EDITORS: Beatriz Sousa Santos, bss@ua.

pt
Ginger Alford, [email protected]

DEPARTMENT: EDUCATION

Educational Data Virtual Lab: Connecting

the Dots Between Data Visualization and
Analysis
Sonsoles Lo pez-Pernas , Universidad Polite
cnica de Madrid, 28040, Madrid, Spain and also University of
Eastern Finland, Kuopio, Finland
Andres Munoz-Arcentales , Carlos Aparicio , Enrique Barra , Aldo Gordillo a
, Joaquín Salvachu , and
cnica de Madrid, 28040, Madrid, Spain
Juan Quemada , Universidad Polite

Educational Data Virtual Lab (EDVL) is an open-source platform for data

exploration and analysis that combines the power of a coding environment, the
convenience of an interactive visualization engine, and the infrastructure needed to
handle the complete data lifecycle. Based on the building blocks of the FIWARE
European platform and Apache Zeppelin, this tool allows domain experts to become
acquainted with data science methods using the data available within their own
organization, ensuring that the skills they acquire are relevant to their field and
driven by their own professional goals. We used EDVL in a pilot study in which we
carried out a focus group within a multinational company to gain insight into
potential users’ perceptions of EDVL, both from the educational and operational
points of view. The results of our evaluation suggest that EDVL holds a great
potential to train the workforce in data science skills and to enable collaboration
among professionals with different levels of expertise.

H
uman-data interaction (HDI) is a recent data sources. This is precisely the aim of the field of
research trend that claims that, in order to visual analytics: to turn decision-makers from passive
gain knowledge from data, users need to observers of visual representations to active users
interact with data instead of only passively consuming of visual analysis tools.2 To take full advantage of
them.1 In other words, visualizing data in a static way the current advances in data science and artificial
is not enough to extract meaningful insights. Instead, intelligence, it is crucial that managers and decision-
data consumers should be able to explore data in an makers acquire the skills to raise and give answers to
active way and understand it on different levels, real-world questions relying on the data available to
according to their individual preferences and capabili- them through an inquiry process.3
ties.1 Having the necessary skills for making sense of In spite of the many training opportunities in data sci-
data is of special relevance in organizations, where ence skills, most of these are often delivered through uni-
domain experts with extensive experience in a given versity courses, MOOCs, or bootcamps.4 Since such
field lack the technical wherewithal to analyze and learning approaches are mostly based on formal educa-
extract the meaningful value of the data they have tion, trainees mistakenly perceive data science as the
available from their own organization or from open application of well-known solutions to well-known prob-
lems, instead of as the research for new methods and
their extension or variation for new and challenging prob-
lems.5 Efforts to teach data science to professionals
0272-1716 ß 2022 IEEE
Digital Object Identifier 10.1109/MCG.2022.3189557 should, instead, be designed in ways that fit around and
Date of current version 3 October 2022. align with work tasks.6 According to Littlejohn,7 “many

76 IEEE Computer Graphics and Applications Published by the IEEE Computer Society September/October 2022
Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
 EDUCATION

