Computer Supported Argumentation Learning: Design of a Learning
Scenario in Academic Writing by Means of a Conjecture Map
Michael Burkhard
1
, Sabine Seufert
1
, Reto Gubelmann
2
, Christina Niklaus
2
and Patcharin Panjaburee
3
1
Institute for Educational Management and Technologies, University of St.Gallen,
St. Jakob-Strasse 21, 9000 St.Gallen, Switzerland
2
Institute of Computer Science, University of St. Gallen, Rosenbergstrasse 30, 9000 St. Gallen, Switzerland
3
Faculty of Education, Khon Kaen University, 123 Thanon Mittraphap, Mueang Khon Kaen, Thailand
Keywords: Argumentation Learning, Academic Writing, Learning Design, Higher Education, Natural Language
Processing, Argument Mining.
Abstract: In academic writing, the competency to argue is important. However, first-year students often have difficulties
to construct good arguments. Advances in natural language processing (NLP) have made it possible to better
analyze the writing quality of texts. New tools have emerged which can give students individual feedback on
their texts and the structure of their arguments. While the use of these argumentation learning support tools
can help create better texts, using them in an academic context also carries risks. Learning scenarios are
needed that promote argumentation competency using argumentation tools while also making students aware
of their limitations. To address this issue, this paper investigates how a learning design with an argumentation
learning support tool can be developed to increase the argumentation competency of first-year students. The
conjecture-mapping technique was used, to visualize our assumptions and illustrate the developed learning
design. As part of a first design cycle, the learning design was tested with 80 students in seven academic
writing classes at the University of St.Gallen in Switzerland. Preliminary findings suggest that the learning
design might be helpful to improve the argumentation competency as well as the data-literacy of students (in
relation to argumentation tools). However, further research is necessary to confirm or reject our hypotheses.
1 INTRODUCTION
The competence of being able to argue is important
in everyday life (Scheuer et al., 2010, p. 2) as well as
in a scientific context (Jonassen & Kim, 2010, p.
440). Argumentation provides means by which we
engage in the rational resolution of issues, questions,
disputes, and problem solving (Jonassen & Kim,
2010, p. 439). In academic writing, argumentation is
one of several important competencies to acquire
(Seufert & Spiroudis, 2017, p. 5; Becker-Mrotzek &
Schindler, 2007). However, for students in their first
year, uncertainty appears to be particularly high
because students are still in transition from high
school to university (Vedral & Ederer-Fick, 2015;
Seufert et al., 2021). Students often lack the
requirements to write research papers for academic
writing (Kruse & Chitez, 2014).
To offer students more guidance in developing
their academic writing skills, text production and
feedback on produced texts can be an important
element of teaching (Seufert & Spiroudis, 2017).
However, for the teacher it is often difficult or at least
very time consuming to provide individual feedback
to each student (Jeong et al., 2019).
As a possible solution, since the 1990’s many
software tools have emerged that aim to support
argumentation (see e.g., Scheuer et al., 2010). Such
tools often have the capability to visualize arguments
graphically and point out missing connections
(Scheuer et al., 2010, p. 12). In this way, these tools
can provide individual feedback to each student. Due
to the advances of artificial intelligence (AI) and
natural language processing (NLP) it has become
possible to better analyze the writing quality of texts
(Crossley, 2020). In the context of academic writing,
new support tools have emerged (Rapp & Kauf, 2018;
Strobl et. al., 2019; Burkhard et al., 2022). A very
powerful recent tool is chatGPT (see ChatGPT Pro,
2023) that can compose entire texts and also support
argumentation (if you ask the chatbot to do so). With
ChatGPT, artificial intelligence (AI) has now made
Burkhard, M., Seufert, S., Gubelmann, R., Niklaus, C. and Panjaburee, P.
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map.
DOI: 10.5220/0011984100003470
In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 1, pages 103-114
ISBN: 978-989-758-641-5; ISSN: 2184-5026
Copyright
c
2023 by SCITEPRESS Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
103
its way into education in schools and universities. The
power of this AI tool has led to widespread concerns
that learners are using it to plagiarize assignments by
creating essays or exam papers rather than taking the
time to develop their own arguments (see e.g.,
(Sharples, 2022; Marche, 2022; Heilweil, 2022).
For students to make sense of the complementary
strengths of such tools (and also to better understand
their limitations), the requirements for developing
good arguments are likely to increase. To this end, it
might be useful to develop learning designs that use
argumentation tools to foster argumentation
competency, but at the same time make students
actively aware of the problems and limitations that
come with the use of these tools. Considering the
identified research desideratum, the following
research question should be addressed:
How can a learning design with an argumentation
learning support tool be developed to increase the
argumentation competency of first-year students?
The objective of the paper at hand is therefore to
develop a learning design by using the argumentation
support tool Artist to foster argumentation
competency of first-year students. The tool Artist is
an adapted version based on the tool Argumentation
Learning developed at the University of St.Gallen by
Wambsganss et al. (2020). The created learning
design was tested with 80 students of the University
of St.Gallen during a first design cycle in the fall term
of 2022. Following an educational design research
(EDR) approach by McKenney and Reeves (2018),
the goal is to contribute to theory and practice
simultaneously.
From a theoretical perspective, the conjectures we
have derived about the learning design can be a
starting point for further research and discussion, as
they highlight the complexity and multiple demands
of technology applications in real classrooms
(compared to a laboratory setting). From a practical
perspective, the paper can serve as a guideline for
other researchers who want to implement similar
projects and explore the potential of the technology in
more detail. It further contributes to a better
understanding on how argumentation learning can be
designed and implemented in the context of academic
writing.
