zur Erlangung des akademischen Grades

TRANSITION Doctor of Philosophy

im Doktoratsstudium
FROM IN-CLASS PhD in Education
Eingereicht von

TO Ben Haas

Angefertigt am
Linz School of Education Department of STEM
OUTDOOR STEM Education

Erstbetreuer

LEARNING WITH Univ.-Prof. Dr. Zsolt Lavicza

Zweitbetreuer

REAL-WORLD Prof. Dr. Thierry Noah Dana Picard

MATHEMATICAL
Mai 2021

MODELLING

Dissertation
 JOHANNES KEPLER
UNIVERSITÄT LINZ
Altenberger Straße 69
4040 Linz, Österreich
jku.at

EIDESSTATTLICHE ERKLÄRUNG

Ich erkläre an Eides statt, dass ich die vorliegende Dissertation selbstständig und ohne fremde
Hilfe verfasst, andere als die angegebenen Quellen und Hilfsmittel nicht benutzt bzw. die wörtlich
oder sinngemäß entnommenen Stellen als solche kenntlich gemacht habe.

Die vorliegende Dissertation ist mit dem elektronisch übermittelten Textdokument identisch.

Linz, Mai

2021

Ben Haas
 2
18.05.2021Haas Ben

Acknowledgements

I want to thank my parents, my three children and the endless support of my creative wife, Talla,
who has been a source of many of the innovative thinking in this research. Without them, I would
never have been able to do this work.

I want to thank my colleague and friend Dr. Yves Kreis for his support, guidance, and
perseverance. Through the last years, we researched mathematics education and taught in
many courses’ young prospective teachers.

I want to thank my supervisor and mentor, Univ.-Prof. Dr. Zsolt Lavicza, who showed me how
important creativity is in mathematics education and supported every idea with the proper
guidance, connection and research methods. Further, I want to thank Univ. Prof. DI Mag. Dr.
Dr.h.c. Dr. Markus Hohenwarter for his support and the opportunity to study at the Linz School of
Education.
I want to thank Dr. Noah Dana Picard for the exchanges, the advice and for showing me how
mathematics are connected to nature, architecture and our world.

I want to thank Dr. Philipp Sonnleitner and Dr. Robert Weinhandl for their participation in

research. I want to thank Dr. Tony Houghton for his support and advice during the publication of

my research.

I want to thank the PhD students from Linz School of Education for sharing their ideas, working
together, and supporting each other, even though long distances.

I want to thank the University of Luxembourg, LUCET, the Ministry of Education in Luxembourg
and Johannes Kepler University to support every research.

I want to thank Georges Ney for the many inspiring conversations, his guidance and support.

I want to thank all the children, parents, teachers and pre-service teachers who took part in the
different studies.

For Robin, Emily and Fleur,

may this inspire you to reach out for your dreams.

3
18.05.2021Haas Ben

Table of contents

1. Introduction ...........................................................................................................................10
1.1. Personal experience......................................................................................................11
1.2. Educational context in Luxembourg
..............................................................................12
1.2.1. National curriculum in mathematics education...................................................12 1.2.2.
National standardised testing results in mathematics education........................14 1.3. Scientific
influences .......................................................................................................16 1.4. Structure of
the cumulative thesis and research questions ...........................................16 2. Theoretical
framework ..........................................................................................................19 2.1. Process
skills and problem-solving approach ...............................................................21 2.1.1. Process
 skills......................................................................................................21 2.1.2. Problem-solving
approach with arithmetic word problems .................................22 2.2. Mathematical modelling
approach.................................................................................23 2.3. Instructional task design
................................................................................................25 2.3.1. Real-world information
learning with an automated tutoring system..................26
2.3.2. Real-world objects learning with augmented reality (AR), digital and physical
modelling
............................................................................................................30
2.3.3. Real-world place and situations learning with outdoor mathematical modelling
within MathCityMap
............................................................................................33
2.4. Summary .......................................................................................................................37 3.
Methodology and methods....................................................................................................38
3.1. Design-based research .................................................................................................38
3.2. Research methods and contexts from our different studies ..........................................40
3.2.1. Real-world information with an automated tutoring system................................41 3.2.1.1.
Context of the study with MathemaTIC ...............................................41 3.2.1.2. Participants of
the study with MathemaTIC .........................................41 3.2.1.3. Research methods of the
study with MathemaTIC..............................41 3.2.2. Real-world objects with digital and physical
modelling.......................................42 3.2.2.1. Context of the experimentation in special needs
education ................43 3.2.2.2. Participants of the study in special needs
education...........................43 3.2.2.3. Research methods of the study in special needs
education ...............43 3.2.2.4. Context of the experimentation in remote teaching
.............................44 3.2.2.5. Participants of the experimentation in remote
teaching.......................44 3.2.2.6. Research method of the experimentation in remote
teaching.............45

4
18.05.2021Haas Ben
3.2.3. Real-world places and situations with integrated STEAM approach in outdoor
trails....................................................................................................................46
3.2.3.1. Context of the study on outdoor trails..................................................46
3.2.3.2. Participants of the study on outdoor trails ...........................................46
3.2.3.3. Research methods of the study on outdoor trails ................................46
3.3. Summary of the methodology and methods..................................................................47 4.
Publications...........................................................................................................................49
4.1. Real-world information with an automated tutoring system...........................................50
4.1.1. Fostering process skills with the educational technology software MathemaTIC
in elementary
schools.........................................................................................50
 4.1.1.1. Short description of the MEDA paper..................................................50
4.1.1.2. Summary, highlights and next developments......................................59
4.1.2. What works: Arithmetic word problems within the educational learning
environment MathemaTIC v.2017
......................................................................59
4.1.2.1. Short description of the IJMEST paper................................................59 4.1.2.2.
Summary, highlights and next developments......................................74 4.2. Real-world
objects with digital and physical modelling..................................................74
4.2.1. Case study on augmented reality, digital and physical modelling with
mathematical learning disabilities students in an elementary school in
Luxembourg
.......................................................................................................74 4.2.1.1.
Short description of the IJTME publication ..........................................74 4.2.1.2.
Summary, highlights and next developments......................................83
4.2.2. Parent's perspectives: The use of augmented reality, digital and physical
modelling in remote teaching in elementary school.
..........................................83
4.2.2.1. Short description of the ESM publication.............................................83 4.2.2.2.
Summary, highlights and new developments ....................................105 4.3. Real-world
places and situations with integrated STEAM approach in outdoor trails .105
4.3.1. Integrated STEAM Approach in Outdoor Trails with Elementary School Pre
service Teachers in Luxembourg
.....................................................................105
4.3.1.1. Short description of the ETS publication ...........................................105
4.3.1.2. Summary, highlights and new developments ....................................121
4.3.2. Connecting the real world to mathematical models in elementary schools in
Luxembourg
.....................................................................................................121
4.3.2.1. Short description of the publication ...................................................121 4.3.2.2.
Summary, highlights and new developments ....................................128 4.4. Main results of
publications .........................................................................................128

5
18.05.2021Haas Ben
5. Discussion...........................................................................................................................132
6. Limitations...........................................................................................................................135
7. Further developments .........................................................................................................136
8. References..........................................................................................................................138
 6
18.05.2021Haas Ben

Abstract

This cumulative thesis presents a PhD research on mathematical modelling with real-world
information, objects, places and situations in the classroom, remote teaching, and outdoor
learning in Luxembourg. We investigated different technology-enhanced tasks, learning, and
teaching settings that could likely engage students in understanding and transferring
mathematical modelling to their living environments through design-based explanatory studies.
The selected design-based research methodology and its characteristics allowed us to adapt
task designs, settings and methods throughout the research process. Hence, firstly, we
investigated process skills learning (e.g., mathematical modelling) with an automated tutoring
system (i.e., the educational technology software MathemaTIC) based on real-world information
from arithmetic word problems within an international project. Although findings were promising,
we redesigned interventions to connect mathematical modelling with real-world objects. Thus,
secondly, we performed mathematical modelling tasks with augmented reality applications, CAD
software and 3D printing on real-world objects in remote teaching and special needs education
with elementary school students. Thirdly, we undertook our final study utilising findings of these
studies and redesigned an intervention that was based on outdoor mathematical trails. We
explored outdoor mathematical modelling with an integrated STEAM (Science, Technology,
Engineering, Arts, and Mathematics) approach. Through such a holistic approach, we collected
data on education-related perceptions of different elementary school education stakeholders
(i.e., students, parents, in service, and pre-service teachers) and developed conceptual
frameworks on task creation, mathematical modelling, and stakeholders’ roles. In this
cumulative thesis, we aspired to contribute and to explain how in-class learning and teaching
could be connected and transferred to mathematical modelling within students’ living
environments. There are numerous crucial moments, method choices and findings within these
studies, highlighted through a selection of peer-reviewed journal articles. Moreover, we will
discuss limitations, delimitations and opportunities different stakeholders’ perspectives on
technology-based learning in mathematical modelling through real-world information.

Keywords: Mathematical modelling, Outdoor Mathematics, Elementary school, Technology

enhanced, Parents, Augmented Reality, Digital and Physical Modelling

7
18.05.2021Haas Ben

Publication portfolio

In this publication portfolio section, a selection of peer-reviewed academic papers are listed as a
requirement for this cumulative thesis. Each article can be considered as milestones in the
design based research progress and summarise essential findings of the ongoing processes.
The main contributions were categorised according to the three major studies that I carried out
with other contributors and with my supervisor during the past five years. Additionally, there are
other supplemental contributions listed in this portfolio in which we shared ideas, experiments,
and writings with the scientific community, and it also offers a background for my scholarly
progress. These publications enabled starting a dialogue and extend a recently rising community
working on technology-enhanced mathematical modelling in Luxembourg.
Real-world information with an automated tutoring system

a. Haas, B., Kreis, Y., & Lavicza, Z. (2020). Fostering process skills with the educational
technology software MathemaTIC in elementary schools. In A. Donevska-Todorova, E.
Faggiano, J. Trgalova, Z. Lavicza, R. Weinhandl, A. Clark-Wilson, & H.-G. Weigand (Eds.),
Proceedings of the 10th ERME TOPIC CONFERENCE (ETC10) on Mathematics Education
in the Digital Age (MEDA) (pp. 199–206). Johannes Kepler University. (Paper A)
b. Haas, B., Kreis, Y., Lavicza, Z. & Sonnleitner P. (In Review). What works: Arithmetic word
problems within the educational learning environment MathemaTIC v.2017. International
Journal of Mathematical Education in Science and Technology. (Paper B)

Real-world object mathematical modelling with technologies

c. Haas, B., Kreis, Y., & Lavicza, Z. (In Press). Case study on augmented reality, digital and
physical modelling with mathematical learning disabilities students in an elementary school in
Luxemburg. The International Journal for Technology in Mathematics Education. (Paper C)
d. Haas, B., Kreis, Y., & Lavicza, Z. (In Review). Parent's Perspectives: The Use of Augmented
Reality, Digital and Physical Modelling in Remote Teaching in Elementary School.
Educational Studies in Mathematics. (Paper D)

Outdoor mathematical modelling with technologies

e. Haas, B., Kreis, Y., & Lavicza, Z. (In Press). Integrated STEAM Approach in Outdoor Trails
with Elementary School Pre-service Teachers in Luxemburg. Journal of Educational
Technology & Society. (Paper E)
f. Haas, B., Kreis, Y., & Lavicza, Z. (2020). Connecting the real world to mathematical models in
elementary schools in Luxemburg. In R. Marks (Ed.), Proceedings of the British Society for
Research into Learning Mathematics: Vol. 40 (2) (pp. 1–6). British Society for Research into
Learning Mathematics. (Paper F)

Additional contributions

g. Kreis, Y., Haas, B., Lavicza, Z., & Weinhandl, R. (In Review). Transition from traditional to
hybrid to online courses for pre-service elementary school teachers at the University of
8
18.05.2021Haas Ben
Luxembourg. International Journal of Science and Mathematics Education. Springer. (Paper
G)
h. Kreis, Y., Haas, B., Reuter, R., Meyers, C., & Busana, G. (2020). Reflections on our teaching
activities in the initial teacher training during the COVID-19 crisis: From "onsite classes" to
"schooling at home". In G. Mein & J. Pause (Eds.), Self and Society in the Corona Crisis:
Perspectives from the Humanities and Social Sciences. Melusina Press. (Paper H)
i. Lavicza, Z., Haas, B., & Kreis, Y. (2020). Discovering Everyday Mathematical Situations
Outside the Classroom with MathCityMap and GeoGebra 3D. In M. Ludwig, S. Jablonski, A.
Caldeira, & A. Moura (Eds.), Research on Outdoor STEM Education in the digiTal Age:
Proceedings of the ROSETA Online Conference in June 2020 (Vol. 6, pp. 23–30). WTM.
(Paper I)
j. Huylebrouck, D. (2020). Abstraktion zum Anfassen: Zwanzig Jahre mathematische Realität mit
GeoGebra. Mitteilungen der Deutschen Mathematiker-Vereinigung, 28(3), 147–151.
 https://doi.org/10.1515/dmvm-2020-280310 (Paper J)
k. Haas, B., Kesting, F., Martin, R., Kreis, Y., & Koenig, V. (2017). Die Lernplattform
MathemaTIC: Digitales Erlernen der kognitiven Prozesse im Sachrechnen im Alter von 8 bis
10 Jahren. In U. Kortenkamp & A. Kuzle (Eds.), Beiträge zum Mathematikunterricht 2017
(pp. 1333–1336). WTM-Verlag. (Paper K)
l. Kreis, Y., Bertemes, J., Kafai-Afif, A., & Haas, B. (2016). The personalised and multilingual
mathematical learning environment MathemaTIC. 13th International Congress on
Mathematical Education, Hamburg, Germany. (Paper I)

9
18.05.2021Haas Ben
1. Introduction

"The role of the teacher is to create the conditions for invention rather than
provide ready-made knowledge." Seymour Papert

One of the core missions in elementary school teachings is to enable young students (aged 4 to
12) to understand and interact with mathematics within their living environment. Each acquired
skills or knowledge in elementary school should allow interactions, understandings and solving of
real-world situations or problems. In mathematics education, process skills and mathematical
modelling processes, the learning concepts focus on applying skills and knowledge in the real
world. Although these concepts are studied during initial elementary school teacher training,
embedded in the national curriculum, or strongly suggested by research, their applications in
classrooms are still infrequent.

Throughout our elementary school experiences, in different teaching and supervision positions,
and during the different studies we led, we observed little interconnections between real-world
and mathematics education. There is a strong tendency to use textbooks in elementary school
teaching in Luxembourg rather than applying mathematical skills and knowledge. Moreover,
there is a major focus on learning through imitation and applying mathematical concepts isolated
(i.e., no interconnections) in closed task settings (e.g., series of written calculations), without
real-world situations or problems. Further, active learning approaches with discoveries of
mathematical concepts and real-world materials decrease, in many reported cases, during
students' school time. Thus, what happens daily in early childhood (e.g., learning through play or
discoveries) is quite sporadic in cycle 4 teachings (ages 10 to 12).

Although there was a significant reform in elementary school education in 2009, with a new
curriculum suggesting more hands-on activities and a stronger focus on skills learning,
elementary schools' teaching culture did not yet respond to these suggestions. However, with
programs like the Digital (4) Education initiative in 2015 or the introduction of specialised
teachers in education (e.g., specialised in technology, special needs or school development),
new opportunities rose in classrooms and schools. Thus, new approaches were used in schools.
Teachers began to use technology-enhanced teaching, and teacher training became more
individualised, observing the needs of schools. Finally, during the latest required distance
schoolings period, due to the sanitary restriction imposed during the Covid-19 pandemic,
requested changes in teaching and enabled new teaching approaches with real-world
mathematical modelling.

Intrigued by these circumstances, we investigated mathematics education and its application to

students' living environments in some of the elementary schools in Luxembourg. We researched
the development and implementation of mathematical modelling tasks, learning, and teaching
settings during the past five years. Furthermore, we inquired about several different approaches,
from automated tutoring systems in a one-to-one setting, remote and in-class digital and physical
modelling on real-word objects and outdoor mathematical trail modelling STEAM integrated
approach—however, each approach connected to the core idea of mathematical modelling with
real-world information. Throughout the different studies, we collected data on the elementary
school stakeholders (i.e., teachers, students, special-needs students, parents, pre-service
teachers) about their elaborated technology-enhanced mathematical modelling tasks, learning
10
18.05.2021Haas Ben
and teaching settings. These studies, findings and developed frameworks allowed us to describe
and support a possible transition from a traditional textbook as well as in-class to real-world
applications of mathematics in outdoor situations.

This section will overview the structure of different studies accomplished during my PhD
research and indicate motivations, context, and scientific influences. Furthermore, it will present
research questions, which guided my work.

Apart from the section personal experience, where the author refers to himself, the author wrote
in the third person. This choice was made since most academic contributions, discussions and
findings were accomplished with other researchers, supervisors, students, and teachers.

1.1. Personal experience

During the past 15 years, I was working within the educational system of Luxembourg. I have
been teaching for more than ten years as an elementary school teacher, developed educational
contents and professional developments for in-service teachers, worked as a youth psychiatry
teacher, and ultimately concentrated on special needs education. Thus, my current position is
special needs teaching with a strong inclusive focus. Since 2015, I am co-teaching mathematical
didactics in the initial training program for future elementary school teachers in Luxembourg.

Over the years, throughout my different appointments, I passionately became interested in

exploring various aspects of mathematics education. I wanted to understand why some children
were at ease with learning and applying mathematics, and others struggled so much with the
subject. I researched and experimented with creating suitable learning opportunities for every
child, rich or poor, from low or high educational households, with or without special needs and
with gender equality. After some time, and through my daily practice, I realised the importance of
process skills necessary for learning mathematics. I observed that those who were high
performing and had healthy self-esteem in mathematics interacted, communicated, and
performed
a variety of modelling within their own environments. These students could quickly recognise
mathematical concepts and apply these concepts according to real-world situations or problems.
Hence, I conjectured that if students could learn these skills in school in a creative and
meaningful way, it would be more likely that they would increase their self-esteem and gain
higher school
based performances.

With this in mind, I joined an international project on the educational technology software
MathemaTIC in 2015 and enrolled for a PhD in psychology at the University in Luxembourg. I
researched how this technology could foster students’ process skills in a one-to-one setting
through arithmetic word problems. After this project completed and finding promising results, I
wanted to continue my research with a new focus. Thus, I enrolled at the Johannes Kepler
University to explore new technologies to connect other mathematical modelling of real-world
applications and identify more tasks and learning situations.

11
18.05.2021Haas Ben
Starting in 2019, I performed two more research studies, one on augmented reality and 3D
printing with real-world objects and one with outdoor mathematical modelling tasks utilising an
integrated STEAM approach. Both studies were conducted mainly during the COVID-19
pandemic, which allowed me to further investigate remote teaching settings.

1.2. Educational context in Luxembourg

Luxembourg's primary education comprises four cycles
(cycle 1,2,3 and 4), from age 4 to 12 (figure 1). Each
cycle lasts two years, and students need to fulfil key
skills to get promoted to the next higher cycle. In case
 students do not reach the required key skill levels, they
can work on their missing skills for an additional year in
the cycle.

Students study three different languages (i.e.,

Luxembourgish, German and French), mathematics,
be developed in the different branches are
Figure 1: Elementary school system in Luxembourg
described in the national
sciences, geography, history, arts, sports and
life and society. The skills and knowledge to
curriculum and serve as guides for teachers to plan and perform their teaching. In general,
teachers teach all topics and thus have a more general approach to didactics. However, pre
service teacher training include an intense learning process on didactical principles and
hands-on teaching in the different topics.

Implemented by law in 2017, teachings are carried out with an inclusive approach and must
recognise students' different (special) needs. Teachers and schools are required to adapt to
students' needs. Teachers receive support and guidance from specialised teachers or a
multidisciplinary team if students' needs are even more challenging for schools and teachers.
1.2.1. National curriculum in mathematics education
Teaching mathematics in elementary schools in Luxembourg involves many different topics, from
number theory to geometry, from measurements to data analysis. In the national curriculum for
elementary schools (MENFP, 2011), there are four domains for content skills (i.e. numbers and
operations, geometry, measurement, and solving arithmetic problems) and process skills
(problem-solving, reasoning and proof, representation and mathematical modelling). There are
subdomains for each cycle for content and process skills and defined key skill levels to achieve.

12
18.05.2021Haas Ben
French (MENFP, 2011)
Figure 2 describes the domain of arithmetic
problem solving and performances from cycle 3
(ages 8 to 9). Teachers can engage students in
performances such as "rephrasing wording in
their own words" or "creating a representation
of the wording", to teach and evaluate the
understanding of arithmetic word problems.
Hence, in addition to this, there are specific
content recommendations (e.g., selection of
algorithms within word problems). To evaluate
skill developments, teachers use scaled
performance descriptions to be able to identify
skill developments and attribute them to key
skill levels. Thus, in figure 3, a key skill level of
understanding and analysing arithmetic word
problems in cycle three is described as
"Students identify and note relevant information
Figure 2: Cycle 3 key skills in arithmetic word problems in
and can discard non-relevant information to be able to solve a basic arithmetic word problem".
The national curriculum
recommends
to elementary school teachers
support active student-centred
learning in their mathematics
education. Furthermore, both
content
and process skills allow
students to
understand their environment
better
and interact with this
environment
through mathematical
modelling
(Selter & Zannetin, 2018).
Hence, the
national curriculum
recommends
connecting mathematics teaching to
real-world situations, close to
students' living environments and
applying mathematical concepts to
real-world problems. These
mathematical models in real-world
connected learning should allow
students to interact with their acquired skills and Figure 3: Key skill progression on arithmetic word
problems in French (MENFP, 2011)
transfer skills and

knowledge to new situations. In elementary schools in Luxembourg, however, most teachers,

based on our findings, tend to teach mathematical modelling commonly through textbook-based
word problems. Moreover, teachers tend to present topics and walk-through manipulations to
students instead of supporting an active mathematical modelling approach.
13
18.05.2021Haas Ben
Furthermore, described by long-time results of national standardised testing in mathematics
education (2014-2020), students' performances in mathematics were lower than expected
throughout elementary school education and revealed numerous inequities.
1.2.2. National standardised testing results in mathematics education
To monitor the quality of teaching and learning in elementary schools in Luxembourg, the
Luxembourg Centre for Educational Testing (LUCET) at the University of Luxembourg performs
evaluation through the standardised large scale testing "Épreuves Standardisées" (ÉpStan), key
skills, class climate and motivation of students. ÉpStan are performed every schoolyear at the
beginning of cycle 2, 3 and 4. Moreover, these tests evaluate language skills in German and
French and mathematical skills built on contents from the national curriculum. Socioeconomic
and sociocultural variables are taken into account during the analysis to ensure a fair
comparison between students, classes and schools.

Based on the last years (2011-2020), LUCET supplied several results and tendencies in relation
to mathematics education suggested by the data analysis. According to this report, students'
performances were lower than expected, and certain inequities could hinder students from
becoming high achievers. Variables such as spoken language, socioeconomic and sociocultural
factors strongly impacted the way students perform in elementary school.

