Applying Uncommon Visualizations to Government Dashboards

Puripant Ruchikachorn

1,2 a

Institute of Public Policy and Development (IPPD), Bangkok, Thailand

Chulalongkorn Business School, Chulalongkorn University, Bangkok, Thailand

Keywords:

Visualization Design, Government, Dashboard.

Abstract:

Many governments provide data dashboards to present the state of the countries or administrative activities.

Their main target audience is typically the citizens but the dashboard design process is usually top-down and

leads to formulaic results. Developing three data dashboard projects for the government of Thailand, we

successfully applied two uncommon data visualizations, grid map and connected scatterplot, despite initial

resistance from the government agencies. We documented the design process including feedback on the two

visualizations and solutions to alleviate their concerns. Academic studies had little success in convincing

stakeholders. In both visualizations, animations helped to frame the concept of the uncommon visualizations.

1 INTRODUCTION

Although the government of Thailand is not a leader

in open data (Lemieux et al., 2015), there has been

an attempt to gather data into a centralized web-

site: https://data.go.th. However, having clean data

in machine-readable formats alone does not yield any

insights. Open data needs a way to visualize, even

as a simple heatmap for its metadata (Carvalho et al.,

2015). Common solutions include data dashboards

that present the economic or administrative statistics

relative to other countries, smaller administrative lev-

els, or historical periods.

A data dashboard is a visual display that con-

solidates various important information on a single

screen so its user can scan the dashboard and per-

ceive the data overview (Few, 2006). With real-time

data, it can be used for operational purpose to observe

outliers and take immediate actions (Sarikaya et al.,

2019). In a dashboard, there are usually many vi-

sualization types; each one for a particular function.

For example, a dashboard may have both a line chart

and a pie chart to show the trend of gross domestic

product (GDP) over the last decade and the compo-

sition of the latest GDP, respectively. A government

dashboard can increase data interpretability (Barcel-

los et al., 2017), empower the citizens through inter-

acting with data (Bowyer et al., 2019), and even pre-

vent corruption (Petasis et al., 2018).

https://orcid.org/0000-0002-2721-6915

In the past three years, we have involved with

three government dashboards for different purposes.

The main tasks were broad and the data varied so

we were able to implement a diverse set of visual-

ization types including a grid map (or a tile grid map)

and a connected scatterplot in the dashboards. To our

knowledge, we had not seen any prior use of such

visualizations in Thailand. Because of their obscu-

rity, there were some reluctance from the government

agencies, not unlike many digital transformation ini-

tiatives (Schirrmacher et al., 2019). We gathered their

feedback and developed the visualizations further to

address their concerns. We found that, instead of prior

studies, animated visualizations could persuade un-

convinced stakeholders.

2 CASE STUDIES

All dashboard projects were unrelated and had differ-

ent data sources and tasks. However, all shared the

characteristic of multiple stakeholders of which one

was in charge of project ﬁnance and also test users in

the ﬁrst few design iterations. Another group of im-

portant stakeholders included end users without any

speciﬁc details and could cover as large as all the cit-

izens. They were usually involved only in the end of

the process.

Dashboard for Spending Data of the Royal Thai

Government. The ﬁrst dashboard was commissioned

Ruchikachorn, P.

Applying Uncommon Visualizations to Government Dashboards.

DOI: 10.5220/0010298102030209

In Proceedings of the 16th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2021) - Volume 3: IVAPP, pages

203-209

ISBN: 978-989-758-488-6

203

by the Comptroller General’s Department (CGD),

the Ministry of Finance, to increase the data trans-

parency of the ﬁnance of the Royal Thai Govern-

ment (Ruchikachorn et al., 2019), similar to an-

other project for the federal ﬁnance of the United

States (B

ohm et al., 2012). The budgetary data is ge-

ographical and updated on a weekly basis. The main

task is to inspect budgetary distribution and locate ar-

eas of particular statistics such as provinces receiving

large budget allocations.