organizations recognize that learning is not effective if EDUCATIONAL DATA VIRTUAL LAB
professionals learn a new process and then do not use In this article, we present Educational Data Virtual Lab
their new knowledge and embed it within their practice.” (EDVL), an open-source platform that enables domain
As such, the materials, tools, and data that are used for experts to become acquainted with data science meth-
learning should ideally be the same as the ones employed ods using the data available within their own organiza-
in work practice. By contextualizing data science prob- tion, ensuring that the skills they acquire are relevant to
lems in the work environment, employees learn by doing their field and driven by their own professional goals. Our
as they pursue a specific goal that is beyond their current main objective is to address the shortcomings of existing
capabilities. For professionals to acquire new skills, com- data science training solutions, which are based on for-
panies need to provide “supportive, caring, learning envi- mal education, by embedding learning into work practice,
following the guidelines of existing literature on profes-
ronments that are responsive to individual learning
sional learning.7 As such, EDVL provides a tool that can
needs and interests,” which are key in lifelong learning,
be seamlessly integrated into an organization’s infra-
especially when it comes to technical skills.8
structure, which allows users to interact with, explore
A myriad of tools exists in the market that enables
and analyze data that is relevant to their own work prac-
data analysis. The most well-known one is probably
tice at different levels. Thus, EDVL presents a favorable
Jupyter,a which provides a notebook-based environ-
environment where professionals can advance their data
ment in which users can write and run code chunks and
science skills with real problems and use the knowledge
view the results in the same interface. Although Jupyter
extracted from data as part of decision making. The
is very widely used in both professional and educational
results and software used for such analysis can be shared
settings, since it is solely code-based, it is often used
among collaborators and even made available publicly.
only by technically skilled users. To enable less-skilled
EDVL manages the whole data lifecycle, including collec-
professionals to perform data analysis by themselves,
tion, storage, publication, access, and use. For that pur-
some more user-friendly tools have emerged, such as
pose, it combines the interactive visualization features
Orangeb or Tableau,c which provide very easy-to-use
provided by Apache Zeppelind along with its support for
interactive user interfaces. Nonetheless, these tools
multiple programming and querying languages, and the
lack the flexibility that a code-based interface provides,
end-to-end data infrastructure enabled by the European
which is a requirement for technically-skilled professio-
FIWARE platform.e As such, EDVL is the result of combin-
nals. Moreover, data access should be facilitated so all
ing Apache Zeppelin as the human–data interface and
stakeholders can have the latest data readily available.
FIWARE as the underlying infrastructure that supports
This feature is not commonly provided by visualization-
the complete data lifecycle. Figure 1 shows the overall
centered tools, which often need a file (CSV or similar)
architecture of EDVL. The EDVL platform is available
as an input, and thus face the problem of data becoming
under an AGPL-3.0 open-source license in the following
rapidly obsolete. Since many different stakeholders are
software repository: https://github.com/ging/edvl.
involved in the sense-making of data,1 they need to be
Apache Zeppelin is a web-based platform that pro-
able to share a common ground for accessing data as
vides collaborative notebooks in which users can perform
well as generating and interpreting results. If the tools
data-driven, interactive analytics. These computational
that are used for data analysis are not the same as for
notebooks can greatly enhance the transparent dissemi-
data interpretation, the users of the latter will lack con-
nation of data science methods and results during
text for interpreting the results and will never be chal-
the publication process, enabling open collaboration.9
lenged to advance their skills and learn how to perform
Instead of exchanging multiple individual data files, code,
the analysis themselves. Ideally, the data analysis envi-
methods, and result files, Apache Zeppelin provides a sin-
ronment should neither pose an entry barrier for less
gle mechanism based on roles for collaborators to share
skilled users (i.e., it should offer interactive ways of data
and interact throughout the whole data curation process.
exploration and analysis) nor hinder the capabilities of
It supports a multitude of interpreters, including Scala, R,
more tech-savvy ones (e.g., it should allow them to carry
and Python, as well as the Angular web framework, which
out complex code-based analyses).
provides convenient ready-to-use interactive visualiza-
tions and enables the creation of customized ones.

a
Accessed: Apr. 20, 2021. [Online]. Available: htt_ps://jupyter.org
b
Accessed: Apr. 20, 2021. [Online]. Available: htt_p s://
d
orangedatamining.com Accessed: Apr. 20, 2021. [Online]. Available: htt_ps://zeppelin.
c
Accessed: Apr. 20, 2021. [Online]. Available: htt_ps://www.tableau. apache.org
e
com Accessed: Apr. 20, 2021. [Online]. Available: htt_ps://fiware.org

September/October 2022 IEEE Computer Graphics and Applications 77

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
EDUCATION

FIGURE 1. Architecture of the EDVL platform.