Figure 4: ÉpStan results on cycle 1 (LUCET, 2021)

Figure 4, evaluating key skills in mathematics after cycle 1, shows a distribution of students who
are either below key skills level requirements (light blue), achieved the required key skill level
(blue), or achieved higher key skill levels (orange). After the first year in cycle 1, we can observe
that most students were reaching their required skill levels, and a large number has even
achieved higher skills. Teaching in cycle 1, happened in general, based on our observations and
recommendations of the national curriculum, in a playful, active and experimental way. Students
regularly manipulated quantities and learned by explorations of new contents within their living
environments.
14
18.05.2021Haas Ben
After cycle 1, students work, as mentioned earlier, more on textbooks and manipulations took a
lesser part in the learning process. Moreover, many contents and skills were learned mostly
through imitation and based on oral guidance. Thus, looking at figure 5, we can observe that
students after cycle 2 change their performances.
 Figure 5: ÉpStan results on cycle 3 (LUCET, 2021)

In elementary school in Luxembourg, however, most teachers, based on findings in our studies,
tend to teach mathematical modelling through textbook-based word problems. Moreover,
teachers tend to present topics and walk-through manipulations with students instead of
supporting an active mathematical modelling approach. According to results from national
testing ÉpStan in mathematics, low process and content skills among students were detected
over the past five years, and the inequity gap (see figure 6 on language background) between
students widened (Sonnleitner et al., 2018). However, such a "traditional" teaching culture in
mathematics education is still widespread in Luxembourgish schools. According to Blomberg et
al. (2014), teachers often rely on their own learning experiences from their own school time and
feel uncomfortable changing their teaching approaches. Also, adopting new ways of teaching
could be risky for accountability.
 Figure 6: ÉpStan results on cycle 3 with social factor
language background (LUCET, 2021)

15
18.05.2021Haas Ben
With the emergence of educational technologies in elementary schools in Luxembourg in 2015,
which were widely supported by national policies (e.g., Digital (4) Education program), some
teachers seemed to consider utilising new methodologies (e.g., interactive tasks in language
learning). Thus, technology-enhanced tasks, learning and teaching settings could be a key
element in changing mathematics education, adapted for each child.

1.3. Scientific influences

Several scholars who described connecting mathematical modelling with real-world information
with or without technology have substantially influenced this cumulative thesis. This section will
briefly introduce their research, and later we will further clarify these ideas in the theoretical
framework section.

Firstly, the research done by Luc Trouche (2004a) on the instrumental genesis and by Raymond
Duval (2006, 2017) on semiotic registers guided the first studies on learning process skills with
technology. Moreover, in every upcoming part, their theoretical inputs were utilised to understand
how different stakeholders experienced technology-enhanced learning.

Secondly, the research carried out by Werner Blum, and Dominik Leiß (2007) on mathematical
modelling approaches described the importance of mathematical modelling and how it connects
with the real world. They described skills and processes which are likely to happen while
performing mathematical modelling tasks. Moreover, we consulted related works on their
mathematical modelling frameworks, which allowed us to discover a wide range of technology
uses to connect mathematical modelling to the real world (e.g., online tasks with interactions,
outdoor mathematical trails, use of augmented reality).

Thirdly, the research performed by Diego Lieban (2019) on digital and physical modelling, where
structured learning with 3D printing and pedagogical usages were researched, served as a role
model. His findings allowed us to adapt the PhD research on mathematical modelling and
consider new tasks, learning and teaching settings.

Finally, the research on outdoor trails by the MathCityMap Team (Cahyono et al., 2020; Jablonski
& Ludwig, 2020; Ludwig & Jesberg, 2015) also influenced this PhD research. Connecting
mathematical modelling directly to students’ learning environments while using the advantages of
technologies (e.g., augmented reality to recreate shapes or complete patterns) led to combining
 all findings from the PhD research project on mathematical modelling with real-world

information. 1.4. Structure of the cumulative thesis and research questions

This cumulative thesis will present the author's motivation, theoretical influences and
frameworks, methodologies and methods, significant academic contributions, discussion of
findings, limitations, delimitations, and an outlook to further research. We structured the different
sections and academic contributions according to the order by which studies were performed to
enable the reader to better understand the evolution and reflections of our research. As
visualised in figure 7, our overall research was utilising a design-based approach (Amiel &
Reeves, 2008; Feng Wang & Hannafin, 2005), and thus, we adapted every subsequent
research studies, in an iterative

16
18.05.2021Haas Ben
design, grounded on findings and discussions carried out during the initial research process.
Hence, studies differed in technology, settings, class or remote teaching, collected data types
and data analysis, and focused on specific elements or perspectives on mathematical modelling
with real-world information. As we will additionally describe in the methodologies and methods
section, we used a mixed-method research approach (Howe, 2012), adapted each time to
opportunities and limitations of given possible data analyses.

Figure 7 Research timeline with studies and stakeholders

We completed three studies, where each study investigated technology-enhanced tasks,
learning and modelling settings with real-world information, situations, objects, or problems.
Each study can be seen as a piece of puzzle contributing to the greater picture of
technology-enhanced mathematical modelling in real-world situations. Moreover, we explored
stakeholders' roles and needs during technology-enhanced mathematical modelling at different
levels (e.g., parents, in
service teachers, pre-service teachers, students, and special needs students) throughout the
three studies. We did this investigation to understand the needs and limitations of transitioning
from in class to outdoor mathematical modelling in elementary school. The studies have been
presented in several conferences (e.g., CADGME, BRSLM, LUXERA, INSTEAD 6, MEDA,
ICME 13, GDM, VARTEL, ACTM, ERME spring school, Linz School of Education and seminars
for SEAMEO, see appendix A), published as proceedings or published (or in review) in relevant
scientific journals (Appendix B).
Thus, our research, throughout the studies, was guided by the following questions:

1. Which technology-enhanced tasks, learning and teaching setting are most likely to foster
active mathematical modelling approaches with real-world information, objects, places and
situations?

2. How should different stakeholders (students, teachers, pre-service teachers and parents) be
supported in learning and teaching with technology-enhanced mathematical modelling?

17
18.05.2021Haas Ben
In the first study, we investigated an automated tutoring system in one-to-one learning (no
intervention from a teacher, parent or peer, only support through technology) on mathematical
modelling and process skills with the educational technology MathemaTIC (mathematic.lu). In
this study, real-world information was presented to the students through arithmetic word
problems, similar to those in textbooks, but animated, with voice guidance and digital
interactions. We aimed
to observe if students can learn process skills based on their interactive arithmetic word
problems with extensive guidance compared to a traditional paper-pencil approach in class. In
this study, our primary focus was on students with various socioeconomic and linguistic
backgrounds and their teachers.

The second study focused on connecting mathematical modelling directly to real-world objects
and problems, using the dynamic geometry software GeoGebra 3D Graphing Calculator
(geogebra.org/3d) and Autodesk® Tinkercad® (tinkercad.com). We particularly employed
augmented reality, digital and physical modelling with 3D printers. Then, we performed two
experiments in this study, one with students having learning difficulties in mathematics and one
in remote teaching, focusing on parents as assistant teachers. Our aim in both studies was to
experiment with technology-enhanced mathematical modelling with real-world objects and
investigate needs from different stakeholders.

In the final study, we extended the mathematical modelling to real-world places and situations
outdoor, outside the classroom. With the GPS supported educational technology MathCityMap
(mathcitymap.eu/en), we investigated outdoor mathematical modelling with STEAM integrated
approach with pre-service elementary school teachers. Also, we experimented with using the AR
function in outdoor trails with the GeoGebra 3D Graphing Calculator. Our focus was on
identifying
teaching patterns when working with outdoor trails and mathematical modelling.

As mentioned earlier, each study could be seen as a piece of a puzzle in how teachers, students
and parents could experience technology-enhanced mathematical modelling with real-world
information and enable the transition from in-class to outdoor learning.

In the upcoming section, we will describe theoretical frameworks of this cumulative thesis.
 18
18.05.2021Haas Ben

2. Theoretical framework

“The only way to learn mathematics is to do mathematics.” Paul Holmos

Connecting mathematics to students' everyday life situations, problems, or objects has been
described in several publications starting in the past Century (Hunter et al., 1993). In almost
every textbook, mathematical concepts were presented in non-mathematical surroundings (e.g.,
arithmetic word problems) and suggested to connect mathematics with real-world information
(Smith & Morgan, 2016). With technology-enhanced teachings, new possibilities emerged to
assist in teaching mathematics based on situations, problems, or objects involving students'
living environments. However, when designing such teaching settings and tasks, there are
several approaches, skills, and instructional designs has to be considered to further enhance
students’ learning opportunities. Not every learning activity connecting mathematics to
real-world information is per se meaningful for students and allows efficient learning. Thus, we
designed a sequence of studies with particular theoretical frameworks that consider essential
aspects of technology-enhanced teachings designs and tasks with real-world information,
objects, places and situations.
 Figure 8 Cell of mathematical modelling

Our theoretical framework (figure 8) was built upon three themes (process skills, problem solving
approaches and mathematical modelling approaches) where each theme aims to describe
students’ mathematical understanding through technology-enhanced tasks and teaching settings
with real-world information, situations, objects or problems. Moreover, the complexity of
instructional design of technology-enhanced tasks and teaching relies on the interconnections
among these different themes. Thus, in the framework, we illustrated the different themes and
their interconnections and how they are built on each other (e.g., process skills are needed in
19
18.05.2021Haas Ben
problem solving approaches). The mastery of each theme, as we will describe in this section,
play a dynamic role in connecting mathematical concepts to real-world or understanding
real-world based tasks. Hence, connecting mathematics to the real-world, by wording, real-world
objects or places, touches multiple different theories (e.g., instrumental genesis, mathematical
modelling approaches, process skills, problem solving or STEAM integrated approach, however,
there are
many similarities and connections, as mentioned before, among these theories. This section will
provide the different themes and illustrate their interconnections and their relevance to our
studies. We did not intend, however, to procure an in-depth analysis of the different theories but
provide the reader with information on our theoretical background and explain why we referred
to them in our studies.

The themes process skills and problem-solving approaches are often used in curricula, policies
or elementary school teacher training, and they are subjects of several studies in literature
(Baffrey Dumont, 1996; Bossé et al., 2010; Franke & Ruwisch, 2010; Liljedahl et al., 2016;
NCTM, 1999; Selter & Zannetin, 2018; Singer et al., 2013; Verschaffel et al., 2000). There is a
common understanding that a mastery of problem-solving approach and related skills are
needed for mathematical modelling (Blum & Leiß, 2007; Blum & Niss, 1991; MENFP, 2011),
although that there are different interpretations (e.g., different descriptions of skills or skill
groups). Thus, in the first part, we will describe the themes of process skills and problem-solving
approaches, this with real-world information based on arithmetic word problem, close to
elementary school students’ experiences in Luxembourg. Our first study (Haas et al., In Review,
2020b) and the beginning of our theoretical research were influenced by these themes and
real-world information tasks. Based on the literature review, we will present a reference for
process skills and problem-solving approaches we designed for our theoretical framework.

While exploring definitions of process skills and problem-solving, we noticed through our
literature review and experiences that these skills or procedures, although essential, are among
a more complex approach when tasks are directly related to real world (e.g., situations or
objects). This approach could be qualified as mathematical modelling approaches. Several
didactical approaches described this approach over time (Blum & Kirsch, 1989; Blum & Leiß,
2007; Blum & Niss, 1991; Freudenthal, 1971; Greefrath & Vorhölter, 2016; Maaß, 2006;
Schukajlow et al., 2012), with different lenses on models (e.g., descriptive or explicative) and
dynamics (e.g., iterative, single way) and in extend to process skills and problem-solving
approaches, there is a clear focus on mathematical modelling skills (Blum & Borromeo Ferri,
2009). We choose the mathematical modelling approach described by Blum and Leiß (2007)
due to its close connections to mathematics didactical teaching in Luxembourg and the
integration of the situation model (Nesher et al., 2003) in the approach, which we identified in
our studies as a vital element with technology enhanced application of mathematical concepts to
real-world information, situations or objects.
Thus, in the second part, we will describe further the mathematical modelling approach through
literature reviews and experiences.

In the third part, we will highlight the different instructional designs that we used in our studies.
Since each study employed different technologies, in class, outdoor or remote, and with different
stakeholders (e.g., special needs students, students, teachers, parents), the utilised instructional
designs were also different. We used the 4C-ID design (Melo & Miranda, 2015) within an
automated tutoring system, the four basic principles of Dienes’ theory on mathematics learning
20
18.05.2021Haas Ben
(Dienes, 1960) with 3D modelling of objects and extended existing designs (Bacca Acosta et al.,
2019; Cahyono et al., 2020; Kelley & Knowles, 2016) of outdoor tasks. Every task design
adapted to the study, however, connected to process skills, problem solving and mathematical
modelling approach. Thus, in the instructional design task part, we will describe for every study
the used framework.

2.1. Process skills and problem-solving approach

In preparation for our first study, we investigated learning arithmetic word problems with the
educational technology software MathemaTIC. Arithmetic word problems are usually used in
Luxembourg's elementary schools to teach or test skills while applying mathematical concepts to
real-world information and are listed as a specific domain in the curriculum. Based on our
observations, arithmetic word problems are often used at the end of teaching sequences in
elementary schools, after students learned arithmetic operation procedures. Word problems
teachings represent a minor part of the overall teachings and are proposed and solved by strong
performing students in many cases. Teachers reported in our studies that students do not
struggle with arithmetic operations but rather battle with the understanding and identification of
the mathematical concept within real-world information.

According to Franke and Ruwisch (2010) and the national curriculum, students need masteries in
two different skills, content and process skills. Content skills are related to structured procedures
(e.g., performing addition of two given numbers, calculating a volume), and process skills refer to
a skill set that allows to identify mathematical concepts within problem wordings or situations,
establish solution paths or self-regulate their solving processes (MENFP, 2011). Throughout our
studies, we investigated mainly process skills since, as mentioned earlier in the introduction,
these skills are instructed less in schools, and the primary focus of teaching focuses on content
skills. Moreover, there are numerous educational technologies designed to assist in acquiring
content skills and the focus of the designed module, in our first study, with the educational
technology MathemaTIC emphasised process skills development.
2.1.1. Process skills
According to Singer et al. (2013) and Liljedahl et al. (2016), however, process skills are critical
elements in connecting mathematical concepts to real-world information and used iteratively
during a solving process (Schukajlow et al., 2012). Thus, we investigated process skills, which
various researchers and organisations have already described. Thus, we wanted to establish a
common understanding between international references (Bossé et al., 2010; NCTM, 1999),
mathematics didactics references used in elementary schools (Selter & Zannetin, 2018) and the
national curriculum (MENFP, 2009, 2011; MENJE, 2020).

We categorised the process skills from different references (Bossé et al., 2010; MENFP, 2011;
MENJE, 2020; NCTM, 1999; Selter & Zannetin, 2018) and created connections between the
definitions. We found significant intersections, and the main differences were due to socio
linguistic aspects used by authors or organisations languages (e.g., NCTM is written in English,
the national curriculum is written in French, didactic principles are written in German). We were
able to identify four major skills based on our analysis of references (table 1).

21
18.05.2021Haas Ben
Table 1 Identified Process skills

Skills Description

Problem Solving Engage in a task where the resolution method is not known in
advance. Transfer skills and knowledge in this task to find a solving
path.

Reasoning and Argue, communicate, analyse, and justify the conjectures made in
Proof the task.

Representation Represent conventional or unconventional models to solve the task.

Modelling Modulate, connect, and communicate mathematical concepts to

solve the task.

The identified process skills are strongly connected to the domain of language-based skills (e.g.,
communicate, argue, or justify) and needed daily practice. Process skills are even described as
transferable skills in literature and curriculum, as these skills engage in an active learning
approach (Franke & Ruwisch, 2010; MENFP, 2011; Selter & Zannetin, 2018). Process skills allow
to identify, modulate, transfer, and apply content skills. Among the four identified process skills,
one described problem-solving. In mathematics education, problem-solving skills are related to
problem-solving approaches, which were widely discussed and researched. Thus, to extend the
understanding of the skill sets students should learn in our technology-enhanced tasks, we
explored different problem-solving approaches (e.g., Mayer, 2005) to identify sub-skills,
procedural and non-procedural knowledge.
2.1.2. Problem-solving approach with arithmetic word problems
In literature, we identified problem solving approaches, and similarly to process skills, we
compared these approaches and adapted them to the curriculum and mathematical didactical
principles present in Luxembourg. Furthermore, based on several problem-solving approaches
(Liljedahl et al., 2016; Mayer, 2005; Vergnaud, 1982; Verschaffel et al., 2000) and connecting the
identified process skills, we elaborated a conceptual framework for our task designs (figure 9).

Figure 9 Problem solving

approach for arithmetic word problems
22
18.05.2021Haas Ben
Each step in this approach (e.g., understanding and representation of wordings, mathematical
modelling, solving arithmetic operations, evaluation of results and mathematical models) requires
the development of specific skills in addition to process skills.

Understanding and representation of the wording: According to Kintsch (1986),

students combine the textbase of the wording with long term memory and create a
situation model. Moreover, in the landscape model (van den Broek, 1995), these
structural representations are built upon two comprehension sources. The first is based
on students' memory, where possible mental representations are generated from the
wordings, and the second is verifying the connections between information and
consistency of the representation (Bianco, 2015).

Mathematical modelling: According to Baffrey-Dumont (1996), students reconstruct

mathematical concepts based on understanding connections between the different
elements in the wording. The mathematical modelling approach depends on the
typologies of arithmetic word problems (Verschaffel et al., 2000).

Solving arithmetic word problems: Once the wording typology (Vergnaud, 1982) is
identified, the according arithmetic operations can be applied. Students procure a result
to the initial result.
 Evaluation of results and mathematical models: Students self-regulate their
assumptions during the wordings analysis, the created schemes and the transfers to
mathematical models (Duval, 2006).

In the first study, we used this framework (figure 9) to identify skills to develop with the
educational technology MathemaTIC, based on arithmetic word problems offering real-world
information. However, in the upcoming studies, we focused our attention on tasks connected
directly to real world objects, situations, or problems out of students' daily lives. Students could
connect process and content skills to real-world problems and situations directly compared to
teaching written tasks in textbooks. Consequently, elementary school students can transfer
information directly from real world to mathematical models, apply mathematics and employ
results in the real world to solve problems. In our studies, process skills and the problem-solving
approach showed, however, to be only one part of the needed skillset when connecting
mathematical concepts to real-world information, objects, situations, or problems. Hence, we
investigated mathematical modelling approaches, which several studies referred to while
connecting directly to real-world objects or situations (Lieban, 2019; Ludwig & Jesberg, 2015). In
the subsequent section, we will describe the mathematical modelling approach we referred to in
our studies.

2.2. Mathematical modelling approach

We carried out a literature review on several studies and frameworks on mathematical modelling
approaches (Blum, 2013; Blum & Kirsch, 1989; Blum & Leiß, 2007; Blum & Niss, 1991; Greefrath
& Vorhölter, 2016; Lieban, 2019; Liljedahl et al., 2016; Maaß, 2006; Selter & Zannetin, 2018;
Singer et al., 2013). Among the consulted studies on mathematical modelling approaches with
real-world, we focused on the framework of Blum and Leiß (2007). This framework is closely
23
18.05.2021Haas Ben
related to curriculum developments in Luxembourg's elementary schools, the initial teaching
training (e.g., pre-service teachers didactics curriculum, new technology developments in
mathematics), and it describes active mathematical modelling process with real-world objects
situations and problems, which is part of the national curriculum. Moreover, within this
framework, we can find the concept of situation model (Nesher et al., 2003), which refers to the
complexity of understanding and decomposing real-world information objects or situations into
schemes and models. The creation of a real-world model to compare to different mathematical
models has shown to be an essential part of the learning of mathematical modelling. Referring
to the collected data on teacher observations and experience from mathematics education
(Haas et al., 2020a), the situation model is a key element in connecting mathematical concepts
to the real world. We adapted the framework for our studies (figure 10) by adding real objects
(e.g., creating daily life objects or completing 3D puzzles) since we used digital and physical
manipulation throughout the second and third study.
 Figure 10 Adapted mathematical modelling approach from Blum and Leiß

According to Blum and Leiß (2007), there are 7 steps, which are partly similar to the problem
solving approach we described in this section before (e.g., creating a situation model or applying
mathematical concepts). Thus, (1) students construct a situation model based on real-world
information; and (2) simplify this model to a real problem (e.g., What is the given problem? What
do we need to solve?). (3) This real problem is then mathematised (i.e., connecting
mathematical knowledge and skills and applying the most appropriate mathematical concept);
and (4) worked mathematically. (5) Students interpret results; (6) proceed to validation by
transferring mathematical results to the real world; and (7) finally exposing them to a real-world
situation. This modulation process can be iterative and further relies on students' content and
process skill development (Moffett, 2012).

In several studies based on Blum and Leiß (e.g., Ludwig & Jesberg, 2015), the use of
educational technology was presented as a critical element to mathematical modelling with
real-world situations, objects or problems. This, either in increasing motivation (Bacca Acosta et
al., 2019),
rendering new possibilities in visualising or training problem skills (Liljedahl et al., 2016) or in
structuring teaching and learning settings (Haas et al., 2020b).

24
18.05.2021Haas Ben
Although this modelling approach reveals several similarities with the problem-solving
approaches that we used with arithmetic our word problems, connecting directly to the real world
makes them different. In this approach, students can manipulate real-world objects, experience
real-time situations, try out several models, and get feedback from the real world. Therefore, the
use of educational technology and the instructional task design were different from our first
study. We shifted from an automated tutoring system with MathemaTIC to digital tools which can
connect the real world with mathematics.

In the last part of this theoretical framework section, we will describe the different instructional
designs that influenced and guided the task designs of our different studies.

2.3. Instructional task design

Due to the characteristics of design-based research (Feng Wang & Hannafin, 2005), we carried
out different studies with technology-enhanced teaching and tasks, each time based on findings
from our previous experiences. However, technologies and teaching differed (e.g., real-world
with word problems, real-world objects with 3D modelling or real-world situations with GPS
supported software). Thus, we investigated technology-enhanced mathematical modelling with
real-world information from wordings (study one with MathemaTIC), 3D modelling of real-world
objects (study two with GeoGebra 3D Graphing Calculator and Autodesk® Tinkercad®) and
modelling with real
world situations (study three with MathCityMap and GeoGebra 3D Graphing Calculator). For
every study, the context and the participants were different. In our first study, we focused on
in-class learning and collected data from students in grade 3, their teachers’ observations and
technical data. (e.g., used hardware in school, Internet connectivity). In our second study, our
focus was on in-class teachings with special needs students, but due to COVID-19, we had to
shift our attention
to remote teaching and how parents assisted teaching at home. In our last study, we focused on
pre-service teachers. We aimed to gain as much information on technology-used mathematical
modelling of different stakeholders in elementary school teaching.