Thai People Map and Analytics Platform

(TPMAP). Developed for National Electronics

and Computer Technology Center (NECTEC),

TPMAP is a dashboard designed to target geo-

graphical areas for poverty alleviation programs

in Thailand (Surasvadi et al., 2019). Unlike the

government spending dashboard, the data is also

hierarchical. The user should be able to not only

spot poor population density but also drill down onto

lower administrative levels.

Development Index Dashboard. The Institute of

Public Policy and Development (IPPD) would like to

compare Thailand to other countries and present the

results to the general public. The underlying dash-

board data were public development data from the

websites of the World Economic Forum (WEF) and

United Nations Development Programme (UNDP)

for the Sustainable Development Goals (SDGs).

There were many time series and the analysts at the

institute wanted to show their pairwise relations in-

cluding correlation and trend.

3 UNCOMMON

VISUALIZATIONS AND THEIR

IMPLEMENTATIONS

There is no thorough worldwide survey on visual lit-

eracy, so it is hard to pinpoint uncommon visualiza-

tions. In this work, we deﬁne uncommon visualiza-

tions as the ones that are not readily available in pop-

ular commercial visualization applications: Tableau

Desktop, Microsoft Power BI, and Google Data Stu-

dio. Common visualizations are a treemap, a choro-

pleth map, a pie chart, a donut chart, a bar chart, a line

chart, an area chart, a scatterplot, a bubble chart, their

stacked and clustered variants—if possible, and their

combinations. Hence, we consider a grid map and

a connected scatterplot as uncommon visualizations.

Although certain customizations in some visualiza-

tion applications support other visualization types due

to their similar visual grammars, we do not count

them as common visualizations because those conﬁg-

urations are not expected to be frequently used or even

discovered.

Our method was comparable to the selection pro-

cess of unfamiliar visualizations (Lee et al., 2016),

which were manually selected and excluded the vi-

sualizations in K-12 curricula such as a pie chart, a

bar chart, a line chart, and a scatterplot. Note that a

treemap was considered as an unfamiliar visualization

but it was available in all commercial visualization ap-

plications that we surveyed. We intentionally differ-

entiate the terms because familiarity implies having

known, read, or created such visualization beforehand

(for example, through classroom materials) while an

uncommon visualization may be created via coding

or other means but is not taught or directly available

for laypeople in any commercial applications.

Federal regulations do not allow any online ser-

vices that require cloud storage servers outside of the

country so we ruled out many cloud solutions and ﬁ-

nally decided to self-host the dashboards. All visual-

izations support minimal interaction, namely, details-

on-demand. They were implemented with web stan-

dard technologies and a JavaScript library D3.js.

3.1 Grid Maps

A common method of displaying quantitative values

over regions is a choropleth map but it may show bias

in favor of large areas due to uneven administrative

region sizes. Some media outlets use a grid map that

presents all regions as equal in size. It can also de-

emphasize the size of map regions that are more cou-

pled with geopolitical history than with data context.

Several related techniques can be applied to create

a grid map. Recently, there is a research project that

directly addresses grid map generation (McNeill and

Hale, 2017). The proposed technique can have grids

or tiles of varied geometries including triangular, rect-

angular, and hexagonal grids with certain constraints.

For the dashboards, we designed and implemented

a hexagonal grid map of Thailand. Among all three

regular tilings i.e. tessellations by convex regular

polygons, a hexagonal grid provides the highest edge-

to-edge connectivity of six compared to three and four

for triangular and rectangular grids, respectively. The

hexagonal grid is therefore more suitable for a dense

map whose regions are well-connected. The central

area of Thailand has that characteristic so we picked

the hexagonal grid for our grid map of Thailand. The

position of each province was manually placed to

closely match the mental image of the geographical

map of Thailand. Since there were only 77 provinces,

IVAPP 2021 - 12th International Conference on Information Visualization Theory and Applications

204

Figure 1: The grid maps of the government spending dash-

board with the same diverging color scale (low in red and

high in blue) but different measures. As an outlier, Bangkok

is shown in green. Both show provincially aggregated bud-

get but the map to the right performs per capita transforma-

tion ﬁrst. Shown are tooltips of the same province, Trat.

it was not necessary to use an algorithm for automatic

layout. A grid map example is shown in Figure 1.

