likewise, the empirical insights reported in this paper are also grounded in the specific

context of the study. Data collection took place in North Rhine-Westphalia; other states and

regions have different legislative arrangements. Moreover, end users did not feature

prominently in the case study. Since corona dashboards are ultimately designed to be an

effective information system for the general public, future research should examine users’

perceptions of their usefulness, for example, through a cross-sectional survey.

Moreover, the focus in this paper was on representational, not informational issues.

Solving these issue may improve the usefulness of corona dashboards, but it will not solve

infodemic issues such as misinformation, amateur data analysis, or fake news (Laato et al.,

2020). It also will not solve issues of data quality, for example through wilfully restricted or

manipulated reporting of data into the systems (Cornish et al., 2020; Pundir, 2020). While my

focus was on representation of “facts”, it is also important to realize that often these facts are

socially constructed corona dashboard rely on data provided by some sort of (private, public,

national, or governmental) infrastructure.

Finally, the design propositions developed in this paper focus on representational

matters, that is, ways to convey information, alone; they do not speak to other purposes or

effects a corona dashboard may have. For example, because they display socially constructed

information as facts, corona dashboards may carry an action-oriented role (Ågerfalk &

Eriksson, 2004; Rittgen, 2006) in that they might instigate certain user responses or
behaviours. I have not looked at how action-orientation could be coupled with my chosen

focus on representation and state-tracking. I do believe, however, that such an analysis would

form a meaningful and desired complement. Such work could, for example, use rule-based

approaches to identify how language could be used in specific user contexts to facilitate

communication and action (Hirschheim et al., 1995, pp. 198-209).

CONLUSION

The fact that both policy makers and regular citizens extensively on corona dashboards

in dealing with the Covid-19 pandemic demonstrates that IS play an important role in solving

this crisis and preventing the next. We do not save lives directly, nor do we develop vaccines

or put more beds in hospital wards. The job of providing the right information to the right

people at the right time in the right format is nonetheless critical. Leveraging our cumulative

knowledge about the design of effective representation and state-tracking systems can be a

promising start. It will not stop the pandemic but it might help managing it in the most

effective and human way possible.

REFERENCES

Abbott, A. (2016). Processual Sociology. University of Chicago Press.