To develop these technology-enhanced tasks and teaching settings, we focused on several

instructional design theories adapted to each situation. However, throughout the different tasks,
we considered the theory of instrumental genesis (Lieban & Lavicza, 2019; Trouche, 2004a),
which describes how users (e.g., students, pre-service teachers, parents) use digital tools with
technology-enhanced tasks and teaching. In instrumental genesis theory, students’ experiences
with digital tools can be analysed within two essential process stages, known as
instrumentalisation and instrumentation stages. The instrumentalisation process describes the
opportunities and the limitations of a digital tool for the user, and the instrumentation process
describes how the user is finally using the digital tool, based on created schemes and
experiences (Lieban & Lavicza, 2019; Ruiz-López, 2018). Thus, we aimed at a double purpose
of our technology use: to facilitate the learning of new skills and to allow students to appropriate
the technology and instrumentalise the digital tools and their functions (e.g., Augmented Reality,
calculation tools, measurement tools, 3D modelling) present in the technologies. A digital tool
alone, independent from the user, could not procure enough information on learning such
complex and personal skill development such as process skills or problem solving. The digital
tool and its

25
18.05.2021Haas Ben
manipulation would create learning situations and, as we hypothesised based on the described
two-phased of instrumental genesis, allow the development of new skills.

The study participants used functions of digital tools in a set of tasks, usually starting with high
guidance, up to free use and application on newly created open learning situations. Our designs
seek each time an increment of users’ evolvement within these digital functions and training of
process skills, problem solving approaches and mathematical modelling approaches. We will
present the different instructional designs we used for the three studies' different technologies
and task designs.
2.3.1. Real-world information learning with an automated tutoring system
Our research began with a one-to-one learning task design (no teacher interference in the
learning process), with the educational technology software MathemaTIC on process skills and
problem solving approach with arithmetic word problems. We created a learning environment
within an international cooperation project to learn process skills embedded in a problem-solving
approach. We referred to the model described earlier in this theoretical framework (figure 9 on
problem solving approach). The objective was to create an automated tutoring system, close to
feedbacks and scaffoldings from a teacher, meaningful tasks with real-world information and
digital functions, supporting process skills in a problem-solving approach with word problems.

Figure 11 MathemaTIC module map

After a literature review, we based on the Competence-Learning-Intervention-Assessment (CLIA)
model by De Corte et al. (2004) and the Four-Component Instructional Design (4C-ID) model by
van Merriënboer and Kester (2005) for learning complex skills to develop an automated tutoring
structure and tasks within the educational technology MathemaTIC. According to Melo and
Miranda (2015), both models have shown to be promising in supporting students either with weak
26
18.05.2021Haas Ben
or strong knowledge of complex skills, which we estimated would respond to the high
heterogeneity in classes. Figure 11, the user interface within the module, visualises the
application of both models. Thus, students experience learning, starting with structured
worked-examples (Kalyuga & Sweller, 2004), designed with high instructional guidance for each
new function, process or content skills. These tasks (blue in figure 11) were related to real-life
situations linked
to student’s living experiences (e.g., visiting a local zoo or travelling by tram) and prepared for
less guided tasks, with procedural information on routines (e.g., type of arithmetic operations to
perform) within the tasks.

To develop non-routine procedures (e.g., problem-solving, understanding wordings,

mathematical modelling and self-regulation), we designed the first tasks in the module to foster
these procedures with tools possessing special functions (e.g., highlighting information in the
wordings and creating a structural representation) with high instructional guidance. This is
similar to the observed classroom teaching, where a teacher would explain with hands-on
activities (e.g., use a colour pen to underline important information or draw a scheme with colour
pens) the procedures and support students (figure 12).
 Figure 12 Digital
tool explanation tutorial
We used the mentioned high instructional guidance, due to the absence of teachers in
MathemaTIC, to fully explain the procedure and several digital tools to train these procedures.
With these tools, we aimed to support students in learning the assigned non-routine procedures.
Overall, we presented three tools to the students: a marker to highlight information and transfer
information to conceptual representations, a planning and calculation tool to identify and perform
mathematical concepts, and a logbook to support students in self-regulating the solving process.
Based on the instrumental genesis theory described by Trouche (2004a), we designed these
tools within the learning environment that enabled students to develop the described non-routine
procedures through the instrumentalisation and instrumentation (section 2.3 on instrumental
genesis) of the digital tools functions (e.g., creating a scheme, decomposition of numbers or

27
18.05.2021Haas Ben
highlighting information in a wording). Students performed cognitive processes supported by
these digital functions (e.g., transferring information of written wordings into a conceptual
representation or code a mathematical model with imminent feedback).
 Figure 13 Planning and calculation tool

We also created a fictitious character for each process skill (table 1 identified process skills)
within the problem solving approach (figure 9) to be guided with (non-adaptive) high instructional
guidance for students during solving tasks. The characters were, a journalist to guide students in
creating a situation model based on the wording of problems; a coach to prepare a plan for the
solution process, a construction worker to solve mathematical problems, and a detective to assist
self-regulate the solution process. These fictitious characters reappeared in the different tasks in
the module, and students could request automated support by clicking on the character (e.g.,
receiving guiding questions on using digital tools). The characters were designed similar to the
teachers' scaffoldings and supports observed in class and perform as “virtual teachers” without
an adaptive approach. Moreover, we designed the characters within the recommendatations and
rules of multimedia design (i.e., low distractors, human guidance, voice guidance) (Mayer, 2005)
and reduced the cognitive load through predefined numbers of stimuli and information in each
stage (i.e., less than 7), according to findings of Kalyuga and Sweller (2004). Finally, characters
were intended to offer a set of scaffolding (i.e., auditive, visual or written language) to support
students in multiple ways.

After completing tasks with high instructional guidance, students worked on tasks with optional
guidance (red tasks in figure 11) and complex (i.e., more than one operation and multiple
distractors in the wording) tasks without guidance (black tasks in figure 11). Hence, similar to 4C
ID (Melo & Miranda, 2015) and CLIA (De Corte et al., 2004), students worked from high
guidance training tasks to complex word problem tasks without guidance. Thus, the intention
was to develop skills at the beginning repetitively through worked-examples (Sweller et al., 2011)
and with various tasks (e.g., different word problem typologies and 1- or 2-step problems). Once
word problem typologies and solving approaches became routines, students instrumented their
digital tools with
28
18.05.2021Haas Ben
their functions for their solving purposes. Thus, during the learning process, tasks increased in
distractions and complexity to let students experience different wordings and train the learned
skills. Further, scaffoldings by fictive characters became optional and students self-regulated
their learning processes by requesting scaffolding when they needed it (Zimmerman, 2002).
 Figure 14 Red task within MathemaTIC
The scaffolding system, based on the multimedia learning theory (Mayer, 2005), cognitive load
theory (Kalyuga & Sweller, 2004) and instructions to increase students thinking processes
(Pressley et al., 1996) consisted in listening to the wordings (auditive support), interacting with
the images and arithmetic operations (visual supports), and automated their solving process
through (non-adaptive) guidance. The tasks were autocorrected, and direct feedback was given
to the students (e.g., the correct results or steps were highlighted in green). All the tasks were
interconnected, and students followed an indicated learning pathway, which was structured
according to the CLIA and 4C-ID frameworks, as mentioned earlier. The automated tutoring was
designed to support students without the need for assistance from teachers or parents. Thus, we
added the characters, digital tools and their functions, a structured learning path with tasks
reaching more complexity during the learning process and involvement of students (e.g., self
regulating the learning). The study we led with MathemaTIC and the design process of the
module showed us how much it is essential for students to create situations based on real-world
information. Our findings, which can be found in papers A and B (see portfolio section), reveal
that students’ performance increase through solving tasks based on written real-world
information. However, there was still a major gap between those students proficient in
languages (i.e.,
Germanophone students compared to non-Germanophone students). This observation confirms
results presented in the introduction section on national testing mathematics education, where
we could observe a similar gap (Sonnleitner et al., 2018). Thus, after this first experience with
technology-enhanced learning, we considered connecting mathematics directly to the real world,
by manipulating objects with 3D modelling software and 3D printing. Our second technology
enhanced tasks then were developed based on real-world objects. Yet, we kept the importance
of the situation model in our design and followed mathematical modelling approaches
recommended

29
18.05.2021Haas Ben
by Blum and Leiß (2007). Since we would create tasks with object manipulations, our
instructional design needed adaptation as well (i.e., Dienes’ theory of mathematical learning).
2.3.2. Real-world objects learning with augmented reality (AR), digital and
physical modelling
In this part, we describe the instructional design from our second study on mathematical
modelling with real-world objects. Like study one, we designed tasks and teachings to engage
students with new skills and use digital tools with specific functions (e.g. AR, 3D modelling or 3D
printing preparation). The theory of instrumental genesis (described in section 2.3) and the
mathematical modelling approach (section 2.2) were used to choose tools and to prepare for the
manipulations of digital tools. However, as mentioned, in a mathematical modelling approach
with real-life objects, technologies were used differently. We did not refer to another automated
tutoring system to learn to apply mathematical concepts to the real world, but we used the digital
tools and their function directly on real-world objects. In this second experience, we could apply
real-world mathematics directly (i.e., no description of situations or objects within a word
problem). There have been several digital tools that were used to manipulate objects in the real
world or develop new manipulatives. Among these, we found GeoGebra 3D Graphing
Calculator, used in various mathematical modelling settings by Diego Lieban (Lieban, 2019) and
the CAD software Autodesk® Tinkercad®, which was used in different educational settings to
explore new opportunities in STEAM learning (O.-L. Ng, 2017).

Figure 15 Use of augmented reality to reconstruct a wooden house and a sandpit

Furthermore, we used the functions of AR, 3D visualisation and manipulation of solids and
forms, and physical modelling with 3D priniting. These functions were also used in several
studies (Bacca Acosta et al., 2019; Lavicza et al., 2020; Liu et al., 2019; O. Ng & Chan, 2019;
Sung et al., 2015; Wu et al., 2013) and positive reinforcement on motivation, understanding of
mathematical concepts, support on representation and visualisation of solids and connecting
real-world to mathematical concepts were reported. These findings led us to use the functions
for mathematical modelling with real-world objects, and we investigated task designs related to
students’ everyday lives. We will describe the different functions in the upcoming paragraphs
and present the employed frameworks in instructional design for this second study, with
GeoGebra 3D Graphing Calculator and Autodesk® Thinkercad®.
30
18.05.2021Haas Ben
AR could be defined as virtual information presented in various ways. Some applications
displayed through smartphones or tablets in allow simple visualisations (e.g., information
real-world environments (Cakir & Korkmaz, that appears while scanning a QR code),
2019). Students experience through some allow modelling of the represented
smartphones or tablets virtual information and virtual information (e.g., mathematical
observe its properties in three dimensions. modelling with geometric shapes in dynamic
This virtual information, however, can be geometry software). According to Wu et al.
 Figure 16 Students modelling an egg cup in remote
teaching

(2013), AR allows visualising an object's hidden structure (e.g., a geometric shape represented
in two dimensions) and, therefore, to render complex ideas more understandable. Moreover,
students’ interactions with the educational content through AR are likely to lower the barrier of
computer-based learning for students and support the development of students’ repertoire of
learning interactions with real-world physical objects (e.g., spatial contiguity) (Bujak et al., 2013).
Several studies on AR in mathematics indicated positive learning effects among students (Bacca
Acosta et al., 2014, 2019), such as increased motivation or better understanding of properties of
geometric shapes (González, 2015; Liu et al., 2019). Further, AR is likely to improve spatial skills
(Martín-Gutiérrez et al., 2015). According to Lieban (2019), similar positive effects on learning
and understanding geometry are supported through digital and physical modelling (e.g., creating
artefacts with a 3D Printer). However, the sole use of AR in mathematical modelling is not
necessarily proofing learning effects in mathematical concepts, spatial skills or applying
mathematics to the real world. Students should use technologies with AR functions in
well-defined settings with clear objectives, according to Billinghurst and Duenser (2012). In our
study on real world objects modelling with special needs students and the remote challenge
during COVID-19, we investigated frameworks for our instructional designs, which could procure
clear and structured settings.

Our previously used frameworks (CLIA and 4C-ID) were aimed at digital learning environments,
however, in this experiment with real-world objects, we used manipulatives (e.g., cubes,
designed 3D printed objects) and thus, we need a framework that would better fit to this
environment. We chose to use the four basic principles of Dienes' theory of mathematics
(Dienes, 1960) as a framework for designing these tasks (table 2). Moreover, this framework
was already successfully used in other experiences with real-world object modelling (Lieban,
2019). In our paper, special needs education and real-world objects (Haas et al., In Press), we
justified our framework choice and highlighted alternatives to the learning by abstraction with
Dienes’ principles (Borasi & Borasi, 1984).

31
18.05.2021Haas Ben
Table 2 Dienes' principles and examples of tasks and scaffoldings

Principles Possible scaffoldings Task content examples

 Dynamic Trial and error Soma Cubes and reproducing
Principle High instructional guidance with Shapes
worked examples

Perceptual Feedback through software and Identification of shape properties

Variability teacher Reconstruction of given cube
Principle puzzles

Mathemati Discussions about possible solutions Completing cube puzzles and forms
cal
Variability Calculating area and perimeter of
Principle given polygons

Constructivi Feedback from teacher/peers/parents Construction of different variants of

ty cube puzzles
Principle

The tasks reached from reproducing and recreations to free construction and combination of
different geometric shapes. With AR and CAD software real-word objects can be transferred into
three-dimensional digital modelling (i.e., Dynamic Principle) or do an approximate modelling of
the existing rea-world objects (i.e., Perceptual Variability Principle) by adding small add-ons.
Further, different objects could be combined (i.e., Mathematical Variability Principle) to a new
geometric model or create an entirely new construction (i.e., Constructivity Principle). Spatial
reasoning in geometry (Sinclair & Bruce, 2015), is supported by the used digital functions, AR
and CAD software (Clements & Sarama, 2011; Liu et al., 2019). During the past years, and
under the restrictions of COVID-19, we searched for opportunities to present tasks for students.
Moreover, we tried to get involved with all the different stakeholders (students, parents and
teachers) in schools. Thus, we started in a regular grade 4 class to work with AR and CAD
software on real world objects, and at the same time, we continued a long-term project in special
needs education. In the middle of our special needs project, we had to stop the in-class
real-world object modelling
because of the pandemic. However, we continued engaging in remote challenges with
elementary school students and their parents and collected data on these task designs and on
parents needs while assisting their children (further details in papers C and D).

Hence, overall, we created several tasks for special needs education, remote teaching and in
class teachings, based on the four basic principles of Dienes (table 2) and mathematical
modelling approaches (section 2.2). While our focus in this second study was to investigate
real-world objects and mathematical modelling, we identified some positive effects on students'
achievements and motivation and parents' needs in assisting their children. We concluded that
students connected mathematics differently with the real world than in the first study (automated
tutoring system with real-world information). We were not able to observe a difference between
students with high and low language proficiencies. However, we were able to observe that
socioeconomic background of students (e.g., technology access at home, education level of
parents or access to support) were still hindering our modelling approaches. Furthermore,
depending on students’ or their parents’ assistance, designs of real-world objects were more
complex and connected to additional mathematical concepts (e.g., octahedrons in egg cups,
physical experimentations with designs). These observations are similar to the results from the
32
18.05.2021Haas Ben
national testing ÉpStan in Luxembourg, presented in the introduction (Sonnleitner et al., 2018).
Therefore, to be able to enable students to connect mathematics with the real world, we
investigated a third type of technology task design and transferred tasks directly to students’
living environments. Utilising the GPS-supported MathCityMap software, we designed outdoor
trails with mathematical modelling tasks. In these tasks, students still needed to create situation
models (Nesher et al., 2003) of real-world problems, with mathematical modelling approaches
(Blum & Leiß, 2007), similarly to our two previous studies (automated tutoring on real-world
information, 3D Modelling with real-world objects). Within MathCityMap, both AR and CAD
software modelling was used. Thus, instrumental genesis on how students instrumentalised and
instrumented digital tools and their functions. In the next section, we will present our design
tasks that we used with our mathematical modelling on real-world outdoor places and situations.
2.3.3. Real-world place and situations learning with outdoor mathematical
modelling within MathCityMap
In the final study, we investigated outdoor
mathematical modelling tasks with the
technology MathCityMap on real-world
places
and situations. MathCityMap is a GPS
supported
educational software in which teachers
can
create outdoor tasks with scaffoldings,
hints,
different questions (e.g., approximations
of
values, recreation of shapes) and solutions.
Tasks are bundled within a trail, and students are
guided from task to task. This technology can be
used on a variety of devices, and trails can be
downloaded and solved without an Internet Figure 17 Task within MathCityMap
connection, which we found increased

accessibility for teachers, students and parents. We investigated teaching patterns with an
integrated STEAM approach in pre-service teacher training based on the mathematical modelling
approach we described: process skill and problem-solving approach. This choice was made
during the COVID-19 pandemic, while we needed to adapt our regular lectures at the university
to follow new restrictions. Moreover, it allowed us to observe how novice first-year pre-service
teachers could engage with these new technology-enhanced tasks and teaching. We started
working
STEAM concepts rather than solely on mathematics tasks since in outdoor mathematics with
real world situations and places, a mathematical approach was insufficient in many cases. While
designing the tasks, we observed that mathematics was connected to arts (e.g., monuments,
paintings, sculptures), engineering processes (e.g., estimate the time of movements or measure
distances of functional objects), science (e.g., spatial orientations, patterns in recreated natural
objects, patterns in astronomic pheromones) or technology (e.g., AR, CAD software) for solving
further tasks. Therefore, it seemed more meaningful to teach through an integrated STEAM
approach (Committee on Integrated STEM Education et al., 2014; Kelley & Knowles, 2016)
instead of creating learning and teaching settings for each domain separately.
In our instructional design, we used an integrated STEAM rather than a STEM approach which is
described as "an instructional approach, which integrates the teaching of science and
mathematics disciplines through the infusion of the practices of scientific inquiry, technological
33
18.05.2021Haas Ben
and engineering design, mathematical analysis, and 21st-century interdisciplinary themes and
skill" (Johnson, 2013, p. 367). In the integrated STEAM approach (Psycharis, 2018), the different
STEAM disciplines are taught jointly (B. Kim & Bastani, 2017), often with real-world problems
(Kelley & Knowles, 2016). Recent research findings included Arts in the STEM disciplines due its
positive impact on students’ motivation, engagement and effective disciplinary learning (Kang,
2019) when solving STEM tasks. As reported in research (Conradty & Bogner, 2018), arts within
STEAM cultivates creative thinking process and further supports students with new
accessibilities to STEAM. Arts in STEAM offered creative approaches to students and
connected mathematical concepts to architecture and a variety of designs (Lavicza et al., 2018).
Kelley and Knowles (2016)
highlighted the importance of connecting real-world problems and situations to different subjects.
According to Connor et al. (2015), STEAM approaches are inevitably linked to problem solving
and fruitful STEAM learning based on problem-based tasks (Quigley et al., 2017). Liao (2016)
discussed that innovative thinking positively impacts the overall learning process when the
STEAM disciplines are taught in a transdisciplinary space with the teaching lessons. Hence, with
an integrated STEAM approach, skills and knowledge can be taught through a holistic view to
prepare students in applying their skills in complex situations in which disciplines are
interconnected (Vasquez et al., 2013). Students should utilise these skills and knowledge from
different disciplines to solve STEAM tasks. According to Land (2013, p. 552), within a STEAM
approach, students “decompose a complex problem using convergent thinking and then apply
the corresponding solution to the real world”. By analysing a task and identifying STEAM
concepts, students are invited to transfer skills, mobilise content knowledge, and discover
myriad connections among STEAM disciplines. Based on these arguments and findings, we
decided to follow an integrated STEAM approach for our study. However, this integrated STEAM
approach in our instructional design was relatively new for us, thus we present the initial phases
of this ongoing process in this thesis (further details can be found in paper E and G).

We used the educational technology MathCityMap to apply an integrated STEAM approach

through outdoor trails. MathCityMap is a GPS supported educational technology allowing
teachers to design, in a guided structure (e.g., predefined criteria, proposed scaffoldings,
feasibility control of tasks) tasks in relation to students’ living environments. Further, teachers
can bundle these tasks into outdoor trails. With MathCityMap in outdoor trails involving
real-world places and situations (Cahyono & Ludwig, 2019; Jablonski & Ludwig, 2020; Ludwig &
Jesberg, 2015), tasks focused on mathematics and numerous interactions with real-world
objects or places. Due to the technology, students used digital functions, in addition to more
traditional analogue measuring instruments. Thus, in the example (figure 20), students needed
to identify the total height of all steps in front of their school building. To measure the height,
students could use a ruler, a measurement application on their smartphones, body parts (e.g.,
hand, elbow-length) or any other tool for their measurements. Students could estimate or
measure precisely and enter their answers. Within the task, they can find hints in MathCityMap.
Hence, several different possibilities of tasks and several possible extensions to the STEAM
disciplines (e.g., continuing possible steps with AR and transforming the architecture) can be
done with MathCityMap.
As a continuation of our previous instructional designs (automated tutoring system with
real-world information and 3D modelling with real-world objects), we focused on the instrumental
genesis theory to identify digital tool functions (e.g., AR, GPS, analogue or digital measurement
tools),

34
18.05.2021Haas Ben
throughout the different studies, as it allowed us
to observe and understand how students
manipulate digital tools and their features.
Thus, we performed a literature review on
mathematical modelling with real
world situations (Freudenthal, 1971; Greefrath &
Vorhölter, 2016; Lavicza et al., 2020; Selter &
Zannetin, 2018), outdoor learning (Jablonski &
Ludwig, 2020; Ludwig & Jesberg, 2015; Moffett,
2012), STEAM integrated approaches (see
previous paragraph), digital tools in outdoor
learning (Cahyono et al., 2020; Cai et al., 2019;
Ludwig & Jesberg, 2015)
Figure
18 STEAM task within MathCityMap and teachers in STEAM teaching
their instrumentalisation and instrumentation. It (Chittleborough, 2014; Cooke & Walker, 2015;
seemed essential to preserve this approach Michaluk et al., 2018) to
identify characteristics of outdoor tasks and trails that we wanted to design and investigate.

Based on this literature review and our preparation for designing outdoor tasks, we developed a
pedagogical framework (figure 19). The pedagogical framework contained two major
components, outdoor trails and STEAM integrated approaches, that we identified during the
literature review in accordance with the MathCityMap.

While MathCityMap was initially designed to be used for mathematics education (Ludwig et al.,
2021), we wanted to illustrate in our framework and instructional design that tasks with
MathCityMap can be related to mathematics and at least one STEAM discipline. For instance,
when students (figure 18) reconstruct a given monument with an additional geometric solid (e.g.,
insert a sphere in a hemispherical monument), student use technology (i.e., AR), mathematics
(i.e., solid properties), connects their knowledge to arts (i.e., monument) and modulate the solid
to fit the monument (i.e., engineering). This interdisciplinary approach within the tasks on real
 Figure 19 Pedagogical framework on STEAM outdoor tasks
35
18.05.2021Haas Ben
world problems allowed us to generate new knowledge and develop further process and problem
solving skills (Stock & Burton, 2011). A STEAM integrated task design could support the
development of new perspectives, content, and connections to STEAM disciplines (Gibbs, 2015).
Thus, the integrated STEAM approach could guide the instructional design of tasks.
Furthermore, students used their process skills (i.e., communication, representation, modelling,
problem solving) and problem-solving approaches (section 2.1) in real-world STEAM tasks. In
the pedagogical framework, we illustrated the importance of process skills and content mastery
in the tasks. In our study (paper E), we analysed several tasks and observed that within each
task, we could find at least two process skills and mathematical content skills.