3.2 Connected Scatterplots

One polyline in a line chart can represent only one

time series while a polyline in a connected scatter-

plot can represent two time series (Haroz et al., 2016).

A correlation between two time series can be visu-

ally observed through the direction of the polyline.

A polyline between the top-left and the bottom-right

corners suggests a negative correlation while the other

diagonal direction implies that the two time series are

positively correlated.

We implemented and used this visualization in a

dashboard that compares Thailand’s development in-

dices to other countries’ in the region. With a lot of in-

dices, putting two indices on a chart at the same time

can provide a good overview of development progres-

sion across countries and time frames. The connected

scatterplot in Figure 2 tells a story that Thai people

work long hours, but the productivity is low. All cir-

cles show their tooltips when they are triggered by a

mouseover event.

4 PRELIMINARY FEEDBACK

AND ITERATIVE DESIGN

As our main concern was not a formal validation of

the visualization techniques that had been previously

evaluated, we did not try to acquire a large group

of users to test a statistical hypothesis (Nielsen and

Landauer, 1993). We adopted an iterative design pro-

cess and tried to apply appropriate validations for all

dashboards with the same groups of stakeholders who

had provided the requirements (Munzner, 2009). We

did not aim to use uncommon visualizations in the

projects, but they were appropriate for the tasks and,

to the development teams, the visualizations seemed

not really unusual. During the ﬁrst design iterations,

we had a feeling that they could feel slightly uncon-

ventional but did not expect unenthusiastic feedback

from some stakeholders.

The uncommon visualizations, a grid map and

a connected scatterplot, were introduced through

sketches and later digital mock-ups of higher ﬁdelity.

All early prototypes were not interactive and there

were quite a few questions regarding the usability of

the visualizations. We experienced some common pit-

falls in visualization design (Sedlmair et al., 2012)

that most government ofﬁcials had speciﬁc visualiza-

tion types and dashboard designs in mind. Expanding

their horizon on visual possibility helped; in our case,

we discussed dashboard taxonomy and various types

beyond the commonly known (Sarikaya et al., 2019).

There were more questions and doubts about the

visualizations after the ﬁrst interactive prototype had

been shown outside of the immediate working team.

The connected scatterplot in the development index

dashboard received a lot of questions second-guessing

whether the general public could read the visual en-

coding or interpret the data. More speciﬁcally, there

were concerns that people might not understand that

both vertical and horizontal axes showed development

indices and the temporal progression did not have to

be left-to-right. Besides a short text to serve as an

instruction, we added an animated visual cue in the

beginning to suggest the direction of time; all points

and lines showed up with cascaded delays. Also, the

dashboard started with one development index on the

vertical axis and the horizontal axis represented year.

The users could add an index to the horizontal axis

and the data lines would morph to the new positions,

serving as a tutorial on the unknown visualization.

We tried to present a previous study as a proof

of usability. For instance, a study on connected scat-

terplots (Haroz et al., 2016) observed many cases of

this visualization usage even in mass media and con-

ducted an experiment to conﬁrm its usability. The re-

Applying Uncommon Visualizations to Government Dashboards

205

Figure 2: The connected scatterplot of work hours per person per year (x-axis, reversed as lower work hours imply a better

quality of life) and productivity measured in GDP per work hour (y-axis). Blue, red, green, yellow, and pink lines represent

Thailand, South Korea, Malaysia, Vietnam, and Indonesia, respectively, while gray lines represent continent-level aggregates

i.e. Africa, Americas, Asia & Oceania, and Europe.

searchers found that the chart might take some time

to recognize but the participants could understand the

visual representation with little explanation and even

engage with the chart more, compared to a line chart.