Therefore, EDVL enables technically-skilled users to Building around the FIWARE Context Broker, a rich
directly query and analyze the data using their program- suite of complementary open-source FIWARE Generic
ming knowledge, whereas nontechnically skilled users Enablers are available, each taking care of a different
can directly interact with the results of these queries and stage of the data lifecycle: data collection, storage, pub-
programs. Since both the code and the results are visible lication, and access. EDVL should be deployed on-prem-
in the Zeppelin notebooks, less tech-savvy users become ise, to integrate all the data sources relevant to an
aware of the provenance of the data, rather than just organization without compromising data privacy. Each
access a curated set of tables and graphs. Moreover, they data source (e.g., a sensor, a car, etc.) is represented as
can progressively become acquainted with the code and, an entity in the FIWARE Context Broker. All data sources
guided by their own goals, make small changes to it until must take care of updating their corresponding entity in
they can learn to write their own code and perform their the Context Broker each time a change in their context
own analyses. An example of a notebook created within information takes place (e.g., a new reading in a temper-
EDVL can be seen in Figure 2. ature sensor, a new position of a car), via the NGSI API.
In turn, FIWARE brings a curated framework of Following is a description of each of the FIWARE
open-source software platform components, termed Generic Enablers involved in EDVL and their role.
Generic Enablers, which can be assembled together Given that the Context Broker only stores the latest
and with other third-party components to build plat- context information available for each entity, the FIWARE
forms that support the development of smart techno- Draco Generic Enabler is needed to persist the historical
logical solutions. The centerpiece of any “Powered by evolution of the data. Every time a change in any entity
FIWARE” platform or solution is the FIWARE Context occurs, Draco stores the updated entity values in one of
Broker Generic Enabler, which supplies a cornerstone the supported database management systems (Post-
function required in any smart solution: management greSQL, MongoDB, MySQL, Cassandra, etc.). Such stored
of context information, enabling to perform updates, data can be made available automatically and immedi-
and bringing access to the context. First, the FIWARE ately as open data through CKAN,f an open-source data
Context Broker provides the NGSI API, used for management system whose main goal is to provide a
the integration of the components within a FIWARE- managed data catalog system for open data.
powered platform and by applications to update or The data persisted by Draco can be accessed
consume context information through a publish/sub- through any notebook created in Apache Zeppelin
scribe mechanism. Thus, data are represented in a uni- and analyzed using any of its supported tools and
fied format, facilitating stakeholders of all kinds to
become familiar with all the data sources available
f
within an organization. Accessed: Apr. 20, 2021. [Online]. Available: htt_ps://ckan.org

78 IEEE Computer Graphics and Applications September/October 2022

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
 EDUCATION

FIGURE 2. Example of a notebook created with Apache Zeppelin.

programming languages. Data coming from public premise to prevent data privacy breaches. However,
APIs and additional data sources are of course acces- data access can be limited within an organization as
sible as well. More interestingly, data can be read as a well. Hence, it is necessary to secure the access to the
stream directly from the Context Broker using the Context Broker data by managing roles and permissions.
FIWARE Cosmos Generic Enabler, which can subscribe FIWARE provides two Generic Enablers that ease this
to changes in the Context Broker entities and receive task: 1) FIWARE KeyRock and 2) FIWARE Wilma. KeyRock
their updated values using Spark Streaming. As such, is an identity management that allows us to configure
not only batch analysis can be performed on the data, roles with specific permissions that allow access only to
but also real-time analysis and visualization. Users can specific entities.10,11 In turn, Wilma enforces the specified
work together in the same notebook and combine dif- data access policies according to the role whenever a
ferent programming languages and libraries, accord- user attempts to retrieve or subscribe to data in the
ing to their own technical skills and preferences, to Context Broker.
curate and analyze data and create interactive dash-
boards and visualizations.
The results from processing the data can be sent EVALUATION
to the Context Broker (through Cosmos) as new enti- With the objective to evaluate EDVL, we used it in a pilot
ties or new values in existing entities, which are then study in which we carried out a focus group to gain
available for internal collaborators to analyze or can insight into potential users’ perceptions of EDVL, both
even be published as open data through CKAN. For from the educational and operational points of view.
instance, one of such data processing operations The focus group was performed in an online video con-
could be anonymizing and aggregating sensitive data ference session with volunteer participants (three
into less granular pieces of information in such a way males and two females). Participants were employees
that it is safe to share them publicly. of a multinational telecommunications company, aged
Finally, there are two aspects of data that are funda- 30–50, with diverse data literacy profiles, including two
mental in any organization: 1) confidentiality and 2) data scientists, one data engineer, and two managers
integrity. As mentioned, EDVL should be deployed on- with no programming knowledge. Two moderators from