Adam, D. (2020a). The Limits of R. Nature, 583, 346‐348.
Adam, D. (2020b). Modelling the Pandemic. Nature, 580, 316‐318.
Adam, M., Werner, D., Wendt, C., & Benlian, A. (2020). Containing COVID‐19 Through Physical
Distancing: the Impact of Real‐time Crowding Information. European Journal of Information
Systems, 29(5), 595‐607. https://doi.org/10.1080/0960085X.2020.1814681
Ågerfalk, P. J., & Eriksson, O. (2004). Action‐oriented Conceptual Modelling. European Journal of
Information Systems, 13(1), 80‐92.
Ågerfalk, P. J., & Karlsson, F. (2020). Artefactual and Empirical Contributions in Information Systems
Research. European Journal of Information Systems, 29(2), 109‐113.
Aggarwal, C. C., & Subbian, K. (2012). Event Detection in Social Streams SIAM International
Conference on Data Mining, Anaheim, California.
Baskerville, R., Baiyere, A., Gregor, S., Hevner, A. R., & Rossi, M. (2018). Design Science Research
Contributions: Finding a Balance between Artifact and Theory. Journal of the Association for
Information Systems, 19(5), 358‐376.
Batra, D., & Antony, S. R. (1994). Novice Errors in Conceptual Database Design. European Journal of
Information Systems, 3(1), 57‐69.
Burton‐Jones, A., & Grange, C. (2013). From Use to Effective Use: A Representation Theory
Perspective. Information Systems Research, 24(3), 632‐658.
Burton‐Jones, A., Recker, J., Indulska, M., Green, P., & Weber, R. (2017). Assessing Representation
Theory with a Framework for Pursuing Success and Failure. MIS Quarterly, 41(4), 1307‐1333.
Carillo, K., Cachat‐Rosset, G., Marsan, J., Saba, T., & Klarsfeld, A. (2021). Adjusting to Epidemic‐
induced Telework: Empirical Insights from Teleworkers in France. European Journal of
Information Systems, 30(1), 69‐88. https://doi.org/10.1080/0960085X.2020.1829512
Carthaus, A. (2020). Corona Confusion: How to Make Sense of the Numbers and Terminology.
Deutsche Welle. Retrieved April 22 from https://p.dw.com/p/3ZeJ7
Chandra Kruse, L., Seidel, S., & vom Brocke, J. (2019). Design Archaeology: Generating Design
Knowledge from Real‐World Artifact Design DESRIST 2019: Extending the Boundaries of
Design Science Theory and Practice, Worcester, Massachusetts.
Cohen, J. (2020). With Global Push for COVID‐19 Vaccines, China Aims to Win Friends and Cut Deals.
Science. https://doi.org/10.1126/science.abf8838
Cornish, L., Jerving, S., & Ravelo, J. L. (2020). Data Around COVID‐19 is a Mess and Here's Why That
Matters. Devex,. Retrieved June 24 from https://www.devex.com/news/data‐around‐covid‐
19‐is‐a‐mess‐and‐here‐s‐why‐that‐matters‐97077
Datta, A. (2020). Here are Some of the Best Maps Tracking Coronavirus Updates. Geospatial Media &
Communications. Retrieved May 12 from https://www.geospatialworld.net/blogs/here‐are‐
some‐of‐the‐best‐maps‐tracking‐coronavirus‐updates/
Diercke, M. (2017). Deutsches Elektronisches Melde‐ und Informationssystem für den
Infektionsschutz (DEMIS). Epidemiologisches Bulletin, 30, 291‐293.
https://doi.org/10.17886/EpiBull‐2017‐039
Dong, E., Du, H., & Gardner, L. (2020). An Interactive Web‐based Dashboard to Track COVID‐19 in
Real Time. The Lancet, 20(5), 533‐534.
Dubé, L., & Paré, G. (2003). Rigor in Information Systems Positivist Case Research: Current Practices,
Trends, and Recommendations. MIS Quarterly, 27(4), 597‐635.
Eriksson, O., & Ågerfalk, P. J. (2021). Speaking Things into Existence: Ontological Foundations of
Identity Representation and Management. Information Systems Journal, 31.
https://doi.org/10.1111/isj.12330
Esri. (2020). ArcGIS Online. Esri. Retrieved November 26 from https://www.esri.com/en‐
us/arcgis/about‐arcgis/overview
European Centre for Disease Prevention and Control. (2020, 24 September 2020). Timeline of ECDC's
Response to COVID‐19. Retrieved November 26 from https://www.ecdc.europa.eu/en/covid‐
19/timeline‐ecdc‐response
Flaxman, S., Mishra, S., Gandy, A., Unwin, H. J. T., Mellan, T. A., Coupland, H., Whittaker, C., Zhu, H.,
Berah, T., Eaton, J. W., Monod, M., Perez‐Guzman, P. N., Schmit, N., Cilloni, L., Ainslie, K. E.
C., Baguelin, M., Boonyasiri, A., Boyd, O., Cattarino, L., Cooper, L. V., Cucunubá, Z., Cuomo‐
Dannenburg, G., Dighe, A., Djaafara, B., Dorigatti, I., van Elsland, S. L., FitzJohn, R. G.,
Gaythorpe, K. A. M., Geidelberg, L., Grassly, N. C., Green, W. D., Hallett, T., Hamlet, A.,
Hinsley, W., Jeffrey, B., Knock, E., Laydon, D. J., Nedjati‐Gilani, G., Nouvellet, P., Parag, K. V.,
Siveroni, I., Thompson, H. A., Verity, R., Volz, E., Walters, C. E., Wang, H., Wang, Y., Watson,
O. J., Winskill, P., Xi, X., Walker, P. G. T., Ghani, A. C., Donnelly, C. A., Riley, S., Vollmer, M. A.