The early users simply disregarded the study, doubted

the study setup, or questioned whether culture or vi-

sual literacy played an important role in understand-

ing the encoding of a connected scatterplot. This

shows that an academic study, at least from an ex-

ternal source, has little effect to convince stakehold-

ers. This is also in stark contrast to a more objec-

tive assessment of unknown visualizations of science

community (Dasgupta et al., 2017a; Dasgupta et al.,

2017b).

A grid map did not raise any serious issues. A

common comment was that it was hard to see the

provinces with the most or least poor population as

the users had to scan throughout the map. We created

another view to sort all hexagons by value. To link

both views together, we introduced an animation to

morph between the two as depicted in Figure 3.

Other qualitative comments such as aesthetics

were taken into consideration during the development

process as well. The users might request contradic-

tory adjustments and were unaware of their trade-

off. For example, one asked for all straight lines in a

connected scatterplot for accurate portrayal while an-

other preferred more curved and smooth lines without

acknowledging the fact that monotonicity might not

have been preserved.

5 DISCUSSION AND FUTURE

WORK

Our use of animations to show the temporal direc-

tion in a connected scatterplot and transitions between

visualization types and views was based on previous

studies on how people make sense of uncommon vi-

sualizations (Ruchikachorn and Mueller, 2015; Lee

et al., 2016). The animations of both grid map and

connected scatterplot also addressed a comment that

the visualizations are not attractive enough. The an-

imations excited some users and invited them to in-

teract. To the users, animated hexagons in a grid map

and lines in a connected scatterplot felt more concrete.

IVAPP 2021 - 12th International Conference on Information Visualization Theory and Applications

206

Figure 3: The source, an in-between, and the target of the transition from a hexagonal grid map of Thailand in TPMAP to a

hexagon list sorted by the data values. Each hexagon’s position is linearly interpolated between the source and target positions

with cascaded delays.

They were helpful to construct a frame to familiarize

themselves with what they do not know (Lee et al.,

2016).

Between two uncommon visualizations, adopting

a grid map seemed to have less resistance. Although

it cannot be directly compared to connected scatter-

plot in terms of commonness or complexity due to

subjectivity and difﬁculty of such measurement, the

relatively easy acceptance of an uncommon map was

quite surprising. This could be due to its more obvi-

ous beneﬁt compared to a choropleth map and that the

hexagons forming as the shape of Thailand still look

familiar and less abstract than a connected scatterplot.

To further analyze the reluctance to initially accept

the uncommon visualizations, we can refer to other

related psychological and sociological topics such as

mental models, previously explored in the context of

security dashboards (Maier et al., 2017), and face

which is a widely referenced sociological concept on

how a person behaves within social relations (Zhang

et al., 2006). In the case of uncommon visualizations,

face can imply the tendency to maintain the status

quo or existing systems. Psychologically, this may

be the clash between System 1 and System 2 thinking

i.e. people tend to ﬁt a new thing within the context

of the already known and fail to see the uniqueness

unless they are given more time and conditions for

slow thinking (Kahneman, 2013). To compete with

known visualizations, particularly in an existing sys-

tem, an uncommon visualization needs not only show

clear advantages but also overcome cognitive biases

and fallacies such as endowment effect and sunk cost

fallacy (Dobelli, 2014).

Another complication that may happen but has

never been explored in the context of visualization is

preference falsiﬁcation. A person may publicly ex-

press their opinion opposite of their real preference

due to social pressure (Kuran, 1987; Kuran, 1997).

This may lead slow and then sudden change as evi-

denced by social movements (Sunstein, 2019). With

the same model, unfamiliar and uncommon visualiza-

tions may be more popular than anticipated and gain

rapid public acceptance as well.

The concepts of preference falsiﬁcation and di-

verse thresholds in behavioral science (Sunstein,

2019) can be applied and studied in the context of

human-computer interaction and visualization. A

common visualization task categorization often im-

plies one user tier of similar experience and expertise.

In reality, people have diverse thresholds to under-

stand or interact with a certain interface. Even within

a single tier, the users may request a more conven-

tional user interface and interaction because of pref-

erence falsiﬁcation.