September/October 2022 IEEE Computer Graphics and Applications 79

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
EDUCATION

the research team were present in the session and easily run the different chunks of code and interact with
directed the discussion. From the use of the focus the results.
group, we aimed to provide a social environment in All of the participants agreed that the possibility
which we could articulate socially formed attitudes and of generating direct queries and having the results
opinions12 toward the use of EDVL. rendered in an interactive way allows extracting infor-
The focus group was conducted following the guide- mation in a very efficient way. One issue they empha-
lines provided by the highly cited paper, “A Practical sized is the difference between queries against
Guide to Focus-Group Research,” by Breen.12 First, we transactional databases, where there is a need for a
welcomed the participants and stated the purpose of the refresh rate, or against more static databases. They
focus group. We informed them that the session would mentioned that knowing at which time the query or
be recorded and ensured that their identity and the video code was last run (explicitly stated in the notebook
recording would remain anonymous at all stages of the interface) was very helpful for knowing whether the
research study. Next, we obtained background demo- information was updated. The participants who were
graphic information and took note of their roles within well-versed in data science highlighted that it was very
the company. Then, we provided an overview of the topic: interesting “that different languages can be mixed
we introduced EDVL and asked the participants to inter- since there are users who make scripts in R and Python
act with a sample notebook. This notebook contained indistinctly,” and also SQL and NoSQL queries. They
queries to a sales database that users could run and commented that combining different programming
explore. After participants had the opportunity to use and query languages in the same environment pre-
EDVL, we began asking questions, starting from general vents users from having to make complex queries in
questions about their experience using EDVL, and build- the console or having different tools for each engine.
ing up to more specific ones. The questions revolved Another core feature that was highlighted by all
around the three main themes of HDI13: legibility (Which the participants was that EDVL was able to create
data were available? How can you extract meaningful shareable interactive reports. The importance of this,
conclusions from the data?), agency (Do you think EDVL according to one of the managers, lies in the fact that
would be useful to analyze the data in your workplace? not only static information is shared with stakeholders
What is your opinion regarding the data security aspect in other departments or levels, but also a tool that
of EDVL?), and negotiability (To which extent should the encourages them to interact and explore the data,
company data be available to all employees? How can providing insights not only about the information
nontechnical professionals benefit from EDVL? Does extracted from the data but also about the process. A
EDVL have the potential to change data management in remarkable conclusion that participants drew from
the workplace?). These questions were used to guide the exploring the interface was that using a real dataset
focus group discussion, but follow-up questions arose brings added value by allowing all stakeholders to
naturally as the conversation progressed. learn more about the organization’s data model. More-
After the session, a member of the research team over, they emphasized the usefulness of being able to
transcribed the conversations, identifying the different hide the code and show only the results to avoid
issues and recurring themes throughout the session. unnecessary information for high-level managers,
First, the whole session was reviewed, taking notes of but still being able to get the code back for auditing
the discussions raised in the focus group. Then, with the if needed.
extracted information, statements, and points of view
were divided into categories, according to the main
topics covered in the discussion, which we present in Security and Data Governance
the following. An interesting debate arose about security at two differ-
ent levels: one about access to notebooks, and another
Interactive Analysis and Reporting one about access to the databases, which was initiated
This first topic addresses participants’ ability to operate by the data scientists. For the participants, access to
EDVL and extract meaningful conclusions from the notebooks based on user and role was mandatory, espe-
query results and interactive charts in the notebook. cially in cases where the teams were separated into
Participants were asked to use the think-aloud proto- small squads to have some confidentiality and access
col14 to describe the information they could extract control. The same issue was addressed on database
from the data. All participants, regardless of their data access. A mandatory requirement was to have control
literacy skills, immediately understood the dynamics of of access to databases, whether to access transactional
the notebook provided by the EDVL tool and could or curated databases.