C., Ferguson, N. M., Okell, L. C., Bhatt, S., & Imperial College, C.‐R. T. (2020). Estimating the
Effects of Non‐Pharmaceutical Interventions on COVID‐19 in Europe. Nature, 584(7820), 257‐
261. https://doi.org/10.1038/s41586‐020‐2405‐7
Gioia, D. A., Corley, K. G., & Hamilton, A. L. (2013). Seeking Qualitative Rigor in Inductive Research.
Organizational Research Methods, 16(1), 15‐31.
Gregor, S., Chandra Kruse, L., & Seidel, S. (2020). The Anatomy of a Design Principle. Journal of the
Association for Information Systems, 21(6), 1622‐1652. https://doi.org/10.17705/1jais.00649
Haldane, A., & Chowla, S. (2020). Fast Economic Indicators. Nature Reviews Physics.
https://doi.org/10.1038/s42254‐020‐0236‐y
He, X., Lau, E. H. Y., Wu, P., Deng, X., Wang, J., Hao, X., Lau, Y. C., Wong, J. Y., Guan, Y., Tan, X., Mo, X.,
Chen, Y., Liao, B., Chen, W., Hu, F., Zhang, Q., Zhong, M., Wu, Y., Zhao, L., Zhang, F., Cowling,
B. J., Li, F., & Leung, G. M. (2020). Temporal Dynamics in Viral Shedding and Transmissibility
of COVID‐19. Nature Medicine. https://doi.org/https://doi.org/10.1038/s41591‐020‐0869‐5
Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design Science in Information Systems
Research. MIS Quarterly, 28(1), 75‐105.
Hirschheim, R., Klein, H. K., & Lyytinen, K. (1995). Information Systems Development and Data
Modeling: Conceptual and Philosophical Foundations. Cambridge University Press.
Iivari, J., Hansen, M. R. P., & Haj‐Bolouri, A. (2020). A Proposal for Minimum Reusability Evaluation of
Design Principles. European Journal of Information Systems, forthcoming.
https://doi.org/10.1080/0960085X.2020.1793697
Kent, W. (1978). Data and Reality. North‐Holland.
Laato, S., Najmul Islam, A. K. M., Nazrul Islam, M., & Whelan, E. (2020). What Drives Unverified
Information Sharing and Cyberchondria During the COVID‐19 Pandemic? European Journal of
Information Systems, 29(3), 288‐305. https://doi.org/10.1080/0960085X.2020.1770632
Langefors, B. (1973). Theoretical Analysis of Information Systems. Studentlitteratur.
Lazzerini, M., & Putoto, G. (2020). COVID‐19 in Italy: Momentous Decisions and Many Uncertainties.
The Lancet, 8(5), E641‐E642.
Lee, A. S. (1991). Integrating Positivist and Interpretive Approaches to Organizational Research.
Organization Science, 2(4), 342‐365.
Limb, M. (2021). Covid‐19: Data on Vaccination Rollout and its Effects are Vital to Gauge Progress,
Say Scientists. British Medical Journal, 372(76). https://doi.org/10.1136/bmj.n76
Looi, M.‐K. (2020). Covid‐19: Is a Second Wave Hitting Europe? British Medical Journal, 371, m4113.
https://doi.org/10.1136/bmj.m4113
March, S. T., & Allen, G. N. (2014). Toward a Social Ontology for Conceptual Modeling.
Communications of the Association for Information Systems, 34(70), 1347‐1358.
Miles, M. B., & Huberman, M. (1994). Qualitative Data Analysis (2nd ed.). Sage.
Minji, L. (2020). S. Koreans Rely on Digital Maps to Track Spreading Virus. Yonhap News Agency.
Retrieved April 22 from https://en.yna.co.kr/view/AEN20200224003800315
Naidoo, R. (2020). A Multi‐level Influence Model of COVID‐19 Themed Cybercrime. European Journal
of Information Systems, 29(3), 306‐321. https://doi.org/10.1080/0960085X.2020.1771222
O'Connor, C., & Murphy, M. (2020). Going Viral: Doctors Must Tackle Fake News in the Covid‐19
Pandemic. British Medical Journal, 369, m1587. https://doi.org/10.1136/bmj.m1587
Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A Design Science Research
Methodology for Information Systems Research. Journal of Management Information
Systems, 24(3), 45‐77.
Pietz, J., McCoy, S., & Wilck, J. H. (2020). Chasing John Snow: Data Analytics in the COVID‐19 Era.
European Journal of Information Systems, 29(4), 388‐404.
https://doi.org/10.1080/0960085X.2020.1793698
Poole, M. S., & van de Ven, A. H. (1989). Using Paradox to Build Management and Organization
Theories. Academy of Management Review, 14(4), 562‐578.
Pundir, P. (2020). 'Missing’ COVID Deaths Raise Questions About India’s Tracking of the Outbreak.
Vice US. Retrieved June 16 from https://www.vice.com/en_au/article/3azdv3/missing‐covid‐
deaths‐raise‐questions‐about‐indias‐tracking‐of‐the‐outbreak
Recker, J., Indulska, M., Green, P., Burton‐Jones, A., & Weber, R. (2019). Information Systems as
Representations: A Review of the Theory and Evidence. Journal of the Association for
Information Systems, 20(6), 735‐786.
Reeves, J. J., Hollandsworth, H. M., Torriani, F. J., Taplitz, R., Abeles, S., Tai‐Seale, M., Millen, M., Clay,
B. J., & Longhurst, C. A. (2020). Rapid Response to COVID‐19: Health Informatics Support for