For future work, we may study and follow the

ISO on human-centered design (International Orga-

nization for Standardization, 2019) to reassure stake-

holders. Also, we would like to formally evaluate

Applying Uncommon Visualizations to Government Dashboards

207

the readability of the grid map and connected scatter-

plot, speciﬁcally in the context of Thailand, to verify

whether Thai audience can understand the uncommon

visualizations without a training, tutorial, or other vi-

sual cues. This can be coupled with visual literacy

assessment (Boy et al., 2014). When all the dash-

boards are released to the public, a larger study to

validate the dashboard and the uncommon visualiza-

tions can be conducted and we hope this will lead to

more adoption of various visualization types beyond

the common ones.

ACKNOWLEDGEMENTS

This work was supported in part by Chulalongkorn

University, CGD, NECTEC, and IPPD. The govern-

ment spending dashboard and TPMAP were devel-

oped by Boonmee Lab.

REFERENCES

Barcellos, R., Viterbo, J., Miranda, L., Bernardini, F., Ma-

ciel, C., and Trevisan, D. (2017). Transparency in

practice: using visualization to enhance the inter-

pretability of open data. In Proceedings of the 18th

Annual International Conference on Digital Govern-

ment Research, pages 139–148, New York, NY, USA.

ACM Press.

ohm, C., Schmidt, M., Freitag, M., Heise, A., Lehmann,

C., Mascher, A., Naumann, F., Ercegovac, V., Hernan-

dez, M., and Haase, P. (2012). GovWILD: Integrating

Open Government Data for Transparency. Proceed-

ings of the 21st international conference companion

on World Wide Web, pages 321–324.

Bowyer, A., Wilson, R., Wheater, S., Snape, M., and Mon-

tague, K. (2019). Human-Data Interaction in the Con-

text of Care: Co-designing Family Civic Data Inter-

faces and Practices. In Extended Abstracts of the 2019

CHI Conference on Human Factors in Computing Sys-

tems - CHI EA ’19, pages 1–6, New York, NY, USA.

ACM Press.

Boy, J., Rensink, R. A., Bertini, E., and Fekete, J.-D.

(2014). A Principled Way of Assessing Visualiza-

tion Literacy. IEEE Transactions on Visualization and

Computer Graphics, 20(12):1963–1972.

Carvalho, P., Hitzelberger, P., Otjacques, B., Bouali, F., and

Venturini, G. (2015). Information Visualization for

CSV Open Data Files Structure Analysis. In Pro-

ceedings of the 6th International Conference on Infor-

mation Visualization Theory and Applications, pages

101–108. SCITEPRESS.

Dasgupta, A., Burrows, S., Han, K., and Rasch, P. J.

(2017a). Empirical Analysis of the Subjective Impres-

sions and Objective Measures of Domain Scientists’

Visual Analytic Judgments. In Proceedings of the

2017 CHI Conference on Human Factors in Comput-

ing Systems, pages 1193–1204, New York, NY, USA.

ACM.

Dasgupta, A., Lee, J.-Y., Wilson, R., Lafrance, R. A.,

Cramer, N., Cook, K., and Payne, S. (2017b). Fa-

miliarity Vs Trust: A Comparative Study of Domain

Scientists’ Trust in Visual Analytics and Conventional

Analysis Methods. IEEE Transactions on Visualiza-

tion and Computer Graphics, 23(1):271–280.

Dobelli, R. (2014). The art of thinking clearly. Harper

Paperbacks.

Few, S. (2006). Information dashboard design. O’Reilly

Media.

Haroz, S., Kosara, R., and Franconeri, S. L. (2016). The

Connected Scatterplot for Presenting Paired Time Se-

ries. IEEE Transactions on Visualization and Com-

puter Graphics, 22(9):2174–2186.

International Organization for Standardization (2019). Er-

gonomics of Human-system Interaction—Part 210:

Human-centred Design for Interactive Systems. ISO,

2 edition.

Kahneman, D. (2013). Thinking, fast and slow. Farrar,

Straus and Giroux.

Kuran, T. (1987). Preference Falsiﬁcation, Policy Continu-

ity and Collective Conservatism. The Economic Jour-

nal.