80 IEEE Computer Graphics and Applications September/October 2022

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
 EDUCATION

Regarding data governance, the two managers infrastructure that supports the complete data life-
highlighted that having the code and results together cycle. The main aim of EDVL is to overcome the main
facilitates the procedures for quality assurance. One limitations of existing data science training opportuni-
of the managers pointed out the possibility of auditing ties, which are excessively inspired in formal educa-
the machine learning models from an ethical and legal tion.5 By integrating EDVL within an organization’s
point of view. infrastructure, employees gain access to real and rele-
vant sources of data, and thus training is driven
Collaborative Learning and by solving real problems, making learning almost
invisible.6
Applicability in the Workplace
As stated in many data visualization tutorials and
Participants agreed that it is necessary to differentiate
manuals, “a picture is worth a thousand words,” and “a
the context in which the tool is used. In the case of
good chart is worth a thousand tables.”15 One of the
working in companies where the data science depart-
main design goals in EDVL was enabling users to inter-
ments are very large and have highly skilled teams, it
act with data according to the HDI framework. Exist-
is the teams that will choose a set of tools, and train-
ing studies on HDI, which draw from the HCI research
ing is given on these tools. In the case of unskilled
tradition,16 focus on user analysis and interaction with
teams, they see it as an advantage “to be able to
the data. Part of the contribution of EDVL is merging
make chunks in different languages and directly to
workplace learning with HDI. According to the findings
make queries that are easily editable to experiment”
obtained from the focus group, one of the most valu-
and “to lose the fear of environments with code.” The
able features of EDVL is being able to create charts
participants did recognize that for collaboration
and graphs directly from queries to databases and
tasks between departments, specifically with market-
being able to interact with them. Literature corrobo-
ing and sales, EDVL can be a good collaboration tool.
rates that HDI must follow some guidelines,17 such as
Cooperation between teams, according to the data
maximizing direct manipulation of data and minimiz-
engineer, improves performance by “speaking the
ing information overload. The fact that EDVL enables
same language.”
users to create and interact with charts and graphs,
The participants of the focus group agreed that
along with changing parameters in code chunks, has
EDVL can be a very fitting tool for training domain
been perceived by the focus group as promising for
experts in data science, as it introduces users to tech-
training and learning. The main advantage of EDVL
niques and systems with workplace databases in a
over purely code-based environments or visualization
simple and interactive environment. Although the
tools is that it supports all the stakeholders involved
data science team recognized that it is important to
in the data curation process, which is a requirement
be trained in the specific tool in which day-to-day
highlighted by HDI literature.18 By integrating code
work will be done, they said that “it is true that for col-
and interactive visualizations in the same environ-
laboration between teams or for professionals who
ment, EDVL provides a smooth learning curve for
can relate to interactive environments, with explor-
users to interact with the data at the visual level and
atory graphs and to approach the field of program-
progressively venture into modifying the code and
ming and data analysis, it can be useful.” The
writing their own. As participants in the focus group
collaborative part is the most powerful and the train-
pointed out, EDVL allows different stakeholders to
ing of nontechnical profiles is the one with the most
speak the same language and share a common
potential.
ground for generating and interpreting results. This is
the main differentiating factor of EDVL with respect to
CONCLUSIONS other data science tools, as confirmed by the partici-
In this article, we presented EDVL, a tool aimed at pants of the focus group.
empowering domain experts to acquire data science Although the main advantage of EDVL might
skills by integrating training into work practice. EDVL appear to be the Apache Zeppelin interface to interact
combines existing open-source technologies that are with data, the underlying infrastructure based on
commonly operationalized in the data science field in FIWARE brings many benefits as well, such as access-
a unified environment that can be incorporated into ing the data stored by FIWARE’s context broker in a
an organization’s infrastructure and used both for unified format (NGSI), allowing to receive data in
training and work practice. Specifically, EDVL is the streaming mode using FIWARE’s Cosmos connector,
result of combining Apache Zeppelin as the human– or relying on FIWARE’s identity manager to allow/con-
data interface and FIWARE as the underlying strain access to data. All these features would have to