Outbreak Management in an Academic Health System. Journal of the American Medical
Informatics Association, In Press. https://doi.org/10.1093/jamia/ocaa037
Rittgen, P. (2006). A Language‐Mapping Approach to Action‐oriented Development of Information
Systems. European Journal of Information Systems, 15(1), 70‐81.
Robert Koch Institut. (2020, November 25, 2020). COVID‐19 (Coronavirus SARS‐CoV‐2) [in German].
Robert Koch Institut. Retrieved November 26, 2020 from
https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/nCoV.html
Rogers, K. (2020). Johns Hopkins' Dashboard: The People Behind the Pandemic's Most Visited Site.
CNN. Retrieved November 26, 2020 from https://edition.cnn.com/2020/07/11/health/johns‐
hopkins‐covid‐19‐map‐team‐wellness‐trnd/index.html
Rogers, Y. (1986). Pictorial Representations of Abstract Concepts Relating to Human‐Computer
Interaction. ACM SIGCHI Bulletin, 18(2), 43‐44.
Rosemann, M., Recker, J., Green, P., & Indulska, M. (2009). Using Ontology for the Representational
Analysis of Process Modeling Techniques. International Journal of Business Process
Integration and Management, 4(4), 251‐265.
Rubin, H. J., & Rubin, I. S. (2004). Qualitative Interviewing: The Art of Hearing Data (2nd ed.). Sage.
Sarker, S., Xiao, X., Beaulieu, T., & Lee, A. S. (2018). Learning from First‐Generation Qualitative
Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and
Evaluators (PART 1/2). Journal of the Association for Information Systems, 19(8), 752‐774.
Scott, L. M., & Janikas, M. V. (2010). Spatial Statistics in ArcGIS. In M. M. Fischer & A. Getis (Eds.),
Handbook of Applied Spatial Analysis: Software Tools, Methods and Applications (pp. 27‐41).
Springer.
Searle, J. R. (1995). The Construction of Social Reality. Free Press.
Silverman, D. (2013). Doing Qualitative Research (4th ed.). Sage.
SimilarWeb. (2020). Estimated total Visits to RKI.de. SimilarWeb. Retrieved November 26, 2020 from
https://www.similarweb.com/website/rki.de
Sorci, G., Faivre, B., & Morand, S. (2020). Explaining Among‐Country Variation in COVID‐19 Case
Fatality Rate. Nature Scientific Reports, 10(18909). https://doi.org/10.1038/s41598‐020‐
75848‐2
Stafford, N. (2020). Covid‐19: Why Germany’s Case Fatality Rate Seems so Low. British Medical
Journal, 369, m1395. https://doi.org/10.1136/bmj.m1395
Stamper, R. K. (1971). Some Ways of Measuring Information. Computer Bulletin, 15(12), 432‐436.
Trang, S., Trenz, M., Weiger, W. H., Tarafdar, M., & Cheung, C. M. K. (2020). One App to Trace Them
All? Examining App Specifications for Mass Acceptance of Contact‐tracing Apps. European
Journal of Information Systems, 29(4), 415‐428.
https://doi.org/10.1080/0960085X.2020.1784046
Tsoukas, H., & Chia, R. (2002). On Organizational Becoming: Rethinking Organizational Change.
Organization Science, 13(5), 567‐582.
Urbaczweski, A., & Lee, Y. J. (2020). Information Technology and the Pandemic: a Preliminary
Multinational Analysis of the Impact of Mobile Tracking Technology on the COVID‐19
Contagion Control. European Journal of Information Systems, 29(4), 405‐414.
https://doi.org/10.1080/0960085X.2020.1802358
Waizenegger, L., McKenna, B., Cai, W., & Bendz, T. (2020). An Affordance Perspective of Team
Collaboration and Enforced Working From Home During COVID‐19. European Journal of
Information Systems, 29(4), 429‐442. https://doi.org/10.1080/0960085X.2020.1800417
Wand, Y., & Weber, R. (1990). An Ontological Model of an Information System. IEEE Transactions on
Software Engineering, 16(11), 1282‐1292.
Wand, Y., & Weber, R. (1993). On the Ontological Expressiveness of Information Systems Analysis
and Design Grammars. Journal of Information Systems, 3(4), 217‐237.
Wand, Y., & Weber, R. (1995). On the Deep Structure of Information Systems. Information Systems
Journal, 5(3), 203‐223.
Weber, R. (1997). Ontological Foundations of Information Systems. Coopers & Lybrand and the
Accounting Association of Australia and New Zealand.
Wong, F., & Collins, J. J. (2020). Evidence that Coronavirus Superspreading is Fat‐Tailed. Proceedings
of the National Academy of Sciences, 117(47), 29416‐29418.
https://doi.org/10.1073/pnas.2018490117
World Health Organization. (2010). What is a Pandemic? World Health Organization. Retrieved
November 27, 2020 from
https://www.who.int/csr/disease/swineflu/frequently_asked_questions/pandemic/en/
Xie, L., Sundaram, H., & Campbell, M. (2008). Event Mining in Multimedia Streams. Proceedings of
the IEEE, 96(4), 623‐647.
Yin, R. K. (2009). Case Study Research: Design and Methods (4th ed., Vol. 5). Sage Publications.
 APPENDIX: CASE STUDY PROCEDURES