In our study, pre-service teachers needed to be aware of different STEAM elements to prepare
hints, scaffoldings and suggest digital tools for students. Moreover, task design was inevitably
related to STEAM knowledge and skills of task creators (Michaluk et al., 2018). Tasks were
structured and conceived according to the findings of Vos (2015) by subject matter experts in
STEAM. Pre-service teachers showed to use educational technologies in different ways (e.g.,
integrating science task
elements, arts or only
referring to mathematics
tasks) and showed
different manipulations of
the
same technology (e.g.,
with more
AR or not, digital
measuring tools
or requesting additional analogue
measuring tools) in their task
designs. According to Slough and
Milam (2013), STEAM task design
should respect three principles
(i.e., making content accessible,
making thinking visible, helping students to learn from each
other). MathCityMap

Figure 20 Students solving tasks within

Thus, with real-world places and situations, using STEAM integrated approaches, we aimed to
create meaningful and engaging tasks (making content accessible). With process skills, problem
solving and mathematical modelling approaches, we created a theoretical framework to support
the creation of tasks (making thinking visible), and finally, tasks were designed that students
solve them in peer groups within the math trails (figure 20).

During this last study on real-world places and situations, we were able to observe different
mathematical modelling experiences. Compared to the first study on real-world information and
the second on real-world objects, we observed that in connecting activities directly to outdoor
learning with technology, students could experience a flow-like state (Csikszentmihalyi, 2014).
Students actively discussed STEAM disciplines, trialled iterative mathematical modelling cycles
and supported each other. Referring to the situation described by the national testing
(Sonnleitner et al., 2018), the use of outdoor tasks with MathCityMap, could provide access to
skill and knowledge to a broader public (free software, access to different devices) and present
a possible solution to reduce educational gap (see section 1.2.2) between the students. Hence,
outdoor tasks could be designed by professionals (teachers), and students could have access to
the tasks outside the schooltime and learn with their parents and peers. In our paper E and G,
we described how we started to train pre-service teachers with outdoor task designs and
outlined results from this study.
36
18.05.2021Haas Ben

2.4. Summary

In the theoretical framework section, we presented an overview of theories, research and

discussions, which guided our studies or elaborated during the research process. We did not
perform an extensive analysis of the different themes. However, in the papers, the reader can
find more in-depth discussions and reflections. Our research on real-world information with a
strong focus on process skills and problem-solving approaches using an automated tutoring
system was a first step in identifying new learning opportunities for all students (independent of
socio
economic, linguistic or performances background). The developed framework (figure 8) showed
that process skills and problem solving were part of the mathematical modelling approach
(section 2.2), which became during our second study on real-world objects important to design
tasks and understanding learning steps of students connecting mathematical concepts to
real-world objects. Finally, while designing tasks outdoor in our last study, we extended the
framework to STEAM. As mentioned before, outdoor tasks were related to several disciplines of
STEAM. While elaborating and adapting the theoretical framework through the different studies
(i.e., different technologies tasks and participants) on three different real-world connections (i.e.,
information, objects and situations/places), we aimed to contribute to our research questions.
Thus, we wanted to identify technology-enhanced tasks for each real-world connection, which
requested different technology and task designs. However, the core framework (figure 8) could
be applied in retrospective to all the different tasks in our studies. We choose to follow a
design-based research methodology (Feng Wang & Hannafin, 2005) to a holistic approach with
technology on different real-world connections with different stakeholders and various
adaptations throughout studies. After every study, we used the data to adapt our research and
respond to our research questions. Moreover, we choose various research methods adapted to
the given possibilities of research in the different studies. The upcoming section will describe
and justify our choices for used methodology and the different research methods in the studies.
 37
18.05.2021Haas Ben

3. Methodology and methods

“There's a big difference between knowing the name of something and

knowing something.” Richard P. Feynmann

We investigated several technology-enhanced mathematical modelling tasks based on real-world

information starting from wordings to real-world objects, then to real-world places and situations.
As outlined in the previous sections, we used different technologies, connections to the real
world and adapted these interventions and task designs based on previous findings,
observations or discussions with different stakeholders (i.e., parents, students, teachers). Our
research was driven by an iterative process, where consultation with different stakeholders in
elementary schools provided us with data (e.g., task-based interviews, semi-structured
interviews, questionnaires) from experiences and suggestions for needed improvements.
Moreover, we tried to understand how students could connect mathematics with their daily lives
and living environments. Research on these three studies lasted five years and underwent
numerous and necessary adaptation (e.g., different real-world accesses, technology
developments, different stakeholders). Therefore, we chose methodologies that were supporting
these iterative processes and adaptations of our research designs.

According to our research needs, we employed design-based research (DBR) (Feng Wang &
Hannafin, 2005) with different designs (e.g., mixed-methods triangulation) and elements for each
study. We completed iterative design cycles (i.e., interactive, iterative and flexible characteristic),
where we performed user experience testing, exchanges with experts, varied collaborators, and
adapted tasks, learning and teaching settings. However, analysing our studies, design-based
research methodology was supplemented by other pragmatic research elements. Therefore, we
sought to use methodologies (e.g., pre- and post-test research designs with our automated
tutoring system or mixed-methods design for our outdoor tasks) and methods (e.g., task-based
interviews, cluster analysis, semi-structured interviews) for the different studies. We chose
designs and methods which best suited our research objects in the study and took a pragmatic
stance on our research approach. Thus, throughout the studies, we varied quantitative and
qualitative research or used mixed methods to better understand situations. This section will
present our overall design-based research methodology, the different designs within studies,
research contexts, data collection approaches, utilised methods, and data analyses. We will
describe how we performed our research over the past five years to obtain findings and
contributions to our research questions.

3.1. Design-based research

Due to the dynamics of our research purpose (e.g., designing technology-enhanced tasks for
teachers, parents and students) and the rapid evolution of technologies, we adapted
design-based research for our overall methodology. Design-based research can be qualified as
an iterative analysis and adaptation of interventions, tasks, methods and evaluations throughout
the research. According to Collins et al. (2004, p. 18):

This approach of progressive refinement in design involves putting a first

version of a design into the world to see how it works. Then, the design is
constantly revised based on experience, until all the bugs are worked.
38
18.05.2021Haas Ben
This methodology allows researchers to continually evaluate and adjust settings (e.g., tasks,
interventions, methods) with social and contextual variables interventions (Dede, 2004). Amiel
and Reeves (2008) described different stages in design-based research (figure 21).

Figure 21 DBR according to the model of Amiel and Reeves (2008)

Our research analysed connections of mathematics teaching and learning to students' living
environments (i.e., real-life). We did this jointly with practitioners and developed our first
technology-enhanced tasks, which were tested (e.g., human-computer interactions, think-aloud
and experimental research with a pre- and post-test design) and adapted. At the end of the first
study, we evaluated these environments and decided to refine some problems (e.g., connect to
real-world without language in wordings with objects). Again, we engaged in discussions with
practitioners and developed new tasks. After each study, we repeated this process.

Wang and Hannafin (2005) wrote that a design-based methodology has several characteristics
(i.e., pragmatic; grounded; interactive, iterative and flexible; integrative; contextual):

Pragmatic: The focus of the research is to inquire pragmatical issues (e.g., learning
tasks or teaching settings) with theory. Both theory and practice are core elements of this
 methodology and influence each other.

Grounded: The research is theory-driven and grounded within existing theories and
designs, embedded in real-world contexts.

Interactive, iterative and flexible: Researchers are a part of the design process and
work closely with the study participants in iterative cycles of analysis, design,
implementation and redesigns.

Integrative: Researchers use mixed research methods during different phases as new
needs and issues arise during the research.

Contextual: The process design and findings change during the research and are
documented. Moreover, results are linked to process and settings, with different content
and depth of design principles.

39
18.05.2021Haas Ben
As mentioned before, with these characteristics in mind, we designed and performed our
research. Initially, we chose a practical issue (i.e., pragmatic characteristic) in teaching and
learning mathematical modelling, tasks connecting mathematics to real-world information with
technology. Our first design aimed to understand and apply mathematical concepts to arithmetic
word problems, commonly used in schools to teach real-world connected mathematics. During
the process, we investigated tasks related to real-world objects, places and situations.

We investigated the issues in theory (i.e., grounded characteristic) through literature reviews,
exchanges with experts and practitioners. As mentioned in the theoretical framework, we
focused firstly on process skills and problem-solving approach and continued with a
mathematical modelling approach during the process design, adapted task designs,
technology-enhanced learning, and STEAM integrated approach in the final studies (i.e.,
grounded characteristic).

We performed iterative design cycles (i.e., interactive, iterative and flexible characteristic), where
we completed user experience testing, exchange with experts, varied collaborators, and adapt
tasks, learning, and teaching settings. Findings from the design-based research vary in depth
and forms; however, they are connected to the research study’s general theory (i.e., contextual
characteristic).

With mixed-methods triangulation (Howe, 2012), we aimed to obtain a more in-depth

understanding through different perspectives. Throughout the different studies, we used various
research methods, adapted to each new evolution in the design process to improve validity for
our research (i.e., integrative characteristic). Thus, our research followed a flow of research
evolution with this design and encountered weaknesses and strengths of tasks, learning, and
teaching settings. In addition, we used a pragmatic research approach to adapt our designs to
given contexts, opportunities and research questions. For example, we could not replicate the
same large-scale experimentation for real-world object connected tasks, as we did for our first
study on the automated tutoring system, due to the limitation of technology accessibility for
participants and the nature of tasks. Moreover, during manipulations with real-world objects, it
was essential for researchers to observe actions and collect the data in-person to understand
students’ actual work (e.g., task-based interviews). Each technology offered different
possibilities of usage and necessitated different research components. Our understanding of
mathematical modelling and how to transition from in-class to outdoor mathematical modelling
was driven by our research results, following a particular design-based research methodology.
The upcoming section will describe the employed research designs, methods, and contexts in
the three different studies (i.e.,
real-world information, real-world objects and real-world places and situations). 3.2.

Research methods and contexts from our different studies

As mentioned before, we performed three studies with three different contexts, technologies and
real-world connections. Each study was done within design-based research. We used various
research methods, each time adapted to technology, context and given data possibilities. Thus,
we used quantitative, qualitative or mixed-methods in data collections to get a deeper
understanding of the tasks and stakeholders needs. In this part, we will present the contexts,
different methods and participants for each study briefly. Since this is a cumulative thesis, the

40
18.05.2021Haas Ben
reader will find more in-depth descriptions in the papers that we will indicate in the upcoming
descriptions.
3.2.1. Real-world information with an automated tutoring system
Our first study investigated how an automated tutoring system could support students, with
different language and socioeconomic backgrounds, in applying mathematical concepts to real
world information. The study was carried out in class, and the real-world information was
presented through arithmetic word problems.
3.2.1.1. Context of the study with MathemaTIC
In 2016, the Ministry of Education in Luxembourg developed,
jointly with the Canadian company Vretta, an educational
technology software for mathematics learning in elementary
schools called MathemaTIC. A multidisciplinary team (i.e.,
teachers, subject matter experts, didactics experts, instructional
designers, and researchers) created a learning module inside
MathemaTIC with an automated tutoring system to foster process
skills and problem-solving approaches in arithmetic words
problems. The objective of this learning module was to create new
learning possibilities and to address low performance. We aimed
Figure 22 Student working with MathemaTIC parents.
for a learning setting in class or at home,
without additional guidance from a teacher or
3.2.1.2. Participants of the study with MathemaTIC
The content of the module on arithmetic word problems was based on the grade 3 national
curriculum. Thus, in this study, we focused on grade 3 elementary school students (age 8 to 10).
48 randomly selected elementary school classes with 667 students in Luxembourg participated
in our quantitative study.
3.2.1.3. Research methods of the study with MathemaTIC
To evaluate possible effects on learning process skills and problem-solving approach with
MathemaTIC, we chose an experimental research design with randomised pre-test–post-test
control group design (Cohen et al., 2018). The experimental research can be used to support the
generalisation of technology-enhanced tasks, particularly if the targeted population is selected
representatively (Oncu & Cakir, 2011). Thus, we wanted students to work on MathemaTIC in a
one-to-one setting compared to their regular courses with teachers. By changing this variable
(i.e.,
teacher replacement by technology), we aimed to observe cause and effect relationships
(Fraenkel et al., 2011) and validate designed tasks through performance evolution comparisons
between groups. An explicit limitation of this design was that we could not eliminate all possible
effects which might lead to performance changes (e.g., teacher motivation).

To establish the groups, we used the variables age, gender, and performance in mathematics
from the national testing base ÉpStan to identify matched pairs and assigned classes to
experimental and control groups. We chose these variables to connect our findings to the results
of ÉpStan (Sonnleitner et al., 2018) and created representative groups of school populations.
Participation in the study was open to all schools, and we received a favourable participation
agreement from 667 students. Since we did not split the enrolled classes evenly, the number of
students in the control and experimental groups differed. Moreover, we also expected drop-outs
by the end of the study (e.g., due to technical issues or students on sick leave). Hence, we
41
18.05.2021Haas Ben
allocated 278 students to the control group, working using paper-and-pencil, and 389 students to
the experimental group, working with the arithmetic word problems module in MathemaTIC. To
evaluate possible effects of task designs within MathemaTIC (e.g., scaffoldings, digital tools and
their functions) compared to teacher-centred teaching, we procured a control group for all tasks
(wordings) used in MathemaTIC with paper-pencil design.

At the end of the study, 34 classes with 472 (8 with a missing post-test) students remained: 246
(2) students in the experimental group and 226 (6) students in the control group. The drop-out
was due to local technical errors (e.g., low WiFi signal or non-working hardware) while using
MathemaTIC or simply because of a missing post-test for the whole class. As mentioned before,
we chose to allow a higher participation number to obtain sufficient data by the end of the study.
However, both groups were still representative (e.g., substantial test results to compare
performances), and we observed no factors that might have made the experiment not internally
valid (e.g., maturation effect, selection bias, experimenter bias) (Leedy & Ormrod, 2010).

We assessed arithmetic word problem-solving using 16 items of varying type (i.e., combination,
transformation, and comparison), complexity (i.e., one or two operations necessary for solving
the problem) and response format (i.e., multiple-choice, open-response, and circuiting). Finally,
we performed a three-way mixed ANOVA with group assignment and language spoken at home
(German & Luxembourgish vs Other) as two between-subjects factors and time as
within-subjects factor. Further details of methodologies and methods can be found in paper A
and B.

The experimental research approach we used in the study on real-world information with
automated tutoring allowed us to evaluate our designed tasks and connect findings to our initial
pragmatic problem. However, at the end of the experimentation, we observed that although we
had positive effects, tasks were still connected to students' language proficiency. Therefore, we
continued the research with real-world objects within our design-based research without
wordings. We changed technologies and developed new tasks. Moreover, we learned from the
first study that a more qualitative approach could procure a better understanding of task solving
and eliminate more side effects in the research (e.g., why a student gained in motivation or
understanding questioning the student). We will describe these modifications to our
methodology and methods for the second study with real-world objects in the next part.
 3.2.2. Real-world objects with digital and physical modelling
In our second study, we investigated how AR, 3D and physical modelling with 3D printing could
support students in learning mathematics with real-world objects. We performed two experiments
in two different contexts. Our first experiment happened during regular school time in special
needs education in elementary school. This experiment was strongly related to our work as a
special needs teacher and in response to the pragmatic problem of supporting students with
learning difficulties in mathematics. The second experiment happened in remote teaching during
the first lockdown in the COVID-19 pandemic, as there was no possibility to continue
mathematical modelling in-class. Furthermore, we wanted to explore how parents who were
assisting their children at home during this period could perform mathematical modelling with
their children. We collected data to understand what kind of support parents need and how
these tasks affect students’ learning motivation. In both experiments, we used structured
task-based interviews to observe mathematical behaviour and grounded theory approaches.
42
18.05.2021Haas Ben
3.2.2.1. Context of the experimentation in special needs education
In special needs education in elementary schools,
specialised
teachers work with students with learning difficulties in
mathematics and other disciplines. Moreover, students with
special
needs have difficulties in spatial orientation, connecting
learned
skills and knowledge to the real world. In many cases,
which we
discussed with practitioners while formulating our
pragmatic
problems, these students mainly work on reduced complexity and
amounts of tasks. Instead of making tasks easier for these
students, however, we wanted to use technologies to support
students in solving age-appropriate problems Figure 23 Student creating digital modelling of a cube
puzzle
and developing new strategies in connecting to
the real world. After an extensive

literature review and idea exchanges with researchers and practitioners, we identified
technologies such as AR, 3D printing and physical modelling as promising directions (section
2.3.2). Thus, we used the dynamic geometry software GeoGebra 3D Graphing Calculator with an
augmented reality function and 3D printing to support these students in learning mathematics in
this setting. Moreover, to improve difficulties in spatial orientation and connecting concepts to the
real world (e.g., shapes and forms), we focused on developing skills that would allow us to
connect mathematics to real-world objects.
3.2.2.2. Participants of the study in special needs education
We received the authorisation for a special needs class in a local school to carry out our
research, in which two students with mathematical learning difficulties were enrolled. Since the
development of technology-enhanced tasks in special needs was not a usual request for
research at that moment, we decided to work with these two students initially and experiment
with new tasks further in upcoming research (which could not be done so far due to restrictions
of COVID-19) with a broader population. Thus, we worked with two special needs education
students from elementary school, one female aged 10 and one male aged 11. Both students,
diagnosed with a mathematical learning disorder, with significant difficulties in the reasoning and
visual-spatial sub-types, visited a regular elementary school class and participated in special
needs courses weekly.
3.2.2.3. Research methods of the study in special needs education In this experiment, we
chose a qualitative research methodology with a case study (Stake, 1995) and structured
task-based interviews (Goldin, 2000) to investigate the use of the developed tasks with special
needs students and to identify developments of mathematical skills through an analysis of
different aspects of performance in process and content skills, technology use and perception)
from two students with difficulties in learning mathematics within special needs courses. With
this methodology, we observed how students perceived technology-enhanced tasks (e.g.,
manipulation, frustration, motivation), the interplay among these tasks and contextual variables
(e.g., teacher, room, time or technical devices) and a possible generalisation of the tasks.

Hence, we performed task-based interviews to collect data on the use of AR, digital and physical
modelling tasks with special needs students (Goldin, 2000). Task-based interviews are often
used in mathematics education to reveal students' mathematical thinking and task management
(Koichu & Harel, 2007). Therefore, we estimated that this clinical approach could be
appropriated to collect data on the designed tasks, and students developed new skills with
technologies (e.g., AR, 3D modelling). Furthermore, task-based interviews, 5 per student (i.e.,
one per task), allowed us certain flexibility on inquiry venues (Goldin, 1997) during interviews
and interaction with the task
43
18.05.2021Haas Ben
environment (Maher & Sigley, 2014). The questions during task-based interviews were designed
to explore students’ skills (e.g., mathematical modelling in mathematics and content skills). Thus,
the interviewer asked students about planning, modelling, representation, and proving
processes. The questions were similar for each task (e.g., How could you find out? What could
you try next? How did you get this answer?). According to Ericsson and Simon (1984), these
questions are likely to engage students in self-observation, like the think-aloud method (Cotton
& Gresty, 2006) or follow interviewer-oriented questions. Thus, we were able to capture user
experiences (e.g., how they manipulate the tools, their expectations, or the moments when they
wanted to use the tools) during the tasks. We collected data from different interviews and
performed data analysis to identify if there were changes in strategies or behaviours. Further
details can be found in paper C.

After this experience and experimentation with tasks in special needs education, we aimed a
second experience on real-world objects with elementary school students. Firstly, however, we
needed to adapt our research to the required sanitary restrictions imposed by the COVID-19
outbreak, which required us to adopt remote learning research, which will be described in the
upcoming section.
3.2.2.4. Context of the experimentation in remote teaching
In the second experiment on real-world objects, we
performed digital and physical modelling in a remote
challenge with the Ministry of Education's support in
Luxembourg. Hence, we organised remote online
teaching using AR, 3D and physical modelling with
3D printing. The task consisted of creating a cultural
artefact (i.e., Easter egg cup) by elementary school
students, with the assistance of their parents, with
digital mathematical modelling in GeoGebra 3D
Graphing Calculator or Autodesk® Tinkercad®. The
final design was then sent to us and printed. To
 support parents and the students, we offered
several tutorial videos to explain the digital
mathematical modelling software's functions and a
dedicated space on the web page “Schoul
Figure 24 3D Prints of students from the remote by mail or Microsoft Teams (for which each
teaching
student in
Doheem”. Moreover, parents could reach out
Luxembourg has a free license) for additional support. The communications platforms we used
allowed us to collect data on students’ performances and motivation on parents’ in their teaching
assistants role. We chose to use a grounded theory approach (J. Corbin & Strauss, 1990) for this
research to collect a variety of data (e.g., open-ended questions from a questionnaire, chats,
calls, field observations, and e-mails) and develop themes describing how both groups
experienced working with these tasks.
3.2.2.5. Participants of the experimentation in remote teaching
The experimentation was open to every student in elementary schools, and we had 44 students
aged 5 to 13 who participated with their parents in the challenge. We advertised our remote
challenge on several social media platforms, websites, schools and even promoted the challenge
in a radio show. Every participant who wanted to join was accepted, and we were able to get
participants from every age category (section 1.2.1). In addition to the students’ tasks
experiences,
44
18.05.2021Haas Ben
we focused in this experimentation on the parents and their teachings (e.g., scaffoldings) as they
took the role of teacher assistants in the remote challenge. Parents needed to fill a new role
during remote teaching, supporting and guiding their children without direct professional support.
Thus, we estimated that it would be important to use this situation and collect data for further
remote teaching or create new opportunities for parents when school would reopen.
3.2.2.6. Research method of the experimentation in remote teaching To
research both students’ and parents’ behaviours in our remote teaching challenge and develop
new theoretical input for this new situation (e.g., first lockdown and remote teaching imposed for
most of the parents), we also used a grounded theory approach (J. M. Corbin & Strauss, 2015; J.
Corbin & Strauss, 1990). Moreover, we did not want to validate any hypothesis as there were too
many unknown effects which we could not take into account. With a grounded theory approach,
we identified themes, and based on them, we created a framework for parent-assisted teaching
in remote mathematical modelling with real-world objects. Furthermore, for the wide variety of
data sources (e.g., semi-structured interviews, open-ended questions, field observations, chats,
e-mails), which needed to be coded and structured, then an iterative coding process (i.e, open,
axial and selective) of grounded theory was used. Thus, we collected data on the use of AR,
digital and physical modelling in remote teaching through parents’ perspectives in various ways.
We conducted 4 semi-structured interviews with parents, 3 semi-structured interviews with
elementary school teachers, one semi-structured interview with a school leader and 8
semi-structured interviews with students from age 7 to 13. Besides, we received data from open
questions (e.g., How did you as parents assist your child? Which parts in the mathematical
modelling process were difficult for you and your child? How was the motivation of your child
during the modelling?) in an online questionnaire with the 13 parents at the end of the remote
teaching. Adding to this, we analysed e-mails, conversations and messages on Microsoft Teams
and WhatsApp groups that we got from parents. The different data sources allowed us to obtain
substantial information without being physically present or observing the entire process, as we
could have done in a standard setting (e.g., in-class observation).