September/October 2022 IEEE Computer Graphics and Applications 81

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
EDUCATION

be implemented from scratch in the databases and 5. E. Serrano, M. Molina, D. Manrique, and L. Baumela,
other systems to which Apache Zeppelin connects if “Experiential learning in data science: From the
the FIWARE infrastructure were not used as a founda- dataset repository to the platform of experiences,” in
tion. Moreover, the FIWARE initiative is very well Proc. 13th Int. Conf. Intell. Environ., 2017, pp. 122–130,
known in the European industry and many companies doi: 10.3233/978-1-61499-796-2-122.
rely on it. Thus, just by adding Apache Zeppelin as an 6. A. Littlejohn, K. Charitonos, and H. Kaatrakoski, “The
additional piece to their infrastructure, such companies role of professional learning in addressing global
could greatly enhance the way their data are opera- challenges: Tensions and innovations associated with
tionalized in the workplace. AMR,” Front. Educ., vol. 4, pp. 122–130, Oct. 2019,
Although the features of EDVL are, according to doi: 10.3389/feduc.2019.00112.
the findings of the focus group, very auspicious for 7. A. Littlejohn, “Learning and work: Professional learning
the purpose of professional training in data literacy, analytics,” in The Handbook of Learning Analytics, 1st
our study is not without limitations. The main one is ed. Alberta, Canada: Soc. Learn. Anal. Res. (SoLAR),
the limited sample used for the focus group. We 2017, pp. 269–277.
framed our research under the denomination of 8. B. L. McCombs, “Motivation and lifelong learning,”
“pilot study” because we first aimed to test EDVL in Educ. Psychol., vol. 26, no. 2, pp. 117–127, 1991,
a real corporate environment with a few potential doi: 10.1207/s15326985ep2602_4.
users, in order to gather feedback for the improve- 9. K. M. Mendez, L. Pritchard, S. N. Reinke, and D. I.
ment of the tool. The HDI literature claims that par- Broadhurst, “Toward collaborative open data science
ticipatory design and evaluation are cornerstones in in metabolomics using jupyter notebooks and cloud
the creation of HDI systems and that the develop- computing,” Metabolomics, vol. 15, no. 10, 2019,
ment of such systems should be an iterative pro- Art. no. 125, doi: 10.1007/s11306-019-1588-0.
cess involving the users.19 The feedback gathered 10. A. Munoz-Arcentales, S. Lo 
 pez-Pernas, A. Pozo, A.
from the focus group was very helpful for validating Alonso, J. Salvachu a, and G. Huecas, “Data usage and
the design of EDVL, and confirming the require- access control in industrial data spaces:
ments gathered from the existing literature. None- Implementation using FIWARE,” Sustainability, vol. 12,
theless, a more exhaustive evaluation of EDVL is no. 9, May 2020, Art. no. 3885, doi: 10.3390/su12093885.
needed with a larger sample, in which the long-term 11. A. Munoz-Arcentales, S. Lo  pez-Pernas, A. Pozo,
learning effectiveness of the tool is assessed and  Alonso, J. Salvachu
A.  a, and G. Huecas, “An
also the impact on revenue or social return of the architecture for providing data usage and access
investment for organizations. control in data sharing ecosystems,” in Proc. 6th Int.
Symp. Emerg. Inf., Commun. Netw., 2019, vol. 160,
pp. 590–597, doi: 10.1016/j.procs.2019.11.042.
REFERENCES 12. R. L. Breen, “A practical guide to focus-group
1. H. Hornung, R. Pereira, M. C. C. Baranauskas, and K. research,” J. Geogr. Higher Educ., vol. 30, no. 3,
Liu, “Challenges for human-data interaction—A pp. 463–475, 2006. doi: 10.1080/03098260600927575.
semiotic perspective,” in Proc. Hum.-Comput. 13. R. Mortier, H. Haddadi, T. Henderson, D. McAuley, and
Interact.: Des. Eval., 2015, pp. 37–48, doi: 10.1007/978-3- J. Crowcroft, “Human-data interaction: The human
319-20901-2_4. face of the data-driven society,” SSRN Electron. J.,
2. J. Kohlhammer, T. May, and M. Hoffmann, “Visual 2014, doi: 10.2139/ssrn.2508051.
analytics for the strategic decision making process,” in 14. C. Lewis, Using the “Thinking-Aloud” Method in
GeoSpatial Visual Analytics. Dordrecht, The Cognitive Interface Design. Yorktown Heights, NY,
Netherlands: Springer, 2009, pp. 299–310, doi: 10.1007/ USA: IBM T. J. Watson Research Center, 1982.
978-90-481-2899-0_23. 15. A. Krause and M. OConnell, Eds., A Picture is Worth a
3. W. G. Pothier and P. B. Condon, “Towards data literacy Thousand Tables. Berlin, Germany: Springer U.S., 2012.
competencies: Business students, workforce needs, 16. N. Elmqvist, “Embodied human-data interaction,” in
and the role of the librarian,” J. Bus. Finance Proc. Embodied Interaction: Theory and Practice in
Librarianship, vol. 25, no. 3/4, pp. 123–146, 2020, HCI, 2011, pp. 104–107.
doi: 10.1080/08963568.2019.1680189. 17. E. Z. Victorelli and J. C. D. Reis, “Human-data
4. K. D. A. Carillo, “Let’s stop trying to be “sexy”—Preparing interaction design guidelines for visualization
managers for the (Big) Data-driven business era,” Bus. systems,” in Proc. 19th Braz. Symp. Hum. Factors
Process. Manage. J., vol. 23, no. 3, pp. 598–622, Jun. 2017, Comput. Syst., 2020, pp. 1–10, doi: 10.1145/
doi: 10.1108/BPMJ-09-2016-0188. 3424953.3426511.