Design
I followed an embedded single-case design, following guidelines for positivist case

study research (Dubé & Paré, 2003; Miles & Huberman, 1994; Yin, 2009). Through the case

study, I tried to learn “facts” (Sarker et al., 2018) about how the design of Germany’s RKI

COVID-19 dashboard was completed and about whether my design propositions were viewed

as important, actable, and effective (Iivari et al., 2020). The objective of the case study was

thus to accurately confront and falsify the design principles with reality.

Data Collection Procedures

Ethical clearance was granted prior to data collection by the University of Cologne (ref

no 200020JR). Table A1 summarizes data sources. Primary data were interviews with

seventeen stakeholders from six different organizations involved with Germany’s RKI

COVID-19 dashboard: the main technology provider (Esri), a public health institute, a

regional public sector computer support and services provider (RegionalIT) plus three

regional public sector institutions from the Cologne region in Germany (a local crisis

management team, an office for public real estate cadastre, and a geo-data management team).

Interviewees were purposefully but not randomly recruited through contacts

established by Esri on basis of theoretical sampling criteria (in particular, roles such as

technical infrastructure developer, user interface designer, public health expert, cartography

expert, regional government, and local crisis response teams). Interviews were conducted in

German and recorded onsite or via videoconferencing between September and November

2020. Average interview length was 58 minutes. Recordings were transcribed, translated, and

verified by interviewees. Additional data collected included notes, memos, system

screenshots, documentations of additional information systems like SORMAS, an open source

early warning and management system (https://sormasorg.helmholtz-hzi.de/), different corona

dashboard variants for regional districts or local crisis management teams (Table A1), and
other relevant web pages, such as the NPGeo hub (https://npgeo-corona-npgeo-

de.hub.arcgis.com/), or the German Fire Protection Association (https://www.vfdb.de/).