From open coding, we were able to develop our initial categories (e.g., parents’ interactions,
parents’ requests or students’ tasks performances). Still, these categories were too vague and
yet insufficient to be used in our framework. Therefore, we went through two additional
re-codings. This process of constant comparative data analysis and re-coding (J. M. Corbin &
Strauss, 2015)
allowed us to identify 6 themes and a conceptual framework on parents' perspectives. Thus,
rereading of transcripts, questionnaire responses, and coding highlighted a set of different
themes that were dominating the data. Further details on the themes and the connections to
data can be found in paper C. At the end of this study, we collected data on the
technology-enhanced actions from students (and parents) with real-world objects. The research
designs and methods offered many details on recalibrating tasks, and we were able to discover
new manipulations. However, our findings were somewhat limited since we could not investigate
how the school population at large (e.g., representative in gender, ages, socioeconomic status)
would experience the technology-enhanced real-world object modelling. We decided to rethink
our pragmatic problem setting and combine the outcomes of both previous studies. With the
educational technology MathCityMap (section 2.3.3), we designed technology-enhanced
mathematical modelling with outdoor tasks and real-world situations and places. Learning from
both previous experiences, we chose a mixed-methods triangulation design, where we
combined quantitative and qualitative

45
18.05.2021Haas Ben
research approaches. In the upcoming section, we will describe the research design, methods
and context of our third study, with real-world places and situations.
3.2.3. Real-world places and situations with an integrated STEAM approach in
outdoor trails
In the third and final study, we investigated technology-enhanced learning and teaching with real
world places and situations within a STEAM integrated approach. The study happened in remote
teaching in our university course with pre-service elementary school teachers during the COVID
19 pandemic. So far, in both previous studies, we collected data on students and parents with
technology-enhanced mathematical modelling. We wanted to gain insights into how teachers
would develop their teaching and tasks. In this case, we observed pre-service teachers.
Teachers are highly important to develop new learning opportunities in mathematics education,
and thus, we needed to identify existing teaching patterns, needed scaffoldings and possible
task designs. We used a mixed-method triangulation (Howe, 2012) to complete quantitative and
qualitative data to better understand task designs and needs from stakeholders.
3.2.3.1. Context of the study on outdoor trails
Amidst the COVID-19 pandemic in 2020, on-campus learning and teaching at the University of
Luxembourg was suspended and changed to schooling at home (Kreis et al., 2020). Considering
the difficulties (e.g., proctoring) of organising an exam in schooling at home, we adopted an
assessment outside the classroom, also using MathCityMap, initially planned for the upcoming
academic year. Further, in a collaborative approach, the pre-service teachers created additional
online resources for elementary schools (i.e., created outdoor trails will be made available to
others). Designed tasks and trails relied on the integrated STEAM teaching approach,
mathematical modelling approach, process skills and STEAM content skills (e.g., areas,
volumes, number theory) of the national curriculum (MENFP, 2011). During the trail creations,
we performed close monitoring through a series of tutorials. Finally, students performed a
self-evaluation and three peer-reviews of the submitted outdoor trails. Similarly to the remote
challenge that we described earlier in this thesis, we needed to collect data from different
sources to understand how pre-service teachers experience task designs, and what scaffoldings
and what skills needed to be developed in upcoming courses. Since we experimented with
teaching technology-enhanced task design in remote areas for the first time, we wanted
participants to participate in their evaluations. Thus, we estimated that while evaluating the trails
of other students, they could consult with different task designs and STEAM integrated
approaches.
3.2.3.2. Participants of the study on outdoor trails
We collected data from first-year pre-service teachers enrolled in the mathematics education
course in the Bachelor's degree at the University of Luxembourg. Out of 100 pre-service
teachers, we received authorisation from 35 female and 14 male teachers. The pre-service
teachers were in their second term and received teachings on basic mathematics educational
principles regarding numbers and operations, geometry, and technology usage to become
elementary school teachers.
3.2.3.3. Research methods of the study on outdoor trails
The study started in the second half of the semester after the schooling at home started and was
thus entirely conducted remotely. The pre-service teachers created 21 outdoor trails with 10 to
18 tasks per trail (259 tasks in total) and received remote guidance from the lectures. We used
mixed methods triangulation (Creswell & Plano Clark, 2011) with quantitative and qualitative
instruments to identify teaching patterns through cluster analysis (i.e., quantitative) based on
coded data (i.e.,
46
18.05.2021Haas Ben
qualitative). Furthermore, we utilised the coded data to describe needs from this first experience.
We oriented our qualitative coding on a designed pedagogical framework (section 2.3.3) to
identify STEAM integrated teaching characteristics with similar grounded theory approaches (J.
Corbin & Strauss, 2014) as earlier. We coded our data through several coding processes, trails,
and tasks based on the pedagogical framework elements (e.g., outdoor trails, integrated STEAM
approach, technologies used, content and process skills). The advantage of a grounded theory
approach was to structure the data from different sources and identify characteristics of tasks
connected to real-world places and situations. However, we had no apparent reference for such
tasks and needed to create theoretical support for future task designs. Thus, we identified codes
that would most likely describe STEAM integrated teaching in outdoor trails (e.g., characteristics
of sciences in the tasks, mathematical skills or technology use). Furthermore, we identified
codes for cluster analysis throughout iterative coding (i.e., open, axial and selective) and
obtained findings on necessary improvements in our teaching settings (e.g., needed scaffoldings
in sciences didactics or peer interactions) during the coding process. Each task was marked by
the dominant content skill (e.g., segment measurement).

Following the findings of Antonenko, Toy, and Niederhauser (2012), we chose cluster analysis
(CA) with variables from the different outdoor trails to identify groups in teaching patterns.
Further, we decided to use CA to identify different teaching patterns and develop differentiated,
tailored support for each cluster in the upcoming courses. Among the different CA methods, we
used a hierarchical cluster analysis (HCA) based on integrating the fields Science, Technology,
Engineering, Arts, and Mathematics (STEAM) in the tasks of the outdoor trails. Before starting
cluster analysis, we converted the five field variables to percentages to get comparable
continuous data. Then, we did the statistical analysis using R. Further details on CA and codes
from grounded theory can be found in paper E, F, G and H.

3.3. Summary of the methodology and methods

We used different designs and methods to adapt to contexts within the design-based research on
technology-enhanced tasks with real-world accesses (real-world information, real-world objects,
real-world places and situations). Further, we used different designs and methods adapted to
studies’ contexts and pragmatic research objects (e.g., different technologies or stakeholders).
The choice of designs and methods was embedded in a long discussion with other researchers
and intensive literature reviews. Moreover, after each study, we questioned the pragmatic
problem (i.e., connection to real-world), consulted practitioners and adapted tasks and
technologies. Finally, we followed a pragmatic approach for each task design and technology
and chose quantitative, qualitative or mixed-methods designs. In retrospection, we would
probably have used more of the mixed-method in all three studies since this offered us sufficient
information to understand the manipulation of tasks and still extract results from a representative
group of the targeted population.

Nonetheless, we identified different connections to the real world, technologies, tasks, and
stakeholders needed for our design-based research. Furthermore, the variety of designs allowed
us to work on different data sets with a variety of approaches, and thus, we were able to identify
and report a complex system when connecting mathematics with technology-enhanced tasks to
the real world. Since this is a cumulative thesis, readers will find in the upcoming section
47
18.05.2021Haas Ben
(pre-)prints of the publications, which will give more insights into the findings, the tasks and the
data analysis. We will present each paper with an introduction and summary section and extract
their main ideas.
 48
18.05.2021Haas Ben

4. Publications

“The important thing is to not stop questioning.

Curiosity has its own reason for existing.” Albert Einstein

We presented our theoretical foundations, the methodology and methods in the previous
sections. Next, readers will find the publications (partly pre-prints) submitted during this PhD
project. We selected two publications for each of the three study periods highlighting our
contributions to the scientific community. Furthermore, readers will find short introductions and
summaries for each publication. Publications will be presented in this section in the following
order:

Real-world information with an automated tutoring system

This section will present two publications related to our study on real-world
information with the automated tutoring system within the educational
software MathemaTIC. The first publication was published during the 10th
ERME Topic Conference (ETC10) on Mathematics Education in the Digital
Age (MEDA) in 2020. We must note that the second publication was
already accepted for publication by reviewers for the Mathematics in
Computer Science journal as a special issue for the Digital Tools in
Figure 25 Presentation of the characters within guest editor of MCS had to choose another
MathemaTIC journal for publishing the paper and selected
Mathematics Education Conference the
(CADGME). However, the publication of the
CADGME special issue was cancelled, and the
International Journal of Mathematical Education in Science and Technology (IJMEST). Currently,
this paper is under an additional review cycle of IJMEST.

• Haas, B., Kreis, Y., & Lavicza, Z. (2020). Fostering process skills with the educational
technology software MathemaTIC in elementary schools. In A. Donevska-Todorova, E.
Faggiano, J. Trgalova, Z. Lavicza, R. Weinhandl, A. Clark-Wilson, & H.-G. Weigand (Eds.),
Proceedings of the 10th ERME TOPIC CONFERENCE (ETC10) on Mathematics Education
in the Digital Age (MEDA) (pp. 199–206). Johannes Kepler University. (Paper A)
• Haas, B., Kreis, Y., Lavicza, Z. & Sonnleitner P. (In Review). What works: Arithmetic word
problems within the educational learning environment MathemaTIC v.2017. International
Journal of Mathematical Education in Science and Technology. (Paper B)

Real-world object with digital and physical modelling

This section will present two AR, 3D and physical modelling publications with
3D printing on real-world objects. Each publication reports on a different
experiment. While the first publication explores technology-enhanced tasks
with special-needs students, the second will describe remote teaching
challenges on mathematical modelling, with a strong focus on parents and
their role during the lockdown due to COVID-19. The first publication was
accepted for publication in a special issue for the CADGME 2020 online
Figure 26 Craftbot 3D Printer Technology in Mathematics
conference in the International Journal for

49
18.05.2021Haas Ben
Education (IJTME). We submitted the second publication to the journal Educational Studies in
Mathematics (ESM), and it is still in the review process.

• Haas, B., Kreis, Y., & Lavicza, Z. (In Press). Case study on augmented reality, digital and
 physical modelling with mathematical learning disabilities students in an elementary school in
Luxemburg. The International Journal for Technology in Mathematics Education. (Paper C)
• Haas, B., Kreis, Y., & Lavicza, Z. (In Review). Parent's Perspectives: The Use of Augmented
Reality, Digital and Physical Modelling in Remote Teaching in Elementary School.
Educational Studies in Mathematics. (Paper D)

Real-world places and situations with an integrated STEAM approach in outdoor trails

This section will present two publications related to our study on

real-world places and situations within a STEAM integrated
approach in outdoor trails. The first paper was accepted for
publication in a special issue on higher education and STEM related
teaching in the journal Educational technology and Society (ESM).
The second paper was published in the 2020 BSRLM online
conference proceedings. The first publication reports on task
publication on reflection to connect
Figure 27 Student measuring the height of a tree
mathematics with the real world in general.
designs and teaching patterns with real-world
places and situations, and the second

• Haas, B., Kreis, Y., & Lavicza, Z. (In Press). Integrated STEAM Approach in Outdoor Trails
with Elementary School Pre-service Teachers in Luxemburg. Journal of Educational
Technology & Society. (Paper E)
• Haas, B., Kreis, Y., & Lavicza, Z. (2020). Connecting the real world to mathematical models in
elementary schools in Luxemburg. In R. Marks (Ed.), Proceedings of the British Society for
Research into Learning Mathematics: Vol. 40 (2) (pp. 1–6). British Society for Research into
Learning Mathematics. (Paper F)

4.1. Real-world information with an automated tutoring system

4.1.1. Fostering process skills with the educational technology software
MathemaTIC in elementary schools
Reference:
Haas, B., Kreis, Y., & Lavicza, Z. (2020). Fostering process skills with the educational technology
software MathemaTIC in elementary schools. In A. Donevska-Todorova, E. Faggiano, J. Trgalova,
Z. Lavicza, R. Weinhandl, A. Clark-Wilson, & H.-G. Weigand (Eds.), Proceedings of the 10th ERME
TOPIC CONFERENCE (ETC10) on Mathematics Education in the Digital Age (MEDA) (pp. 199–
206). Johannes Kepler University. (Paper A)
4.1.1.1. Short description of the MEDA paper
In the publication entitled “Fostering process skills with the educational technology software
MathemaTIC in elementary schools,” we reported on the initial findings from the experimental
research design with a randomised pretest-posttest control group on the use of the automated
tutoring system MathemaTIC with elementary school students (grade 3). This publication will
outline the design and results of this part of the study.

50
18.05.2021Haas Ben

Fostering process skills with the educational technology software

MathemaTIC in elementary schools
Ben Haas1, Yves Kreis² and Zsolt Lavicza1
1
Johannes Kepler University, Linz School of Education, Austria,
 ben.haas@outlook.com, ²University of Luxemburg, Luxemburg
This study reports the use of automated tutoring and scaffolding implemented in the
module “arithmetic word problem” in the educational technology software
MathemaTIC in grade 3 (age 8 to 10). We examined 246 students with access to
MathemaTIC and receiving tutoring and scaffolding through a one-to-one learning
setting with this technology. The control group (n=226) had access to the same
learning tasks and worked with paper-and-pencil without MathemaTIC but with their
teachers. Results showed that the experimental group finished with higher outcome
scores than the control group. This paper will outline the study and attempts to explain
these results.
Keywords: educational technology, process skills, elementary school, mathematics,
problem-solving

INTRODUCTION
Teaching arithmetic word problems in grade 3 (age 8 to 10) is reported by the teachers
in Luxemburg, as one of the most challenging topics in mathematics. Teachers in our
study suggested that students struggle in class to solve arithmetic word problems, due
to a lack of comprehension in reading, understanding of the wording and identifying
the arithmetic operations to execute. In these tasks, students required process skills
such as arguing, communicating, representing, and problem-solving (Selter &
Zannetin, 2018). Moreover, based on the result of the national school monitoring
EpStan in mathematics and language, students with low reading skills are also those
who are low performing in mathematics (Sonnleitner et al., 2018). Similar to the
findings of LeBlanc and Weber-Russel (1996), there is a connection between well
developed skills in reading and understanding of the mathematics course language and
mastering process skills. Therefore, many students need continuous assistants from a
teacher while learning to solve arithmetic word problems. In class, however, the group
of students is heterogeneous, and a close follow-up is challenging to realize.
In 2016, the Ministry of Education in Luxemburg developed, jointly with the Canadian
company Vretta, an educational technology software for mathematics learning in
elementary schools called MathemaTIC. A multidisciplinary team created a module
inside MathemaTIC with an automated tutoring system to foster process skills in
arithmetic word problems in grade 3 in order to create new learning possibilities and
to address the low performances. The instructional design of the module aimed for one
to-one learning in class and at home for students without additional guidance from
teachers or parents. We carried out a quantitative study to obtain findings on the use of

51
18.05.2021Haas Ben

this arithmetic word problem module by addressing the following two research
questions:
RQ1: Are students who learn process skills in arithmetic word problems with
MathemaTIC likely to improve at the same degree compared to a traditional
paper-and-pencil setting with the guidance of the teachers?
 RQ2: What are the limitations and opportunities of a one-to-one setting with
MathemaTIC?
Hence, we will present the design of the automated tutoring system and some results
of the quantitative study in which we examined 246 students with access to
MathemaTIC in a one-to-one learning setting without their teachers compared to a
control group (n=226), that had access to the same learning items using paper-and
pencil, but worked with their teachers.

THEORETICAL FRAMEWORK

Figure 1: Main view of the module “arithmetic word problem”

The structure of the automated tutoring in the arithmetic word problem module was
based on the Competence-Learning-Intervention-Assessment model by De Corte et al.
(2004) and the Four-Component Instructional Design (4C-ID) model by van
Merriënboer& Kester (2005) for learning complex skills. Both models suggested that
students should learn through guided learning tasks and then apply the skills in tasks
that are gaining in complexity and lowering in guidance. Furthermore, students should
develop a cognitive structure applicable for these complex skills with meta-tools
(Trouche, 2004) in the new tasks. Hence, in the module arithmetic word problems
(compare figure 1), students started working on guided learning tasks (blue), followed
52
18.05.2021Haas Ben

by semi-guided tasks (red) and finally complex tasks without guidance (black). The
scaffolding system, based on the multimedia learning theory (Mayer, 2005), consisted
in listening to the wordings, interacting with the images and arithmetic operations, and
self-regulating their solving process through (non-adaptive) guidance from a fictitious
character (one for each of the four solving steps). The tasks were autocorrected, and
direct feedback was given to the student.
The different arithmetic tasks were addition and subtraction word problems as
recommended in the curriculum for grade 3 (MENFP, 2011), based on the criteria for
constructing and solving arithmetic word problems (Franke & Ruwisch, 2010) and
structured through the semantic classification of Vergnaud(1982) in transformative,
compositional and comparative problems. The tasks were related to situations and
places from the students’ living environment: “Luc does a bike tour from Luxembourg
to Echternach with his 3 friends. The odometer on his bike is at 125 km at the start of
his trip. The tour is 42 km long. What will the odometer show when they get to
Echternach?”.

Figure 2: Guided use of a meta-tool: highlighting information and creating a scheme

The first part of the module was dedicated to learning process skills (first 14 “blue”
tasks after the key in figure 1). Students practised different process skills separately in
guided learning tasks. These tasks then lead to discovery and manipulation of meta
tools supporting the different process skills (i.e. identifying relevant information in the
wording with a highlighter tool and creating a resolution scheme, compare figure 2) to
make it easier for students to solve the problem.

53
18.05.2021Haas Ben
Figure 3: “Black” item without guidance
The second part focused on using the learned process skills in the identified arithmetic
word problems (15 tasks organised as 3 ovals on the top of figure 1). Thus, students
solved the different typologies of arithmetic word problems (combination,
transformation and comparison) with the help of the learned meta-tools (compare
menu bar on the left of figure 3). Each typology was presented in a set of three levels
from guided tasks (blue), semi-guided tasks (red) to complex tasks (black). In the
guided tasks (blue) in each typology, students needed to follow four steps solving
procedure using the learned meta-tools: they analysed the wording, modulated the
content into a resolution plan, executed the arithmetic operations and verified their
results. In the semi-guided tasks (red), students were asked if they wanted to use the
meta-tools, but could choose not to utilise them. In the complex tasks (black), they
solved tasks with multiple arithmetic operations and no scaffolding was offered. They
could use the meta-tools, but without additional guidance.

METHODOLOGY
In this section, we describe the quantitative pre-/post-test approach we utilised to
measure if students in grade 3 (age 8 to 10) who learn arithmetic word problems with
MathemaTIC in a one-to-one setting are likely to improve at the same degree
compared to a traditional paper-and-pencil setting with teachers. The experimental
group worked with the arithmetic word problem module in MathemaTIC in a
one-to-one setting, without a specific teacher guidance. Their teachers did only ensure
access to
54
18.05.2021Haas Ben

MathemaTIC and helped with technical issues. The control group did the same tasks
using paper-and-pencil, however with the guidance and assistance of their teachers.
Both groups worked for 20 hours (2 hours per week over a period of 10 weeks) on
process skills in arithmetic word problems. During the study, we observed three
different moments in the learning behaviour of the students within the experimental
and control groups. Besides, we interviewed their teachers on their perception of the
students’ learning with or without MathemaTIC based on the research questions RQ1
and RQ2.
Participants of this quantitative study were 48 randomly selected classes with 667
students in grade 3 (age 8 to 10) in elementary schools in Luxemburg. We used the
variables age, gender, and performance in mathematics of EpStan to identify matched
pairs and assigned classes to experimental and control groups. We allocated 278
students to the control group, working using paper-and-pencil, and 389 students to the
experimental group, working with the arithmetic word problems module in
MathemaTIC. At the end of the study, 34 classes with 472 (8 with a missing post-test)
students remained: 246 (2) students in the experimental group and 226 (6) students in
the control group. The dropout was due to local technical errors (low WiFi signal or
non-working hardware) while using MathemaTIC or simply because of a missing post
test for the whole class. Students in both groups performed an identical pre-test and
post-test with 15 items based on the different typologies of arithmetic word problems
(combination, transformation and comparison) with one or two operations and one
item with a combinatorial problem. This combinatorial item allowed us to observe if
students would transfer their learned skills into another typology of problem. Both
tests have been created based on the experiences from author groups from the national
school monitoring and based on the skills from the curriculum.

RESULTS AND DISCUSSION

Results from the experiment suggest that the use of the module “arithmetic word
problem” in MathemaTIC is a promising approach to foster process skills in
mathematics in a one-to-one setting. The statistical analysis below was carried out
using the software R (R Core Team, 2020).
Cronbach’s alpha (Revelle, 2019) indicates a good reliability for the pre-test (α=0.774)
and the post-test (α=0.787). The detailed analysis shows that dropping one of the 16
items will only slightly increase the reliability for question 1 of the pre-test (α=0.777)
and that there are no reverse-scored items. Thus, from this point of view, all items are
to be kept in the tests. However, several questions in the pre-test (Q1: 0.16, Q3: 0.27,
Q11: 0.24) as well as in the post-test (Q1: 0.26, Q3: 0.26) have an item-rest correlation
below 0.3. Hence, they do not correlate very well with the scale overall. The success
rates are 81% vs. 76% for question 1 (low difficulty level), 18% vs. 38% for question
3 (high difficulty level) and 3% vs. 9% for question 11 (very high difficulty level).
Although extremely easy or difficult items only poorly allow to discriminate, they were

55
18.05.2021Haas Ben

needed to sample content and objectives adequately. Thus, we kept all items for further
analysis.
Figure 4: Increase in score over time for the group
We used lme4(Bates et al., 2015) to perform a linear mixed-effect analysis of the test
result score predicted by the fixed effects time (pre-/post-test), control/experimental
group and their interaction as well as the random effect student. Visual inspection of
residuals plots revealed minor deviations from homoscedasticity and normality, which
we accounted for by using bootstrapped confidence intervals. The main effect time has
an estimate of 1.14 points (95% CI [0.56, 1.72]) for the control group. Thus, post-test
scores of students from the control group were significantly higher than those in the
pre-test. The main effect group has an estimate of -0.29 points (95% CI [-0.89, 0.28])
in the pre-test. On the one hand, both groups were comparable at the beginning of the
study, because the confidence interval contains 0, and on the other hand, the
experimental group probably had, in the pre-test, slightly lower test results than in the
control group. Finally, the interaction effect time x group had an estimate of 0.90
points (95% CI [0.15, 1.68]). This effect underscores the fact that the performance
gains of the experimental group, working with the educational software MathemaTIC,
were significantly larger than those in the control group, resulting in somewhat better
post test performance although starting with a somewhat lower pre-test performance
(compare figure 4).
During the classroom observations and the interviews, we were able to collect data in
both groups on the motivation, participation and transfer of skills. Thus, in the
experimental group, teachers reported that students' motivations to solve and discuss
56
18.05.2021Haas Ben

arithmetic word problems were higher than during the regular course (without
MathemaTIC) and that they voluntarily exchanged on the tasks after the resolution.
Teachers attributed the increase of motivation to the gamification aspect of
MathemaTIC as well as the guidance and direct feedback given by the educational
technology. According to teachers’ reports, some students suggested in other teaching
hours (without MathemaTIC) to use the learned process skills to solve mathematical
tasks (i.e.: Calculating the area of the classroom floor). In the control group, teachers
stated that there was no change in motivation and some students had significant
difficulties (i.e. understanding wording or findings of the arithmetic operation) to solve
all the given tasks on paper.