82 IEEE Computer Graphics and Applications September/October 2022

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.
 EDUCATION

18. A. Freitas and E. Curry, “Big data curation,” in New ENRIQUE BARRA is currently an assistant professor with the Tel-
Horizons for a Data-Driven Economy. Cham, cnica de
ematics Engineering Department, Universidad Polite
Switzerland: Springer International Publishing, 2016,
Madrid, 28040, Madrid, Spain. His research interests include
pp. 87–118.
technology-enhanced learning, video conferencing, and the use
19. E. Victorelli, J. Reis, A. Santos, and D. Schiozer, “Design
of games in education. Contact him at [email protected].
process for human-data interaction: Combining
guidelines with semio-participatory techniques,” in Proc.
21st Int. Conf. Enterprise Inf. Syst., 2019, pp. 410–421,
doi: 10.5220/0007744504100421. ALDO GORDILLO is currently an assistant professor with the
cnica de
Computer Science Department, Universidad Polite
Madrid, 28040, Madrid, Spain. His research interests include the

SONSOLES LOPEZ-PERNAS is an assistant professor with
field of technology-enhanced learning, with a special focus on
cnica de Madrid, 28040, Madrid, Spain. She
Universidad Polite
learning analytics, e-learning resources, game-based learning, and
is also with the University of Eastern Finland, Kuopio, Finland.
computer science education. Contact him at [email protected].
Her research interests include technology-enhanced learning
and real-time analytics, with a focus on educational escape
rooms and learning analytics. She is a graduate student  is currently an associate professor
JOAQUÍN SALVACHUA
member of IEEE. She is the corresponding author of this cnica de Madrid, 28040, Madrid, Spain.
with Universidad Polite
article. Contact her at [email protected]. His research interests include advanced cloud and edge
architectures, big data infrastructures, data privacy and
ANDRES MUNOZ-ARCENTALES is a researcher with the usage control, NoSQL Databases, and Blockchain. Contact
cnica
Department of Telematics Engineering, Universidad Polite him at [email protected].
de Madrid, 28040, Madrid, Spain. His research interests include
data engineering, data fusion, and machine learning in smart
environments. Contact him at [email protected]. JUAN QUEMADA is a full professor in the Telematics Engi-
cnica de Madrid
neering Department, Universidad Polite
CARLOS APARICIO is currently a lecturer for interaction for (UPM), 28040, Madrid, Spain. His research interests include
design and web-development in the ESDMadrid—Superior collaborative and social application architecture for the Inter-
School of Design, Madrid, Spain, and also a Ph.D. candidate net and the Web. Contact him at [email protected].
with the Telematics Engineering Department, Universidad
cnica de Madrid, 28040, Madrid, Spain. His research inter-
Polite
ests include HCI to interactive data visualization in Big Data Contact department editor Beatriz Sousa Santos at [email protected] or
and AI contexts. Contact him at [email protected]. department editor Ginger Alford at [email protected].

September/October 2022 IEEE Computer Graphics and Applications 83

Authorized licensed use limited to: UNIVERSITY OF LECCE. Downloaded on December 08,2022 at 16:24:21 UTC from IEEE Xplore. Restrictions apply.