Table A1. Data sources

Organization Esri Public RegionalIT Local crisis Office for District geo- Total
health management public real data
institute team estate management
cadastre team
Interviews 6 1 3 3 2 2 17
Key Executive, Geo-health GIS Department Department Geo-data
Informants business specialist developer, manager, manager, manager,
development application public geomatics crisis
manager, consultant relations specialist management
solution manager, team
architect, GIS coordinator
cartographer, coordination
user manager
interface
designer
Length 316 70 51 38 49 524
(minutes)
Length 93 20 17 14 17 161
(pages)
Notes 32 6 7 5 5 55
(pages)
Additional 4 1 1 1 3 1 11
documents
Relevant https://npgeo https://healt https://civite https://experi https://rhein- https://www.ar
local -corona- h- c.maps.arcgi ence.arcgis.c erft- cgis.com/apps
dashboards npgeo- mapping.de/ s.com/apps/ om/experienc kreis.maps.a/opsdashboar
and web de.hub.arcgi 7TageInzide opsdashboa e/0f5d0f9aa1 rcgis.com/apd/index.html#/
portals s.com/); nz/ rd/index.html 1a4bababe77 ps/opsdashb252af02201ee
https://npgeo #/0453cba0 5b5c14b735f oard/index.h 4a70bf4190b3
-corona- 2245458e86 tml#/6bd759 39731eee
npgeo- 9241f4070a ccf5694481
de.hub.arcgi 4393; 82e9551134
s.com/app/3 https://www. 183300
a132983ad3 giscloud.nrw
c4ab8a2870 .de/corona-
4e9addefaba dashboard.h
tml

Data Collection Protocol

Interviews followed a semi-structured interview protocol, which evolved in four

iterations over the course of the study. It is registered online at 10.17605/OSF.IO/Q3WGV.

The protocol was designed with a set of pre-planned questions to cover the subject area

(Rubin & Rubin, 2004). During the interviews, follow-up inquiries were used in addition to

the protocol to gain a deeper understanding of the subject matter or to clarify individual
responses. The protocol consisted of four parts: First, clarifying the context and role of the

interviewee; second, understanding the design of the corona dashboard, key features, key use

cases, and key changes since March 2020; third, evaluating the design principles for state-

tracking; and fourth, gathering additional insights and comments.

To evaluate the design principles for state-tracking, I relied on the recommendations

and template by Iivari et al. (2020). They suggest five criteria, accessibility, importance,

novelty, actability, and effectiveness, to examine whether the prescriptive knowledge for

creating instances of IS artefacts contained in the design principles is in fact applicable and

helpful for practitioners. Of these, accessibility of the design principles (the degree to which

practitioners can understand the design principle) was ensured in the interviews by providing

clear definitions, explanations, and several examples relevant to corona dashboards (see the

interview protocol, part three), rendering the evaluation of accessibility irrelevant. Likewise,

novelty (the degree to which a design principles bears potential to surprise practitioners) was

deemed irrelevant to the objectives of the case study. In turn, the interview covered

evaluations of importance, actability, and effectiveness.

Data Analysis
A two-pronged strategy was followed. The primary data analysis strategy was

hypothetico-deductive (Lee, 1991; Sarker et al., 2018): Data-grounded, inter-subjectively

valid, and verifiable claims (Silverman, 2013) were generated to validate or falsify the four

design principles for state-tracking. Analysis thus involved primarily pattern matching (Yin,

2009), to compare empirically generated evidence (primarily quotes made in the discussion of

each criteria for each design principle) with a logical evaluation outcome pattern (a broad

categorization in low, medium, and high) for each design principle. Table A2 provides several