CONCLUSION AND OUTLOOK

Our findings highlighted that students in the experimental group improved their
performances in arithmetic word problems significantly using the educational
technology software MathemaTIC in one-to-one setting. Students learned meta-tools
on process skills and successfully solved addition and subtraction word problems in all
topologies without the direct guidance of a teacher or a parent. Teachers reported a
high acceptance in class and an overall increase in motivation and participation of the
students in mathematics courses. Thus, the module on arithmetic word problems in the
educational technology software MathemaTIC is a viable alternative to the traditional
paper-and-pencil course. Over time it could be a valuable asset to support students
individually or in groups or even the entire class with MathemaTIC within traditional
courses.
We will perform further investigation on the fixed effects of gender, age, nationality,
spoken language (L1) and the random effect school. Additionally, we will investigate
all process skills in detail by performing a qualitative comparison of the pre-test and
the post-test in our future analyses. Hopefully, we can further narrow the origin of the
observed significant performance gains of the experimental group and we will report
these analyses in future publications.

ACKNOWLEDGEMENTS
The Ministry of Education (MENJE) and the company Vretta are thanked for the
collaboration and initiatives on the educational technology software MathemaTIC. The
University of Luxemburg (LUCET) is thanked for the support in this research.

REFERENCES
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting Linear Mixed-Effects
Models Using lme4. Journal of Statistical Software, 67(1), 1–48.
De Corte, E., Verschaffel, L., &Masui, C. (2004). The CLIA-model: A framework for
designing powerful learning environments for thinking and problem solving.
European Journal of Psychology of Education, 19(4), 365–384.

57
18.05.2021Haas Ben

Franke, M., &Ruwisch, S. (2010). Didaktik des Sachrechnens in der Grundschule (F.
Padberg, Ed.; 2nd ed.). Springer Spektrum.
LeBlanc, M. D., & Weber-Russell, S. (1996). Text integration and mathematical
connections: A computer model of arithmetic word problem solving. Cognitive
 Science, 20(3), 357–407.
Mayer, R. (2005). The Cambridge Handbook of Multimedia Learning (R. Mayer, Ed.).
Cambridge UniversityPress.
MENFP. (2011). Plan d’études: École fondamentale. Ministère de l’Éducation
nationale et de la Formation professionnelle. http://www.men.public.lu/catalogue
publications/themes-transversaux/cen/cens/plan-etudes/fr.pdf
R Core Team. (2020). R: A Language and Environment for Statistical Computing
(4.0.1) [Computer software]. R Foundation for Statistical Computing.
https://www.R-project.org/
Revelle, W. (2019). psych: Procedures for Psychological, Psychometric, and
Personality Research (1.9.12) [Computer software]. Northwestern University.
https://CRAN.R-project.org/package=psych
Selter, C., &Zannetin, E. (2018). Mathematik unterrichten in der Grundschule:
Inhalte—Leitideen—Beispiele. Friedrich Verlag GmbH.
Sonnleitner, P., Krämer, C., Gamo, S., Reichert, M., Muller, C., Keller, U., &Ugen, S.
(2018). Schülerkompetenzen im Längsschnitt—Die Entwicklung von Deutsch
Leseverstehen und Mathematik in Luxemburg zwischen der 3. Und 9. Klasse.
Luxembourg Centre for EducationalTesting, UniversitätLuxemburg; Service de
Coordination de la Recherche et de l’Innovation pédagogiques et technologiques
(SCRIPT).
Trouche, L. (2004). Environnements Informatisés et Mathématiques: Quels usages
pour quels apprentissages? Educational Studies in Mathematics, 55(1), 181–197.
van Merriënboer, J. J. G., & Kester, L. (2005). The Four-Component Instructional
Design Model: Multimedia Principles in Environments for Complex Learning. In R.
Mayer (Ed.), The Cambridge Handbook of Multimedia Learning (pp. 71–94).
Cambridge University Press; Cambridge Core.
Vergnaud, G. (1982). A classification of cognitive tasks and operations of thought
involved in addition and subtraction problems. Addition and Subtraction : A
Cognitive Perspective, 39–59.

58
18.05.2021Haas Ben
4.1.1.2. Summary, highlights and next developments
We applied our theoretical model on process skills and problem-solving approach (see section
2.1) to an online learning environment that allowed students to progress individually and obtain
higher performances. Furthermore, digital tools and their functions (e.g., highlighting information
in the wording or self-regulating the solving process), which created on non-technology
enhanced research and literature, influenced solving processes of word problems by students
within the experimental group. Based on our findings, we suggested that most students received
sufficient guidance (e.g., auditive, visual) to develop new skills in connecting mathematical
concepts to real
world information. However, as we concluded, more data analysis was needed to identify the
groups of students, depending on their socio-lingual backgrounds, and how students improved
their performances in solving word problems. Furthermore, we wanted to investigate students’
motivations that they experienced during the use of MathemaTIC and understand if the elements
in MathemaTIC (e.g., gamification, animations or self-paced learning) influenced students’
motivation in solving word problems.

In the second publication, we outlined a more in-depth analysis of students' performances in this
study.
4.1.2. What works: Arithmetic word problems within the educational learning
environment MathemaTIC v.2017
Reference:
Haas, B., Kreis, Y., Lavicza, Z. & Sonnleitner P. (In Review). What works: Arithmetic word
problems within the educational learning environment MathemaTIC v.2017. International Journal
of Mathematical Education in Science and Technology. (Paper B)
4.1.2.1. Short description of the IJMEST paper
In this publication, we detailed the design of the arithmetic word problem module (i.e., tasks,
learning paths, problem structures) and further explained digital tools and their functions based
on instrumental genesis (Lieban & Lavicza, 2019; Trouche, 2004a) with MathemaTIC.
Furthermore, hindrances and opportunities (e.g., technological equipment in schools, technical
issues) were described in this second publication. We further analysed our data to identify
performance gains for different groups in connection to their socio-lingual backgrounds.
Moreover, we investigated students’ motivational factors, which turned out to be essential for
their mathematics learning with MathemaTIC.

59
18.05.2021Haas Ben
What works: Arithmetic word problems
within the educational learning environment MathemaTIC v.2017

Ben Haasa, Yves Kreisb, Zsolt Laviczac and Philipp Sonnleitnerd

a
Johannes Kepler University Linz, Linz School of Education, Austria; bUniversity of
Luxembourg, Department of Education and Social Work, Luxembourg; cJohannes Kepler
 University Linz, Linz School of Education, Austria; dUniversity of Luxembourg, Luxembourg
Centre for Educational Testing, Luxembourg

ARTICLE HISTORY
Compiled May 10, 2021

ABSTRACT
According to teachers in Luxembourg’s elementary schools, a major challenge for
grade 3 (age 8 to 10) students is solving arithmetic word problems. Due to the high
classroom heterogeneity, teachers report difficulties in teaching word problems to
their students. Thus, students’ different social and language profiles (e.g., luso
phones students in german-speaking courses) require much support from teachers,
often not feasible within school time. Moreover, teachers report that students have
difficulties applying the in-class learned solving strategies in arithmetic word prob
lem tasks. Hence, solutions in supporting students and teachers were sought with
educational technology. One technology was the arithmetical word problem module
within the educational technology software MathemaTIC, with automated tutor ing.
The module was designed to gradually increase skill levels and give students
feedback on their learning outcomes. Several digital tools have been developed to
support in developing new schemes in arithmetic word problem-solving. This paper
describes designs and structures of automated tutoring, scaffolding, and digital tools
implemented in MathemaTIC. We compared MathemaTIC to teachings in elemen
tary schools in Luxembourg. This quantitative analysis performed during the initial
design process with an experimental group (n=246) with MathemaTIC and a control
group who performed the same tasks on paper-pencil (n=226) will be discussed.

KEYWORDS
Automated Tutoring System; Scaffolding; Arithmetics; Elementary Schools;
Educational Technology; Problem Solving

1. Introduction

In Luxembourg, arithmetic word problem solving is an elementary school topic

where most students experience prevalent learning difficulties. In our studies in
grade 3 (age 8 to 10), teachers reported low comprehension skills in understanding
wordings and identifying operations in arithmetic word problem tasks. The result of
the national testing in mathematics and course language comprehension reported
similar observa tions, where students with low reading skill scores are also those
who are low per forming in mathematics [36]. Moreover, according to LeBlanc and
Weber-Russel [21] well-developed process skills are related to well-developed skills
in reading and under

CONTACT Ben Haas. Email: ben.haas@outlook.com

60
18.05.2021Haas Ben
standing of the mathematics course language. These skills are among those
process skills in mathematical thinking and content skills [12], needed for arithmetic
problem skills. Hence, to develop skills in arithmetic word problem solving, a
majority of stu dents require professional assistance from teachers through constant
feedback. In class, however, this is challenging for teachers as there are large
learning groups with a high heterogeneity in spoken languages and social
background. Therefore, a close follow-up is often difficult to realize by the teachers.
In response to support teachers and stu dents in elementary school in grade 3, a
multidisciplinary team (i.e., subject-matter experts, researchers, teachers,
instructional designers and developers) designed, within an international
governmental project in Luxembourg, an automated tutoring system to foster skills
in arithmetic word problems solving. This system was embedded in a learning
module on arithmetic word problems in the educational software Mathe maTIC [13,
14]. The instructional design aimed to let students learn in a one-to-one setting
 accessible in class and at home, independently of their performances (i.e.: low,
medium or high performer according to their classroom results), without the need for
constant feedback from teachers. In addition to the automated tutoring system, we
created digital-tools (e.g., logbook to structure the solving process or highlighter to
highlight information in the wording), which could influence solving behaviours of the
students. We aimed to be as close as possible to teachings and textbooks used in
classrooms, to get a broader acceptance among teachers. Thus, we investigated
which different teaching methodologies were used by teachers and analyzed them.
This in vestigation guided the design process of the automated tutoring system.
During the initial design process, we performed experimental research with a
pre-/post-test de sign. Our research was guided by following research questions:
RQ1: Are students who learn arithmetic word problems with MathemaTIC likely to
improve at the same degree, or higher, compared to a traditional paper-and-pencil
setting with the guidance of the teachers?
RQ2: What factors (e.g., spoken language by students), if there are any, abet or hin
der learning arithmetic word problems within MathemaTIC, compared to a traditional
paper-and-pencil setting with the guidance of the teachers?
We collected data on performance evolution, with an experimental group (n=246)
working during 9 weeks on the arithmetic word problem module in the educational
technology software MathemaTIC, without their teachers, compared to a control
group (n=226), doing the same tasks on paper-pencil with their teachers. We
collected data on spoken language , gender and attitude towards arithmetic word
problems of the participants. We were able to find a significant improvement in
performances of stu dents working with the arithmetic word problem module in the
educational technology MathemaTIC [15]. In this paper, however, we will further
describe the automated tu toring system, the task design and the digital tools.
Moreover, we will describe factors which we were able to observe in both groups,
related to the fixed effects of attitude, gender, nationality and the spoken language
of students.

2. Theoretical background and design process

2.1. Design process

Similar to earlier developments within other modules in MathemaTIC we tried to mir
ror the national curriculum and teachers methodologies, to get a broader
acceptance among teachers, similar to Peterson and Herrington [30], who stated
that complex

61
18.05.2021Haas Ben
problems should be developed in real contexts with practitioners. Hence, during the
design process, we tested tasks in classes, observed students and teachers while
try ing to use digital tools or technological functionalities. We designed the arithmetic
word problem module and tasks in iterative development cycles. Thus, the instruc
tional design process was a design-based research process [5], rather than
predictive research [1]. After literature reviews on skills in arithmetic word problems
[12, 34, 40], the structure of learning in educational technology [11, 24] and digital
tools design [8, 9, 37], we started to develop the module and tasks in a
multidisciplinary team with the Ministry of Education in Luxembourg (SCRIPT), the
Company Vretta from Canada, the University of Luxembourg and Luxembourg
Institute of Socio-Economic Research (LISER). In the first development, we
performed an expert review (i.e., math ematics teachers, elementary school
teachers, human-computer-interaction researchers, assessment experts,
subject-matter experts). With these experts’ inputs, we did two rounds of
development and in-class testing (i.e., two grade 3 classes). In classroom testing,
 we presented tasks and digital tools to students in grade 3 (age 8 to 10). With a
think-aloud protocol [6, 10], we collected information on the students’ cogni tion and
manipulations with the developed digital tools and tasks in the automated tutoring
system. To consider teachers’ needs and to improve the handling of the tasks, we
collected observations from the class teachers. The final design was then used for
quantitative testing with 667 students.

2.2. Theoretical framework for solving arithmetic word problems

We started the development with an analysis of the national curriculum [26, 27] to
identify content and process skill levels in arithmetic word problems. Arithmetic word
problem solving is among the four domains in the curriculum. Each domain is then
divided in key skill developments throughout the elementary school teachings (e.g.,
oral picture problem solving in grade 2 to open ended complexe problems in grade
4). In grade 3, teachings focus on solving written arithmetic word problems by de
veloping key skills in recognizing and applying mathematical concepts from
wordings. These are one step or two step arithmetic word problems. We
interconnected the skill levels students need to develop with didactical principles
used in teaching method ologies in elementary schools in Luxembourg [4, 34] and
international standards on process skills [29]. The different references, national
curriculum, didactic principles and international standards, had high intersections
and differed mainly due to socio linguistic aspects of the different languages (e.g.,
NCTM in English, the curriculum in French, didactic principles in German). Based on
these reviews, we identified four major process skills (i.e., problem solving,
reasoning and proof, representation and modelling). We compared process skills to
different problem solving approaches, to identify transitions from wordings to
mathematical models and interconnections be tween the different process skills.
Moreover, as we designed the module on arithmetic word problem solving in a
one-to-one setting (e.g., without assistance from the teacher or parent), we wanted
to integrate all needed developments and applications of skills in solving arithmetic
word problems, to present tutoring and scaffolding accordingly for students. Hence,
in addition to the four identified process skills, we specified further skill
developments (e.g., structural representations and understanding of wordings). After
consulting grade 3 teachers and performing several observations in class, we
connected observed and reported scaffoldings to different researches in problem
solv ing [12, 22, 24, 33, 35], to identify skill developments, which we needed to
implement

62
18.05.2021Haas Ben
in the module on arithmetic word problems.
 Figure 1. Problem solving approach in the arithmetic word problem

In figure 1, we assembled procedural steps in solving arithmetic word problems,

identified from intersections of analysed approaches in problem solving. Each step
(e.g., understanding and representation of wordings, mathematical modelling,
solving arithmetic operations, evaluation of results and mathematical models)
requires devel opment of specific skills. In an arithmetic word problem, the algorithm
is not visible immediately and students have to apply a problem solving approach to
identify the required mathematical concepts. As described by Kintsch [19], students
combine the textbase of the wording with long term memory and create first a
situation model. Hence, students use their linguistic skills [21] to understand the
message of wordings and create structural representations. According to the
landscape model [38], struc tural representations are built upon two sources of
comprehension. The first is based on the memory of students, where possible
mental representations are generated from the wordings and the second is verifying
the connections between information and con sistency of the representation [3].
Furthermore, Miller and Johnson-Laird [28, p. 693] stated: “to know the meaning of
a sentence is to know how it could be verified”. Hence, within the module students
should work on skills to verify or self-regulate their assumptions during the analysis
of the wordings, the created schemes and the transfers to mathematical models [8].
According to Baffrey-Dumont [2], students re construct mathematical concepts
based on understanding of connections in between the different elements in the
wording. Further, the mathematical modelling approach is depending on the
typologies of arithmetic word problems [41]. To develop tutoring and digital tools
which should guide students, we investigated several instructional design
frameworks.

63
18.05.2021Haas Ben
 Figure 2. Design of the arithmetic word problem module overview (v.2017)

2.3. Instructional design of the module and tasks

Based on the Competence-Learning-Intervention-Assessment (CLIA) model by De
Corte et al. [7] and the Four-Component Instructional Design (4C-ID) model by van
Merri¨enboer and Kester [39] for learning complex skills, we developed a tutoring
struc ture within the module (figure 2) and tasks. According to Melo Miranda [25],
these models have shown to be promising in supporting students either with weak
or strong knowledge of complex skills, which we estimated would respond to the
high hetero geneity in classes. Thus, students started with worked-example tasks
[16], designed with high instructional guidance for each new typology of word
problems. These tasks (blue in figure 2) were related to real-life situations related to
student’s living expe riences (e.g., visiting a local zoo or travelling by tram) and
prepared for less guided tasks, with procedural information on routines (e.g., type of
arithmetic operations to perform) within the tasks. To develop non-routine
procedures (e.g., problem solving, understanding wordings, mathematical modeling
and self-regulation), we designed first tasks in the module to foster these
procedures with digital-tools and high instructional guidance. This approach is
similar to the observed classroom teaching, where a teacher would explain with
hands-on activities (e.g., use a colour pen to underline important information or draw
a scheme with colour pens) the procedures and support students. Thus, we used
the mentioned high instructional guidance, due to the absence of the teacher in
MathemaTIC, to fully explain the procedure and several digital tools to train these
procedures.
With these digital tools, we aimed to support students in learning the non-routine
procedures. Thus, based on the instrumental orchestration described by Trouche
[37], we designed digital tools (artefacts) in the learning environment to enable
students to develop the described non-routine procedures, through the use and
instrumentalisation

64
18.05.2021Haas Ben
of these digital tools. Thus, with the digital tools in arithmetic word problems,
students perform cognitive processes supported with technology (e.g., transferring
information of written wordings into a conceptual representation or code a
mathematical model with imminent feedback).
 There were overall 3 digital tools presented to the students, a marker to highlight
information and transfer information to conceptual representations, a planification
and calculation tool to identify and perform the mathematical concepts, and a
logbook, to support students in self-regulating the solving process. For each process
skill, we created a fictitious character to guide with (non-adaptive) high instructional
guid ance the student during the learning process. These fictitious characters
reappeared in the different tasks in the module and students could request
automated support by clicking on the character (e.g., receiving guiding questions on
using digital tools). The characters were designed similar to the teachers’
scaffoldings and supports observed in class and perform as “virtual teachers”,
however without an adaptive approach.
After completing tasks with high instructional guidance, students worked on tasks
with optional guidance (red tasks in figure 2) and complex (e.g., more than one op
eration and multiple distractors in the wording) tasks without guidance (black tasks
in figure 2). Hence, similar to 4C-ID and CLIA, students worked from high guidance
training tasks to complex word problem tasks without guidance. This structure
aimed to leave the choice to students on which digital tools they would want to
resort to. The scaffolding system in the tasks was based on the multimedia learning
theory [24] and dual coding theory [32], to adapt to the cognitive load students could
handle in grade 3. We performed several in-class testing, as mentioned earlier,
calibrate the amount of information and visual stimuli in a task.

Figure 3. Task view with digital tools

The instructional guidance (compare figure 3) consisted in listening to wordings
(e.g., read aloud), interacting with images and creating conceptual representations
with guiding questions (e.g., questions suggesting alternatives or missed
information). Fur-

65
18.05.2021Haas Ben
ther, students interacted with arithmetic operations (e.g., using a digital tool to con
struct the operations), and self-regulated their solving process through
(non-adaptive) guidance from fictitious characters (one for each of the four solving
steps). The dif ferent tasks were autocorrected, and direct feedback was given to
the student with visual stimuli (e.g., green-lighting button for correct answers and
red lighting button for incorrect answers).
 2.4. Typologies of arithmetic word problems
In the beginning of grade 3 [27] the relevant wordings and operations in our study
were addition and subtraction word problems.
Table 1. The different types of arithmetic word problems used in the module arithmetic word problem
Narrative Description Examples (in the module arithmetic word problem solving)
Composition A composition of two or three el They see food for animals for 9¿, a cap for
ements: The student is led to a 15¿. a raincoat for 17¿ and candies for 12¿.
search for the compound or one They buy the animal food and candies. How
of the measures much do they spend?
A plane takes off at 11:00 from Findel to land
Transformation A transformation of one of the at 12:50 in Porto. How long does it take to get
elements: The student has to to Porto?
search for the transformation or
one of the states (initial or final) Linda and Susan are comparing how far they
Comparison A comparison of two elements: The can jump. Linda jumps 178cm. Susan jumps
student compares two ele 153cm. How many more cm does Linda jump
ments than Susan?
Lara and Max go to the zoo souvenir shop.

The typologies of word problems from national curriculum and textbooks (i.e.,
Zahlenbuch 3 [43]) were similar to the classification of Vergnaud [40], who described
three different narratives and operations in addition and subtraction (table 1). For
each of these types of arithmetic word problems (15 tasks organised as 3 ovals in
fig ure 2), we created learning paths with tasks with high instructional guidance,
optional guidance and complex without guidance. Moreover, we started for each
type (com position, transformation and comparaison) with rather simple wording
structures (no distractors) up to complexe wordings (with at least two distractors).
The distractors (e.g., numbers and quantities not related to the problem) were
similar to the tasks used in class and based on experiences reported from teachers.
Students learned the typologies and how to identify the according wordings in
MathemaTIC with support and guidance, similar to experiences students could
receive from their teachers in class.

2.5. Participants
48 randomly selected classes with 667 students in grade 3 (age 8 to 10) in
elementary schools in Luxembourg participated in this quantitative study. Based on
the variables age, gender, language (i.e., Luxembourgish and German speaking v.s.
others) and per formance in mathematics (i.e., national testing EpStan), we

identified matched pairs ´ and assigned classes to experimental and control groups.
We allocated 278 students to the control group, working on paper-and-pencil tasks
with their teachers, and 389 students to the experimental group, working with the
arithmetic word problems mod ule in MathemaTIC without their teacher. During the
intervention, the control and experimental group worked for 20 hours (2 hours per
week over a period of 10 weeks) on tasks related to arithmetic word problems. The
control group received the same

66
18.05.2021Haas Ben
tasks (i.e., wordings of the arithmetic word problems) as the experimental group,
how ever using paper and pencil and with scaffoldings from their teachers. Teachers
used their usual teaching approaches, which were similar to the ones we observed
during the design of the learning environment and didn’t rely on technology (e.g.,
calcula tors). In the experimental group, students performed the tasks within
MathemaTIC and received no additional guidance from their teachers. At the end of
 the study, 34 classes with 472 (8 with a missing post-test) students remained: 246
(2) students in the experimental group and 226 (6) students in the control group.
The dropout was due to local technical errors (low WiFi signal or non-working
hardware) while using MathemaTIC or simply because of a missing post-test for the
whole class. Between both groups, no significant differences were found concerning
gender, nationality or language.