Kuran, T. (1997). Private truths, public lies: The social

consequences of preference falsiﬁcation.

Lee, S., Kim, S.-H., Hung, Y.-H., Lam, H., Kang, Y.-A.,

and Yi, J. S. (2016). How do People Make Sense

of Unfamiliar Visualizations?: A Grounded Model

of Novice’s Information Visualization Sensemaking.

IEEE Transactions on Visualization and Computer

Graphics, 22(1):499–508.

Lemieux, V. L., Trapnell, S. E., and Excell, C. (2015).

Transparency and Open Government: Reporting on

the Disclosure of Information. JeDEM, 7(2):75–93.

Maier, J., Padmos, A., S. Bargh, M., and W

orndl, W.

(2017). Inﬂuence of Mental Models on the Design

of Cyber Security Dashboards. In Proceedings of the

12th International Joint Conference on Computer Vi-

sion, Imaging and Computer Graphics Theory and

Applications, pages 128–139. SCITEPRESS.

McNeill, G. and Hale, S. A. (2017). Generating Tile Maps.

Computer Graphics Forum, 36(3):435–445.

Munzner, T. (2009). A Nested Model for Visualization De-

sign and Validation. IEEE Transactions on Visualiza-

tion and Computer Graphics, 15(6):921–928.

Nielsen, J. and Landauer, T. K. (1993). A mathematical

model of the ﬁnding of usability problems. In Pro-

ceedings of the SIGCHI conference on Human factors

in computing systems - CHI ’93, pages 206–213, New

York, NY, USA. ACM Press.

Petasis, G., Triantaﬁllou, A., and Karstens, E. (2018). Your-

DataStories: Transparency and Corruption Fighting

Through Data Interlinking and Visual Exploration.

pages 95–108. Springer, Cham.

Ruchikachorn, P. and Mueller, K. (2015). Learning Vi-

sualizations by Analogy: Promoting Visual Liter-

acy Through Visualization Morphing. IEEE Trans-

IVAPP 2021 - 12th International Conference on Information Visualization Theory and Applications

208

actions on Visualization and Computer Graphics,

21(9):1028–1044.

Ruchikachorn, P., Suveeranont, R., and Luangaroonlerd,

T. (2019). A Case Study of Open Data Visualiza-

tion System for Government Transparency in Thai-

land. In 2019 IEEE Paciﬁc Visualization Symposium

(PaciﬁcVis), pages 353–354. IEEE.

Sarikaya, A., Correll, M., Bartram, L., Tory, M., and Fisher,

D. (2019). What Do We Talk About When We Talk

About Dashboards? IEEE Transactions on Visualiza-

tion and Computer Graphics, 25(1):682–692.

Schirrmacher, N.-B., Ondrus, J., Tan, F., Loh, Y., and

Hardoon, D. R. (2019). Overcoming Status Quo Bias:

Nudging in a Government-Led Digital Transforma-

tion Initiative. In ICIS 2019 Proceedings.

Sedlmair, M., Meyer, M., and Munzner, T. (2012). Design

Study Methodology: Reﬂections from the Trenches

and the Stacks. IEEE Transactions on Visualization

and Computer Graphics, 18(12):2431–2440.

Sunstein, C. R. (2019). How Change Happens.

Surasvadi, N., Ruchikachorn, P., Siripanpornchana, C.,

Thajchayapong, S., and Plangprasopchok, A. (2019).

TPMAP: A Data Analytics and Visualization Platform

to Support Thailand Target Poverty Alleviation Pro-

grams. In 2019 IEEE Paciﬁc Visualization Symposium

(PaciﬁcVis), pages 331–332. IEEE.

Zhang, Q., Chintakovid, T., Sun, X., Ge, Y., and Zhang,

K. (2006). Saving Face or Sharing Personal Informa-

tion? A Cross-Cultural Study on Knowledge Sharing.

Journal of Information & Knowledge Management,

05(01):73–79.

Applying Uncommon Visualizations to Government Dashboards

209