illustrative examples.
 Table A2. Examples for deductive analysis of data

Design Criteria Example quote Matched

principle evaluation
pattern
Mapping Actability Well, the problem is that the state variables that would Low: data
be most interesting do not exist in the same processing
chronological order as the reported data. There are complicated
many different possibilities that could be considered in
parallel to the reporting system. But the statistical
processing is tedious.
Effectiveness We also thought about determining the R-value. I then Medium: possibly
took a look at the RKI, how is it actually derived. There effective but with
are so many ways to calculate it inside, so many associated risks
different results. I would keep my hands off it. We did
then too. We're careful with that. We don't want to give
the wrong signal to the outside world.
Tracking Actability This recovery data is not recorded directly. That is also Low: lack of data
very difficult. That's why the whole process is missing in availability
the health care system, where only the illness is
recorded, and estimated values are available. But of
course it would also be a great help if you had the data
with a better time stamp for a crisis management team.
External Importance I think it also makes much more sense to show the High: logical
event correlation between the measures we have taken and arguments that
the effects. So I think this makes a lot of sense in support
principle. implementation
Actability You would have to find the right place. I think that these Medium:
graphs on the bottom right are ideal for this purpose. I implementation
would first have to investigate a little more closely possible but effort
whether we could display such individual events and required
what a suitable representation would look like.
Superimposed events are, I believe, not technically
intended. You need to build something like this from
scratch.
Effectiveness If the result now comes out, for example, "despite the Medium: Risk of
use of a mask, the numbers are going up again", that misinterpretation
ultimately leads to people all saying "The mask is possible
nonsense. I'll leave it out". […] That would lead to
misinterpretations among citizens. In this respect, you
have to weigh more carefully what you are representing.
Sequencing Actability If the data is in a reasonable form that it can be Medium: data
processed, then it is relatively easy to put it on a card. preparation
When it starts with getting the data from the individual required
federal states or the individual districts, then how do you
bring it together? What is the data model behind it? Can
that be unified? In one rural district the limit is perhaps
50, in the other at 100. That's where the problems start, I
think, but if you have a clean database, the
representation or visualization on the map is still the nice
accessory , but the main work is in data preparation.
Sequencing Importance I think it is absolutely necessary to go in this direction, High: Very
that one somehow depicts the complexity with all its relevant solution
interactions. You can do that. It's not a big problem with limited
mathematically either. complexity
 The secondary data analysis strategy was inductive (Miles & Huberman, 1994), with

the aim to develop deeper insights into the rationale of the evaluation of the design principles,

that is, why a particular design principles was not deemed actable, and how a design principle

was implemented if at all in the in situ artefact-in-use (Chandra Kruse et al., 2019). A

thematic analysis was carried out using Gioia et al.’s (2013) methodology to build a data

structure that grouped 1st order concepts adhering to interviewees’ terms into 2nd order

conceptual themes, which I captured as tensions (structural paradoxes that require

reconciliation, Poole & van de Ven, 1989) to express the similarities and differences in the 1st

order concepts. Table A3 provides illustrative examples.

Table A3. Examples for inductive analysis of data

Example quote 1st order 2nd order theme

concept
So this is of course also a huge challenge to collect these data Federalism in Administrative
and federalism is not exactly playing into our hands, if you like data reporting tension: federated
to put it that way, because every region has its own publication legislation versus
mechanisms and platforms to disclose such things. And in centralized
some cases, you can no longer find them at all. information
infrastructure
I know that in [a different district], for example, the dashboard Information Administrative
is operated by the data processing centre and not by the ownership and tension: federated
district administration. And the public health department is solution legislation versus
included in the district administration. We don't have a data sovereignty centralized
processing centre, we only have an IT department, and we information
ourselves are more or less independent, and we are also infrastructure
members of the crisis team. This means that we ourselves can
react quickly but do not have to access a data processing
centre. […] There is always the question: who maintains the
data there?
If you compare that with weather forecasts, where you have Learning about Temporal tension:
collected something over the years, you have models, you forecasting fromAccurate historical
have that. You can also use a dashboard and visualize it. If other scenarios data versus uncertain
you apply it to this case, you simply have to gain a little forecast data
experience. […] And then you can perhaps get more valid
forecasts. […] The most important thing is official and really
verified data, because there is also enough fake news and
whatever else is thrown around.
Of course, we also have this forecast data, but it has been Availability and Trust tension:
shown, or was agreed, that such data does not belong in a vulnerability of Confidence in data
public dashboard. Only the pure facts should be included, in forecast data processing versus
order not to make oneself vulnerable to attack. data interpretation
And so that means such uncertainty - we've discussed this, Provision Trust tension:
can one represent the uncertainty of the numbers? It can be versus Confidence in data
represented, but we know that the public in particular cannot interpretation of processing versus
deal with it. uncertain data data interpretation

View publication stats