2.6. Measures

2.6.1. Mathematical word problem-solving

We assessed arithmetic word problem-solving using 16 items of varying type
(combina tion, transformation, and comparison), complexity (one or two operations
necessary for solving the problem) and response format (multiple-choice,
open-response, and circuiting). Those items were curriculum-based and developed
by the authors in close collaboration with psychometric and content experts from the
Luxembourgish school monitoring program.
To rule out any forms of differential item functioning (DIF) masking effects of the
training itself, we drew on IRT-framework [20] to rigorously analyze the pretest item
pool with regard to group (experimental vs. control), score (high vs. low performers),
and gender. Additionally, we checked whether items worked differently for students
speaking the language of instruction at home (Luxembourgish or German speakers
compared to all other languages) or depending on students’ nationality (Luxembour
gish vs. Other). Andersen’s Likelihood-Ratio tests (LRT) and Wald tests revealed
DIF for items 1, 14 (both median split), and 12 (gender). After excluding those items,
the final item pool proved to fit the Rasch model (LRT χ2 = 16.56, df = 12, p = .17)
and could thus be considered fair and reliable with respect to the investigated
criteria.

2.6.2. Students’ attitude towards arithmetic word problems

At the beginning of pre- and post-test, students were asked how much they liked
solving arithmetic word problems. Students responded on a non-verbal, three-point
Likert scale. Hence, this item was used to infer students’ attitude towards dealing
with this kind of problems.

2.6.3. Data analysis

All data analyses were conducted within the R-framework [31]. Test scaling was
done using the package eRm [23], descriptives, group comparisons, and three-way
mixed ANOVA were performed using tidyverse [42], ggpubr [17], and rstatix [18].
Type I error was set to α = .05 for all analyses.

67
18.05.2021Haas Ben
Table 2. Means, standard deviations and quartiles
Measure M SD Min Q1 Q2 Q3 Max
Experimental group (n = 244)
Pretest 3.98 2.64 0 2 4 6 11
Posttest 5.81 3.08 0 3.75 6 8 13
Control group (n = 220)
Pretest 4.25 2.83 0 2 4 6 12
Posttest 5.25 2.88 0 3 5 7 11
 3. Results and discussion

3.1. Attitude towards solving arithmetic word problems

On a descriptive level, both groups showed a similar attitude towards solving arith
metic word problems at the beginning of the study. Half of the students (52-54%)
reported an indifferent attitude to deal with those kinds of problems and around one
third (34-38%) liked such tasks. This, however, changed at the end of the study with
al most half of the students of the experimental group (47%) reporting a positive
attitude towards mathematical word problems compared to an unaltered proportion
within the students of the control group (36%). Although a χ2-test between the
groups’ attitude when taking the posttest was not significant (χ2 = 5.55, df = 2, p =
.06), we interpret this shift as indication that working with MathemaTIC might
improve the student’s motivation to deal with arithmetic word problems in general.
Note that this effect cannot be attributed to the experimental group’s better
performance in the posttest since the related question was asked at the beginning
of the test session.

3.2. Training effects of MathemaTIC

3.2.1. Group differences at the beginning of the study

At the beginning of the study, the administered test appeared to be too difficult for
the majority of the students resulting in a skewed distribution and floor effect in the
control as well as the experimental group. 50% of all participants were not able to
solve more than four (out of 13) items with none of them solving all presented word
problems (see table 2 for descriptives). Hence, no significant differences could be
found in problem-solving ability between the groups with regard to variance
(Levene’s F = 0.37, df = 1, p = .55), or total sum score (student’s t = 1.03, df = 462,
p = .30). Thus, both groups were comparable in terms of their pre-intervention ability
to solve mathematical word problems.

3.2.2. Group differences after the intervention

During the course of the study, all participants improved their ability to deal with
arithmetic word problems, resulting in a much better fit between students’ ability and
test difficulty. Whereas the control group increased its performance on average by
one additionally solved item, the experimental group managed to solve almost two
items more (¯xdiff = 1.83) compared to the pretest (see table 2). Whereas three
students in the experimental group succeeded in solving all 13 items, maximum
score of the control group even dropped from 12 (pretest) to 11. A slight increase of
variance could be observed in the experimental group, rising from SDpre = 2.64 to
SDpost = 3.08.
To further evaluate the impact of the experimental group’s training with Math
emaTIC while simultaneously controlling potential language effects, we performed

68
18.05.2021Haas Ben
a three-way mixed ANOVA with group assignment and language spoken at home
(German & Luxembourgish vs. Other) as two between-subjects factors and time as
within-subjects factor. No extreme outliers (3 times interquartile range below Q1 or
above Q3) were detected and homogeneity of variances was observed for all
conditions (p > .05). Except the aforementioned floor effect of pretest performance,
inspection of QQ-plots led us to assume reasonable normality distribution of the
data.
 Although no statistically significant three-way interaction was found, results
clearly showed significant interactions between language spoken and time (F =
23.8, dfs = 1, 385, p < .01). In addition, substantial main effects for time (F = 74.25,
df = 1, 385, p < .01), language spoken (F = 26.42, df = 1, 385, p < .01), and group
(F = 5.47, df = 1, 385, p = .02) were found. These results confirmed that arithmetic
word problem-solving ability improves independently from the treatment condition
but that this development was substantially influenced by the experimental condi
tion the students were in, or which language they spoke. Importantly, however, the
spoken language did not interfere with the effects of the training. Post-hoc t-tests
showed, that although both groups significantly improved their word problem-solving
skills (p < .01 for both groups), working with MathemaTIC substantially added to this
development: whereas for the control group a small effect was found (Cohen’s d =
.25), data showed a medium effect (Cohen’s d = .53) in the experimental group. In
other words, working with MathemaTIC accounted for one additionally solved item in
the posttest. Likewise, speaking German or Luxembourgish substantially helps with
arithmetic word-problems.
Figure 4 demonstrates the effect of MathemaTIC and language spoken at home.
Importantly, non-German and non-Luxembourgish speaking students only minimally
improved without intervention (+ .33 items). Working with MathemaTIC, however,
the performance of such students increased by 1 item on average (+ 1.08 items).

4. Conclusion

Throughout the study, we were able to observe several positive effects on arithmetic
word problem-solving while using MathemaTIC. Thus, students in the experimental
group, working in a one-to-one learning setting, significantly improved their perfor
mances in arithmetic word problem solving. The educational technology supported
students with scaffoldings and thus students were able to learn without additional
help from a teacher or parent. Yet, some students requested support for technical is
sues (e.g., browser update or internet connection). These issues were mainly solved
by the teachers or researchers during the sessions. Nonetheless, MathemaTIC
showed to be an alternative to use in class, to differentiate teachings and moreover
support students with close assistance. Thus, teachers could make different use of
their time and focus their support on those students with higher needs of assistance.
During the times of a pandemic, where students need to work at home or with social
distancing, MathemaTIC could be a valuable asset in learning arithmetic word
problem solving.
Language skills play a vital role in the acquisition of skills needed to solve
arithmetic word problems. Hence, students with difficulties in the instruction
language are clearly disadvantaged when learning these competencies: Whereas
German speaking students significantly improved their skills even without special
intervention (control group condition); no increase in performance was found for
their non-German speaking peers. Crucially though, working with MathemaTIC
substantially improved performance of non-German speaking students by around 1
item and analyses revealed no differential

69
18.05.2021Haas Ben
 Figure 4. Differences in arithmetic word problem-solving depending on working with MathemaTIC (experi mental
group) and language spoken at home.

training effect depending on student’s language. Thus, our results suggest that
using educational platforms such as MathemaTIC could help to close the
performance gap caused by differences in mastering the language of instruction.
Finally, in the experimental group, we observed a shift towards a more positive
attitude to solve arithmetic word problems. We assume that this effect is related to
the gamification of the tasks and the close guidance. Students got immediate
feedback and guidance in a playful way, which was different from the traditional
formal teachings. Further, with the digital tools, they proceeded to an active
mathematical modelling and worked at their own pace.

5. Outlook and further research

The output of the study on arithmetic word problem learning with MathemaTIC was
promising. However, a deeper analysis and comparison of the different word
problems and solving procedures students applied after the use of MathemaTIC
could give us more information on the success rate. The word problems differed in
their complexity (e.g., one or two step problems, number of distractors or the nature
of the wording) and requested lower or higher skill developments. We will therefore
analyse and present in upcoming publications, how students solved the different
word problems. This will certainly clarify and give more details to the difference in
learning with or without

70
18.05.2021Haas Ben
MathemaTIC.

Acknowledgment
 The Ministry of Education (MENJE) and the company Vretta are thanked for the
collaboration and initiatives on the educational technology software MathemaTIC.
The University of Luxembourg (LUCET) is thanked for the support in this research.

References

[1] T. Amiel and T.C. Reeves, Design-Based Research and Educational Technology:
Rethink ing Technology and the Research Agenda, J Educ Technol Soc 11 (2008), pp.
29–40. [2] V. Baffrey-Dumont, R´esolution de probl`emes arithm´etiques par des enfants
de huit ans, Rev sci ´educ 22 (1996), pp. 321–343.
[3] M. Bianco, Du langage oral `a la compr´ehension de l’´ecrit, Regards sur l’´education,
Presses universitaires de Grenoble, Grenoble, 2015.
[4] W. Blum and D. Leiß, How do Students and Teachers Deal with Modelling Problems?,
in Mathematical Modelling (ICTMA 12): Education, Engineering and Economics, C.
Haines, P. Galbraith, W. Blum, and S. Khan, eds., Woodhead Publishing, Chichester,
2007, pp. 222–231.
[5] A. Collins, D. Joseph, and K. Bielaczyc, Design Research: Theoretical and
Methodological Issues, J Learn Sci 13 (2004), pp. 15–42.
[6] D. Cotton and K. Gresty, Reflecting on the think-aloud method for evaluating e-learning,
Br J Educ Technol 37 (2006), pp. 45–54.
[7] E. De Corte, L. Verschaffel, and C. Masui, The CLIA-model: A framework for designing
powerful learning environments for thinking and problem solving, Eur J Psychol Educ
19 (2004), pp. 365–384.
[8] R. Duval, A Cognitive Analysis of Problems of Comprehension in a Learning of Mathe
matics, Educ Stud Math 61 (2006), pp. 103–131.
[9] R. Duval, Understanding the Mathematical Way of Thinking - the Registers of Semiotic
Representations, Springer, Cham, Switzerland, 2017.
[10] K.A. Ericsson and H.A. Simon, Verbal reports as data, Psychol rev 87 (1980), pp.
215–251. [11] L. Fiorella and R.E. Mayer, What works and doesn’t work with instructional
video, Comput Hum Behav 89 (2018), pp. 465–470.
[12] M. Franke and S. Ruwisch, Didaktik des Sachrechnens in der Grundschule, 2nd ed.,
Mathe matik Primarstufe und Sekundarstufe I + II, Spektrum Akademischer Verlag,
Heidelberg, 2010.
[13] B. Haas, F. Kesting, R. Martin, Y. Kreis, and V. Koenig, Die Lernplattform MathemaTIC:
Digitales Erlernen der kognitiven Prozesse im Sachrechnen im Alter von 8 bis 10
Jahren, in Beitr¨age zum Mathematikunterricht 2017, U. Kortenkamp and A. Kuzle,
eds., M¨unster. WTM-Verlag, 2017, pp. 1333–1336.
[14] B. Haas, Y. Kreis, V. Koenig, and R. Martin, L’utilisation de l’environnement num´erique
MathemaTIC pour d´evelopper les comp´etences dans le domaine de la r´esolution de
probl`emes arithm´etiques, Paris, France. 2016. Available at
https://evaluationmaths.sciencesconf.org/conference/evaluationmaths/
pages/Axe2_Recherche_Ben_Haas_Koenig_Kreis_Martin.pdf.
[15] B. Haas, Y. Kreis, and Z. Lavicza, Fostering process skills with the educational
technology software MathemaTIC in elementary schools, in Proceedings of the 10th
ERME TOPIC CONFERENCE (ETC10) on Mathematics Education in the Digi tal Age
(MEDA), A. Donevska-Todorova, E. Faggiano, J. Trgalova, Z. Lavicza, R. Weinhandl,
A. Clark-Wilson, and H.G. Weigand, eds., Linz, Austria. Johannes Ke-

71
18.05.2021Haas Ben
pler University, 2020, pp. 199–206. Available at https://drive.google.com/file/d/
1onuTtshYC9SzkaswGT4twtM2QE0yWpsX/view.
[16] S. Kalyuga and J. Sweller, Measuring Knowledge to Optimize Cognitive Load Factors
During Instruction., J Educ Psychol 96 (2004), pp. 558–568.
[17] A. Kassambara, ggpubr: ’ggplot2’ Based Publication Ready Plots (2020). Available at
https://cran.r-project.org/package=ggpubr.
[18] A. Kassambara, rstatix: Pipe-Friendly Framework for Basic Statistical Tests (2020).
Available at https://cran.r-project.org/package=rstatix.
 [19] W. Kintsch, Learning from Text, Cogn Instr 3 (1986), pp. 87–108. [20] K.D. Kubinger,
Psychological Test Calibration Using the Rasch Model - Some Critical Suggestions on
Traditional Approaches, Int J Test 5 (2005), pp. 377–394. [21] M.D. LeBlanc and S.
Weber-Russell, Text integration and mathematical connections: A computer model of
arithmetic word problem solving, Cogn Sci 20 (1996), pp. 357–407. [22] P. Liljedahl, M.
Santos-Trigo, U. Malaspina, and R. Bruder, Problem Solving in Mathe matics Education, in
Problem Solving in Mathematics Education, P. Liljedahl, M. Santos Trigo, U. Malaspina, and
R. Bruder, eds., ICME-13 Topical Surveys, Springer, Cham, 2016, pp. 1–39.
[23] P. Mair, R. Hatzinger, and M.J. Maier, eRm: Extended Rasch Modeling (2020).
Available at https://cran.r-project.org/package=eRm.
[24] R.E. Mayer (ed.), The Cambridge Handbook of Multimedia Learning, 2nd ed.,
Cambridge University Press, New York, 2014.
[25] M. Melo and G.L. Miranda, Learning Electrical Circuits: The Effects of the 4C-ID In
structional Approach in the Acquisition and Transfer of Knowledge, J Inf Technol Educ
Res 14 (2015), pp. 313–337.
[26] MENFP, Loi du 6 f´evrier 2009 portant organisation de l’enseignement fondamental. -
Legilux, no. A20 in M´emorial, Gouvernement du Luxembourg, Luxembourg, 2009,
Avail able at http://legilux.public.lu/eli/etat/leg/loi/2009/02/06/n3/jo.
[27] MENFP, Plan d’´etudes : ´ecole fondamentale, no. Num´ero Sp´ecial in Courrier de
´
l’Education nationale, Minist`ere de l’ Education nationale et de la Forma- ´ tion
professionnelle, Luxembourg, 2011, Available at http://www.men.public.lu/
catalogue-publications/themes-transversaux/cen/cens/plan-etudes/fr.pdf.
[28] G.A. Miller and P.N. Johnson-Laird, Language and Perception, reprint 2014 ed.,
Harvard University Press, Cambridge, MA, 2013.
[29] NCTM, NCTM Principles and Standards for School Mathematics, National Council of
Teachers of Mathematics (NCTM), Reston, VA, 1999.
[30] R. Peterson and J. Herrington, The State of the Art of Design-Based Research, in Pro
ceedings of E-Learn 2005–World Conference on E-Learning in Corporate,
Government, Healthcare, and Higher Education, G. Richards, ed., Vol. 2005,
Vancouver, Canada. Asso ciation for the Advancement of Computing in Education
(AACE), 2005, pp. 2302–2307.
[31] R Core Team, R: A Language and Environment for Statistical Computing (2020). Avail
able at https://www.R-project.org/.
[32] M. Sadoski and A. Paivio, Imagery and Text: A Dual Coding Theory of Reading and
Writing, 2nd ed., Routledge, New York, 2012.
[33] S. Schukajlow, D. Leiss, R. Pekrun, W. Blum, M. M¨uller, and R. Messner, Teaching
methods for modelling problems and students’ task-specific enjoyment, value, interest
and self-efficacy expectations, Educ Stud Math 79 (2012), pp. 215–237.
[34] C. Selter and E. Zannetin, Mathematik unterrichten in der Grundschule: Inhalte -
Leitideen - Beispiele, 2nd ed., Friedrich Verlag, Seelze, 2019.
[35] F.M. Singer, N. Ellerton, and J. Cai, Problem-posing research in mathematics
education: new questions and directions, Educ Stud Math 83 (2013), pp. 1–7.
[36] P. Sonnleitner, C. Kr¨amer, S. Gamo, M. Reichert, C. Muller, U. Keller, and S. Ugen,
Sch¨ulerkompetenzen im L¨angsschnitt: Die Entwicklung von Deutsch-Leseverstehen
und Mathematik in Luxemburg zwischen der 3. und 9. Klasse, in Nationaler
Bildungsbericht Luxemburg 2018, LUCET & SCRIPT, Luxembourg, 2018, pp. 39–58.
Available at https:

72
18.05.2021Haas Ben
//www.bildungsbericht.lu/media/ul_natbericht_de_web_1.4.pdf. [37] L. Trouche,
Environnements Informatis´es et Math´ematiques: quels usages pour quels ap
prentissages?, Educ Stud Math 55 (2004), pp. 181–197.
[38] P. van den Broek, A ’Landscape’ Model of Reading Comprehension: Inferential
Processes and the Construction of a Stable Memory Representation, Can Psychol 36
(1995), pp. 53–54.
[39] J.J.G. van Merri¨enboer and L. Kester, The Four-Component Instructional Design Model
: Multimedia Principles in Environments for Complex Learning, in The Cambridge Hand
book of Multimedia Learning, R. Mayer, ed., Cambridge Handbooks in Psychology,
 Cam bridge University Press, Cambridge, 2005, pp. 71–94.
[40] G. Vergnaud, A classification of cognitive tasks and operations of thought involved in
addition and subtraction problems, in Addition and Subtraction : a cognitive
perspective, T.P. Carpentier, J.M. Moser, and T.A. Romberg, eds., Lawrence Erlbaum
Associates, Hillsdale, NJ, 1982, pp. 39–59.
[41] L. Verschaffel, B. Greer, and E. de Corte, Making sense of word problems, Contexts of
Learning, Swets & Zeitlinger Lisse, Lisse, 2000.
[42] H. Wickham, M. Averick, J. Bryan, W. Chang, L. D’Agostino McGowan, R. Fran¸cois, G.
Grolemund, A. Hayes, L. Henry, J. Hester, M. Kuhn, T. Lin Pedersen, E. Miller, S. Milton
Bache, K. M¨uller, J. Ooms, D. Robinson, D. Paige Seidel, V. Spinu, K. Takahashi, D.
Vaughan, C. Wilke, K. Woo, and H. Yutani, Welcome to the tidyverse, J Open Source
Softw 4 (2019), p. 1686.
[43] E.C. Wittmann and G.N. M¨uller, Das Luxemburger Zahlenbuch 3, Minist`ere de

l’Education nationale, de l’Enfance et de la Jeunesse, Luxembourg, 2016. ´

73
18.05.2021Haas Ben
4.1.2.2. Summary, highlights and next developments
We observed the evolution of student learning with the automated tutoring system MathemaTIC.
Technology-enhanced learning was perceived as motivating for students similarly to many other
studies, for instance, findings of Chun-Yi Lee and Ming-Jang Chen (2015). According to our
results, this could happen because students received explanations and self-paced training with
gamification elements and immediate feedback. Furthermore, working with MathemaTIC
improved even non-German speaking students' performances, and analyses revealed that there
was no differential training effect depending on students’ languages. Our results suggested that
using MathemaTIC could reduce the performance gap caused by differences in language
proficiencies. However, improvements in performances for German-speaking students were far
 more significant. Thus, the connection to the real world through wording (e.g., needed language
skills) seemed still to be a hindrance in making real-world mathematics accessible for every
student. These findings were similar to discussions led by Fuchs et al. (2007) and Boonen et al.
(2013), who signalled the importance of comprehension and morphological language skills for
word problem-solving. Accordingly, technology-enhanced learning could provide new possibilities
compared to more traditional non-technological teachings. Though, with our tasks, we still
observed the importance of written language in the learning process.

In the second study section, we will present two publications on two different experiments with
real-world objects and technology-enhanced mathematical modelling.

4.2. Real-world objects with digital and physical modelling

4.2.1. Case study on augmented reality, digital and physical modelling with
mathematical learning disabilities students in an elementary school in
Luxembourg
Reference:
Haas, B., Kreis, Y., & Lavicza, Z. (In Press). Case study on augmented reality, digital and
physical modelling with mathematical learning disabilities students in an elementary school in
Luxemburg. The International Journal for Technology in Mathematics Education. (Paper C)
4.2.1.1. Short description of the IJTME publication
This publication presents a case study with two elementary school students having mathematical
learning difficulties, using AR and 3D modelling with GeoGebra 3D Graphing Calculator and a
3D printing device in mathematics classes. Both students worked on modelling geometric
shapes (cubes, cuboids, squared pyramids, and octahedrons) and forms (polygons) by
combining real
world objects with AR and creating a copy of geometric shapes or missing parts with a 3D
printer. We will describe designs and outcomes of technology-enhanced tasks based on Dienes’
theory of mathematical learning (Dienes, 1960) and discuss findings on developed process skills
and mathematical concepts.

74
18.05.2021Haas Ben

Case Study on Augmented Reality, Digital and Physical Modelling with

Mathematical Learning Disabilities Students in an Elementary School in Luxemburg
By Ben Haas1, Yves Kreis² and Zsolt Lavicza1
1
Johannes Kepler University, Linz School of Education, Austria, ben.haas@outlook.com
²University of Luxemburg, Luxemburg

Received: 30 October 2020 Revised: 21 December 2020

This paper reports on a case study of two elementary school
students with mathematical learning disabilities (MLD) (ages
10 and 11) using augmented reality (AR), digital and
physical modelling in mathematics class. MLD students
worked on modelling geometric shapes (cubes, cuboids,
squared pyramids, and octahedrons) and forms (polygons) by
combining real-world objects with AR and creating a copy of
geometric shapes or missing parts with a 3D printing device.
The study focused on the development of process skills and
mathematical concepts, tried to identify changes in the
visual-spatial memory, and documented the learning
behaviour in class. Further, we collected data through task
based interviews with both students. Based on our findings,
we present settings and manipulatives which are likely to
foster process skills and mathematical concepts in geometry
tasks suitable for MLD elementary school students.
1 INTRODUCTION
class with MLD students. We choose to conduct a case study,
in this explanatory research setting, with two students (ages
10 and 11) to identify possible learning settings and
manipulatives. In this case study, MLD students used
augmented reality (AR), digital and physical modelling with
the dynamic geometry software GeoGebra 3D Graphing
Calculator (GeoGebra 3D), to perform tasks on geometric
shapes and forms (e.g., visualisation, mathematical
modelling, arithmetic calculations on geometry).

As practitioners and researchers in special needs Based on the collected qualitative data (i.e., task based
education in elementary school, we work regularly with interviews) within a case study approach, we describe in this
students with mathematical learning disabilities (MLD). paper, tasks and observations that we made during the trial.
These students experience difficulties in follow-up Further, we discuss settings and manipulatives which are
mathematics courses (Waiyakoon et al., 2015), show low likely to foster process skills and mathematical concepts in
mastery of mathematical concepts (e.g., written calculation or geometry tasks suitable for MLD students with dominant
visualisation in area and perimeter) and process skills (e.g., weaknesses in visual-spatial and reasoning subtypes, in
mathematical modelling), comparable to APA’s Diagnostic elementary school.
and Statistical Manual of Mental Disorders description of
learning disorders (APA, 2013). According to Fuchs et al. 2 LITERATURE AND THEORETICAL
(2007), external factors or general intelligence cannot FRAMEWORK
explain these difficulties, and so MLD students need specific
teachings in mathematics courses, to develop specific In the beginning of our task development, we
mathematical strategies (e.g., compare mathematical performed a literature review on the use of AR, digital and
concepts) and skills (e.g., use and understand adapted physical modelling in class. Further, we wanted to
materials to perform in mathematics). Further, we observed understand how these technologies could support MLD
that MLD students are less confident in class and tend to quit students in mathematics courses and identify, possible
tasks in mathematics courses often. A metareview on MLD scaffoldings.
highlighted its multidimensionality and Karagiannikis et al.
(2014) identified several subtypes of difficulties (core AR could be defined as virtual information, which is
number, memory, reasoning and visual-spatial). Each MLD displayed, through smartphones or tablets on real-word
student shows strengths and weaknesses among these environments (Cakir & Korkmaz, 2019). A student could see
subtypes. through a smartphone or a tablet a virtual information and
observe its properties in three dimensions. This information,
During our work, we investigated different teaching however, can be presented in various ways. Some
approaches to support MLD students (ages 10-11), male and applications allow simple visualisations (e.g., information
female, in developing strategies to gain in confidence and to which appears while scanning a QR code) or some allow
overcome these difficulties. Those MLD students were modelling of the represented virtual information (e.g.,
mainly struggling in visuo-spatial (e.g., written arithmetic mathematical modelling with geometric shapes in dynamic
calculations in relation with geometry or confusion in geometry software). According to Wu et al. (2013), AR
mathematical symbols or shapes and figures) and reasoning allows to visualize the hidden structure of an object (e.g., a
(e.g., problem-solving, mathematical modelling, geometric shape represented in two dimensions) and
communicating or arguing) subtypes, according to their therefore to render complex issues more understandable. The
special needs diagnose. interactions with educational content through AR are likely
to lower the barrier of computer-based learning for students
With the emergence of new technologies in and support the development of the student’s repertoire of
classrooms (e.g., augmented reality and 3D printing), we
orientated our research to the use of these technologies in
18.05.2021Haas Ben
learning interactions with real-world physical objects (e.g., lead to these learning effects. Students should use AR in
spatial contiguity) (Bujak et al., 2013). prepared settings with clear objectives according to
Billinghurst and Duenser (2012). Thus, tasks such as
In recent studies, we used AR in mathematical discovering the properties of an octahedron, or completing
modelling of real-world objects or buildings in elementary complex figures, should be built up based on mathematical
school teachings (Haas et al., 2020). Moreover, several didactical principles.
studies on AR in mathematics indicated positive learning
effects among students (Bacca Acosta et al., 2019) such as According to Lieban (2019), similar positive effects
increase in motivation or better understanding of properties on learning and understanding geometry are supported
of geometric shapes (González, 2015; Liu et al., 2019). through digital and physical modelling (e.g., creating
Further, AR is likely to improve spatial skills (Martín artefacts with a 3D Printer). Moreover, Ng and Chan (2019)
Gutiérrez et al., 2015). According to Liu et al (2019), the use reported that students learning through play and not through
AR could increase student’s gains in traditional mathematics procedural or formula-driven learning, gain a better
courses. However, simply using AR does not necessarily understanding in mathematics.

In the literature, several findings describe MLD
students’ difficulties in learning abilities and developments
(e.g., visual-spatial ability, mathematical cognition, problem
solving skills, performance perception) in mathematics
(APA, 2013; Fuchs et al., 2007; Geary et al., 2012; Murphy
et al., 2007; Passolunghi & Mammarella, 2012). According
to the classification from Karagiannakis et al. (2014) on
MLD, there are four major subtypes. These subtypes are (a)
core number (e.g., internal representation of quantity), (b)
memory (e.g., working memory, semantic or phonological
memory), (c) reasoning (e.g., decision-making, entailment,
strategic planning, process skills) and (d) visual-spatial
(visuo-spatial working memory/reasoning/perception). As
mentioned earlier, MLD students are not necessarily week
performing in each subtype and depending on the student
there are subtypes where student do well perform. Compared
to our work in special needs, however, most MLD students
showed low performing in reasoning and spatial-visual
subtypes. These low performances could be related to the
teachings and learning they received so far in school, but this
assumption would need further investigations. However, we
observed in our practice among MLD students, with major
difficulties in reasoning and visual-spatial memory, that they
struggle with procedural and standard mathematics class
instructions and tend to lose confidence in their mathematical
performances. Thus, not adapted teaching to their needs
could further increase difficulties and lower motivation to
practice.
According to Szucs et al. (2013), tasks and settings
with MLD students should be adapted and support
developments of visual-spatial and working memory, to
develop the mentioned learning abilities and to increase
performances in mathematics. Therefore, considering the
different findings on MLD students, AR and digital and
physical modelling, we developed technology-enhanced
geometric tasks. We referred to the four basic principles of
Dienes’ theory of mathematics (1960) for the task design
(i.e., dynamic principle, perceptual variability principle,
mathematical variability principle and constructivity
principle). The four principles can be described as follows.
Within the dynamic principle, students learn in playful
settings to discover technologies (e.g., AR, GeoGebra 3D)
and connect basic geometric tasks (e.g., reproducing shapes
and patterns) to technologies. In the perceptual variability
principle, students work on core mathematical concepts, but
with variables changing during the setting (e.g.,
reconstruction cube puzzles in different sizes and dissections,
18.05.2021Haas Ben
standard mathematics tasks instructions. Therefore, we
extended Dienes’ different principles with developed
scaffoldings from AR, digital and physical modelling
educational experiences (Cahyono et al., 2020; Duh &
Klopfer, 2013; Haas et al., 2020; Ng & Chan, 2019). The
amount of instructional guidance by the software and the
Principles Scaffolding
Dynamic Principle Trial and error
High instructional guidance with worked
examples
while using technologies). Within the mathematical
variability principle, students work on the learned
mathematical concepts in different tasks and settings.
However, in this principle, mathematical concepts were
embedded in a more complex learning setting and students
needed to identify concepts during the tasks. In the
constructivity principle, students develop their own concepts
in an open setting (e.g., design new cube puzzles).
Dienes’ concept of learning by abstraction was
discussed critically and some authors suggested that it
needed adaptions or alternatives (Borasi & Borasi, 1984).
Based on the work done by Diego Lieban (2019) with the
four basic principles of Dienes’ theory of mathematics,
digital and physical modelling, we estimated, however, that it
was a reliable structure for our task design to work on
elementary transformation in geometry. Moreover, we
wanted to continue to adapt this theoretical reference with
new technologies, through the present case study.
AR, as mentioned before, could support learning and
engage students in geometry understanding (Thamrongrat &
Law, 2019). Hence, we connected the concept of learning by
abstraction with AR, digital and physical modelling within
the dynamic geometry software GeoGebra 3D (Lavicza et al.,
2020). This software allowed to explore mathematical
modelling with AR and the designs were transferable to a 3D
printing device. Thus, with GeoGebra 3D, we designed tasks
with discoveries of geometric properties (e.g., volume or area
of a cube) through playful trial and error with digital and
physical modelling.
Furthermore, we aimed that the tasks develop process
skills (NCTM, 1999) and be related to real-world
manipulations (Blum, 2013; Freudenthal, 1971). Learning
process skills (i.e., problem solving, reasoning and proof,
representation, and modelling) with real-word manipulations
has been shown to be promising in several studies (Blum &
Leiß, 2007; Schukajlow et al., 2012).
3 TASK AND LEARNING DESIGN
For the task and learning design, based on the four
basic principles of Dienes’ theory of mathematics, we
defined scaffoldings and contents (Table 1). According to
Dede (2008), tasks and learning settings with AR need
structured scaffoldings, to exploit the value added of the
technologies. Moreover, MLD students need these
scaffoldings to understand and develop procedural and
teacher will be larger in the first tasks and decrease along the
study. We aimed to increase peer interactions and
discussions as possible scaffoldings throughout the study, to
develop process skills (e.g., which are essential to develop
for MLD students with difficulties within the subtype
reasoning).
Task content
Soma Cubes and reproducing Shapes
 Perceptual Feedback through software and teacher Identification of shape properties
Variability Reconstruction of given cube puzzles
Principle

Mathematical Discussions about possible solutions Completing cube puzzles and forms
Variability Principle
Calculating area and perimeter of given
polygons

Constructivity Feedback from teacher/peers/parents Construction of different variants of cube

Principle puzzles

Table 1: Task descriptions, scaffolding and Dienes’ principles

and the printing device (i.e., Dynamic Principle and
We started by developing trial and error tasks with Perceptual Variability Principle). In the first learning session,
Piet Hein’s Soma Cubes (Berlekamp et al., 2003) and with these tasks, high instructional guidance was given to
different geometric shape puzzles. We designed the tasks as explain the use and functions (e.g., AR) of the dynamic
worked examples (Sweller et al., 2011), with high geometry software GeoGebra 3D.
instructional guidance in view of their limited working
memory capacity, to familiarise students with the software

Figure 1: Reconstruction of a cube with GeoGebra 3D Graphing Calculator

Later students worked in a more open task setting by
completing AR compositions of geometric shapes and
reconstructing puzzles of geometric shapes through digital
and physical modelling (Figure 1). During the execution of
these tasks, students communicated plans on how to solve the
given task and thus developed their process skills. In these
tasks, student worked essentially on recognizing learned
mathematical concepts within complex tasks mathematical
variability principle).
At the end, MLD students created their designs of
cube puzzles with the digital and physical modelling. In a
peer setting, they needed to complete each other’s puzzle
with AR first and then design the missing parts. MLD
students searched for all possibilities to complete the puzzle.
The puzzle was also given to their parents to solve. They
defined the problem to solve, developed multiple solutions,
and optimised their designs (The Constructivity Principle).
77

The use of AR was not limited to reconstructions of

geometric shapes. MLD students used AR as well in tasks
where they needed to calculate areas and perimeters of a
given polygon. Therefore, based on a textbook item in
18.05.2021Haas Ben
mathematics courses, we printed a polygon with the 3D
printing device and a set of forms. With AR and the
collection of forms, MLD students calculated the area and
perimeter of given polygons (Figure 3).

Figure 2: Measuring area and perimeter of polygons with AR
The different tasks were done in a series of learning
sessions, where we collected data on MLD student’s content
and process skills, reasoning and visual-spatial difficulties
development, and their overall learning motivation. In each
learning sessions, students worked iteratively on the tasks,
similar to the mathematical modelling approach described by
Blum and Leiß (2007). MLD students received a given
problem/question (e.g., complete the cube puzzle), identified
the correct mathematical concept to apply with the use of
AR, performed mathematical calculations/measurement, and
realized a digital and physical modelling. This process was
redone, or at least some parts, when the result was not correct
(e.g., cube puzzle part was not matching) or when
improvements could be done (e.g., create different complex
puzzle parts).
4 METHOD
We choose a case study approach (Stake, 1995) to
investigate the use of the developed tasks with two MLD
students. Further, we aimed to identify developments of
mathematical skills among MLD students through an
analysis of different aspects (performances in process and
content skills, technology use and perception) from the two
participants in the case study.
In our study, we worked with two MLD elementary
Sessio Tasks Student A
n
1&2 Initiation A was intrigued by Soma cubes.
through Soma Neither solving strategies were visible nor
Cubes and self assurance in mathematics.
reproducing A followed the instructional design and
Shapes explanation. A reproduced shapes and soma
cubes with AR in GeoGebra 3D.
school students, one female aged 10 and one male aged 11,
from November 2019 to March 2020 (beginning of the
COVID-19 confinement in Luxemburg). Overall, we had ten
sessions of 55 minutes per MLD student (2 per week). Tasks
were performed over two sessions, due to the time MLD
students needed in modelling and preparing 3D printings.
Both students were diagnosed with MLD, with major
difficulties in the reasoning and visual-spatial sub-types,
visited a regular elementary school class and participated in
special needs courses weekly.
To collect data on the use of AR, digital and physical
modelling tasks with MLD students, we choose to perform
task-based interviews (Goldin, 2000). The task-based
interviews, 5 per student (i.e., one per task over two
sessions), allowed us a certain flexibility on inquiry venues
(Goldin, 1997) during interviews and interaction with the
task environment (Maher & Sigley, 2014). Moreover, we
were able to observe when and why students used
technologies and how they proceeded in the solving process.
MLD students, worked on the tasks during two distinct
sessions and the interview recording for one task was done
through both sessions. The questions during the task-based
interviews were designed to explore MLD students process
skills in mathematics (NCTM, 1999; Selter & Zannetin,
2019) and content skills (e.g. calculating areas or perimeters,
recognise shape patterns). The interviewer asked students
about planning, modelling, representation, and proofing
process. The questions were similar for each task and were
repeated iteratively during the solving process (e.g., How
could you find out? What could you try next? How did you
get this answer?). The questioning process was repeated
several times, as the mathematical modelling, in this case
with technology on real-world objects, was done in iterative
cycles, before students solved the tasks (Blum, 2013).
During the tasks and learning settings, we collected
data from the different interviews and performed a data
analysis.
5 RESULTS AND DISCUSSION
In Table 2, we summarise the results from task-based
interviews during the different sessions. Both students
(marked as A and B) responded to the interviews during the
tasks. During the sessions, MLD students had a tablet (with
GeoGebra 3D), a computer (3D printing device software) a
3D printing device, ruler, scissors, and papers at their
disposal.
Student B
B tried to solve the Soma Cube in multiple
attempts but could not solve them.
B stopped solving and asked for support.
B felt discouraged.
B followed the instructional design and
explanation. B reproduced shapes and soma
cubes with AR in GeoGebra 3D.
3&4 Reconstruction A asked to use GeoGebra 3D for the solving B tried to solve the tasks with GeoGebra 3D and
of different process. A tried to explain and proof parts of the asked to represent using a ruler, scissors and
cubes solving process. papers. B stopped and asked to use GeoGebra 3D
to continue

18.05.2021Haas Ben

A solved the task. the solving process.

B asked for guidance and reexplanation on how to
use AR.
B solved the task.

5&6 Completing A explained how a cube could be completed by B used GeoGebra 3D to complete the
cube puzzles using GeoGebra 3D. cube. B asked re explanations on how
A used the AR function. to use AR. B completed 3 out of 6
A completed 4 out of 6 puzzles. puzzles.
A modulated missing parts with GeoGebra B modulated missing parts.
3D. A modulated larger parts. B modulated many small parts.
A proofed the answers with 3D printing. B proved his results within GeoGebra 3D.

7&8 Area and A tried to solve the task first with a ruler. B explained how to measure the given polygons
perimeter of A draws a solving scheme based on the AR with AR and GeoGebra 3D.
polygons function. A calculated area and perimeter of B explained the difference between area
different polygons. A proofed results with AR and perimeter with AR visually.
function. B recognized forms in the given polygons. B
proofed his results with a second measurement
with a ruler.

9& Construction A explained the design process and essential B explained the design process and essential
10 of different steps. A created two variants of cube puzzles (one steps. B created one variant of cube puzzles
variants of with five parts and one with nine parts). (one with 6 equal parts).
cube puzzles A printed the cube puzzles. B printed the cube puzzles
A explained how to solve the cube puzzles. B explained how to solve the cube puzzles.

Table 2: Summarized results from task-based interviews

of AR, digital and physical modelling, increased motivation
Both MLD students showed different solving and thus triggered a higher understanding of learning. With
behaviours from session one and two on “Initiation and the technology, MLD students tried out different and new
Soma Cubes and reproducing Shapes”. We observed strategies. At some point, both students started to plan the
throughout the tasks, that A seemed to be more confident to solving process, with the use of the GeoGebra 3D. This
use educational technology and used it more readily for change within the task solving process, could be an indicator
planning, modelling, representing, and proofing. B, tended to of improvement within the MLD subtype reasoning. It is
use traditional tools during the tasks, but could not perform a possible that the task or the established routine led MLD
structured solving with a reliable result without technology in students to perform with a structured solving process.
the tasks. However, as the case study was limited to two students and
we did not propose other tasks (without technologies or
Similar to Szucs et al. (2013), solving behaviours in without the scaffoldings), this change was maybe only
visual-spatial tasks could be related to the working memory related to the nature of the task and learning setting.
of MLD students and differ from one to another. A and B The two MLD students showed a gain in confidence
seemed to have different degrees of severity in the visual and enjoyment during the tasks and learning settings, which
spatial subtype of MLD, which was observable during the could be a change factor in perception of their own
first tasks. Further, we observed that B was less confident in mathematics performances (Tapia & Marsh, 2004).
mathematics than A (e.g., B said it was impossible to solve However, as mentioned earlier, this change of perception
because of bad performances in mathematics). The lack of could be strongly related to the tasks and learning settings
confidence, which although increased for both students itself. Although the two MLD students expressed an increase
during the trial, seemed to be a strong hindrance for MLD on motivation in mathematics, a long-time investigation is
students. Both, in the beginning, wanted to abandon the tasks needed to measure if the positive perception endures in other
with the first problem encountered (visual representation of tasks and learning settings. We will investigate perception
the right Soma Cube side). They seemed to have no changes through these tasks in a future study, once
additional strategies or possibilities to try in a different way. restrictions in elementary school, due to COVID-19, are
However, we observed, like Bacca et al. (2014), that the use lowered.
 Based on the interviews during the tasks (e.g.,
foremost area and perimeter of polygons and construction of
different variants of cube puzzles), we observed that the use
of GeoGebra 3D (e.g., AR function and digital modelling)
became -trough the different tasks- a reliable support for both
MLD students in their solving behaviour.
Both MLD students, with student B using the
GeoGebra 3D on their own, in the end used AR and 3D
printing to support their answers and to perform the solving
process. Comparing the first and the last session, students
seemed at the end to be in a flow, described by
Csikszentmihalyi (2014), where the task was appropriated,
the goal clear, and accepted by both MLD students. The
mathematical modelling with AR, digital and physical
modelling on real-world objects, was experienced by both
students in different ways. A, as mentioned before, was able
to use GeoGebra 3D earlier in the sessions and experimented
with the technology (Figure 3). While completing or creating
cube puzzles, A designed parts composed of multiple small
cubes (e.g., a row of three small cubes and one cube added
on top) for the cube puzzle. B, however, mostly used the
single cube shape for his creative process and only at the end
tried to combine one, two or three small cubes. Moreover, A
was experimenting with the AR functions and digital
modelling (e.g., slice cubes, use colour distinction or
combine shapes within AR).
Figure 3: Creation of missing puzzle parts
The difference in both MLD student’s solving
behaviour was present from the beginning on. Therefore, we
assumed, as mentioned, that the severity in MLD and
previous learning experience could explain the difference.
Nevertheless, for both MLD students, a change in the
modelling process was observed.
The time investment in playful tasks with real-world
objects design seemed to increase mathematical modelling
within a structured approach and supported cognition in
mathematics (Ng & Chan, 2019).We were able to observe
the use and improvement of process skills (e.g.,
representation, modelling, problem solving, and
communication) in combination with the technology. Both
MLD students, with difficulties in reasoning (MLD subtype),
showed process skill developments and expressed several
times during the sessions their new positive perception
towards mathematics using technologies like AR and 3D
printing with GeoGebra 3D.
During the sessions and the tasks, scaffoldings
evolved from strong instructional guidance to peer
discussions or support. Hence, we were able to notice that for
both MLD students, although at different moments, guidance
from the teacher was less needed or requested. Both MLD
students even gave feedback to each other in the final tasks
(mathematical variability principle and constructivity
principle). Beside positive comments on the realisations
(e.g., cube puzzle designs), they used mathematical language
and concept knowledge. Furthermore, they shared knowledge
on AR, digital and physical modelling manipulations (e.g.,
slice AR representations in GeoGebra 3D).
Finally, we noted, which was important for us to
know, that both MLD students seemed to improve their
visual-spatial ability, with the use of AR, digital and physical
modelling. AR supported them in seeing and experiencing
geometric shapes in 3D instead of a traditional 2D paper
representation (González, 2015). The feedback we received
from both MLD students was largely on how they
understood and could more accurately visualize shapes and
forms. Thus, both MLD students asked to use GeoGebra 3D
in the regular mathematics courses on different topics (e.g.,
calculation volume of a bowl in class). However, with the
confinement of COVID-19, we were not able to follow up
this development in regular classes.
6 CONCLUSIONS
Our case study was limited to two MLD students and
in this paper, we described several observations which
seemed to be promising in learning geometry and developing
visual-spatial and reasoning MLD subtypes. With the
dynamic geometry software GeoGebra 3D, both MLD
students developed new process and content skills, which
were likely to be attributed to the tasks and learning settings
they performed. The solving behaviour in geometric tasks
became more structured and the visual-spatial ability seemed
to be developed. Further, both MLD students seemed to gain
in confidence and experience, through a playful approach to
mathematical tasks with enjoyment. This could be changing
their own perception of their mathematical performances.
The technology was accepted and used differently by both
MLD students, foremost in the mathematical modelling
process, which needs to be investigated in further research, to
develop new tasks and adapt scaffoldings. However, we will
need to investigate these first case study findings in other
studies, with more participants, to validate our observations.
We will perform further studies on the use of AR, digital and
physical modelling with MLD students. Therefore, we plan
to do a study with GeoGebra 3D and its AR function in
kindergarden and do a follow-up of both MLD students who
participated in this trial, after the schools reopened.
A broader trial, with more MLD students, could
certainly bring more findings in the future. Moreover, we
wanted to connect the findings in further studies to
performance results in mathematics course tests and observe
teachers with MLD students in regular courses using the
GeoGebra 3D. Our tasks and setting could be tested by
teachers in the upcoming studies.
 R., Graf, S., & Kinshuk, K. (2014). Augmented Reality
ACKNOWLEDGEMENTS Trends in Education: A Systematic Review of Research and
Applications. Journal of Educational Technology & Society,
We would like to thank the students, parents and teachers 17(4), 133–149.
supporting this research.
Bacca Acosta, J. L., Baldiris Navarro, S. M., Fabregat Gesa,
REFERENCES R., & Kinshuk, K. (2019). Framework for designing
motivational augmented reality applications in vocational
APA. (2013). DSM-5: Diagnostic and statistical manual of education and training. Australasian Journal of Educational
mental disorders (5th ed.). American Psychiatric Association Technology, 35(3), 102–117.
(APA). https://doi.org/10.14742/ajet.4182

Bacca Acosta, J. L., Baldiris Navarro, S. M., Fabregat Gesa,

80
18.05.2021Haas Ben