To this end, the paper is structured as followed:
Section 2 lays the foundation for our design by
elaborating on the theoretical background of
argumentation competency for academic writing and
how it can be fostered. Section 3 provides information
about the applied research design and the methods
used. Section 4 describes the learning design to foster
argumentation competency of first-semester students.
Section 5 gives insights into the testing of the learning
design and critically reflects on the chosen approach.
Section 6 concludes with some final remarks.
2 THEORETICAL
BACKGROUND
2.1 Argumentation Competency
for Academic Writing
Argumentation can be defined “as the valid
combination between claims and premises” (Rapanta
et al., 2013, p. 483). In the philosophy of logic,
"validity" is used in different ways, depending on the
specific relationship between premise(s) and claim
(Gubelmann et al., 2022). With a deductively valid
inference, it is not logically possible that the premise
is true, while the conclusion is wrong. Deductively
valid inferences then divide into inferences that are
deductively valid due to the form of premise(s) and
conclusion. Such formally valid inferences are the
domain of formal logics. Other inferences are
deductively valid due to the content, or meaning, of
premise and claim. They are usually called materially
valid. In addition to deductively valid inferences,
there are defeasible valid ones, where the truth of the
premise(s) gives reason to accept the truth of the
conclusion without guaranteeing it. Many everyday
inferences are of this sort, variously called inductions
or abductions (inferences to the best explanation). An
overview on this terminology is given in Figure 1.
Figure 1: Kinds of valid inferences. Source: Gubelmann et
al. (2022).
For education, argumentation competence is
considered important, because it is associated with
higher-order thinking, helps students to connect
information across contexts, separates relevant from
irrelevant information and increases the ability of
students to explain their knowledge (Rapanta et al.,
2013, p. 484). Being able to argue is important in
everyday life (Scheuer et al., 2010, p. 2) as well as in
a scientific context (Jonassen & Kim, 2010, p. 440).
Fostering argumentative activities incorporated in
CSEDU 2023 - 15th International Conference on Computer Supported Education
104
learning environments support productive thinking as
well as conceptual change (Jonassen & Kim, 2010, p.
439). In the context of academic writing,
argumentation is an important though not the only
competency to acquire (Seufert & Spiroudis, 2017,
p. 5; Becker-Mrotzek & Schindler, 2007). By
explicitly supporting their claims with premises,
students examine and reveal their assumptions about
knowledge domains, which likely leads to a more
relativistic, differentiated view (Jonassen & Kim,
2010, p. 440).
In education, argumentation can be approached
from two different perspectives: 1) the arguing to
learn approach and 2) the learning to argue
approach (Jonassen & Kim, 2010; Rapanta et al.,
2013, p. 486). In the 1) arguing to learn approach,
“learning emerges as a natural result of an
argumentative intervention” (Rapanta et al., 2013, p.
486). An example for arguing to learn in the context
of academic writing would be, if students peer-review
each other’s drafts, critically discuss and argue about
their texts and learn from that interaction. In the 2)
learning to argue approach, this relationship is
reversed. In this approach, the focus lies on
argumentation itself, how it can be fostered as well as
its benefits. An example for learning to argue in the
context of academic writing would be if students
learn more about the logic of text structures in order
to create more convincing arguments. This paper
investigates the topic from the second perspective and
adopts a 2) learning to argue approach.
Rapanta et al. (2013, pp. 489-491) further
distinguish between 1) argument as form, 2)
argument as strategy and 3) argument as goal. From
the perspective of 1) argument as form, arguments are
primarily investigated as products consisting of
different forms of premise-claim statements. From
the perspective of 2) argument as strategy, the focus
of interest lies “in the procedure of the argument
exchange” (Rapanta et al., 2013, p. 491). As
arguments are often embedded in a dialogical context,
arguments are analyzed from a strategic view based
on different argumentative moves (Rapanta et al.,
2013, p. 490). Finally, from the perspective of 3)
argument as goal, the focus of interest lies on the
overall discursive process, which traditionally has
been persuasion (Walton, 1989; Rapanta et al., 2013,
p. 491). The critical discussion in general or the
negotiation of content to reach consensus might be
other goals of argumentation (Baker, 1999; Rapanta
et al., 2013, p. 491).
In this paper, we primarily adopt the view of 1)
argument as form, which can be represented by
Toulmin’s argument pattern (TAP) (Rapanta et al.,
2013, p. 489). TAP (see Figure 2) is a prominent
model of rhetorical argumentation developed by
Figure 2: Revised version of Toulmin’s argument pattern (TAP). Source: Own representation based on Toulmin (2003, p. 97)
and Amhag (2011, p. 4).
Qualifier (Q)
Related to the claim and indicates the degree
of strength in the claim of using peculiar
comments
Data (D)
Information which the claim is based
(previous research, personal experience,
common sense or statements) and are used
as evidence to support this claim
Claim (C)
Assertions about what exists or the
justification of the norms or values
that people hold or desire for
acceptance of the claim
Warrant (W)
Explicit of implicit argument that explains
the relationship between data and claim
“Because” / “Since”
Backing (B)
Connected directly to the
warrant, with often implicit
motives underlying underwriting
and claims
“Because of” / “On account of”
Rebuttal (R)
Connected to the qualifier with the
statements of facts that either
contradict the claim, data or rebuttal
or qualify an argument
“But”
“Unless
“Therefore”
“Hence”
Mandatory elements
Optional elements
“Probably/ “Maybe”
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map
105
Toulmin in 1958 (Jonassen & Kim, 2010, p. 440).
According to Toulmin (2003, pp. 89-100), the
argumentation model consists of the elements claim
(C), data (D), warrant (W), qualifier (Q), backing (B)
and rebuttal (R). As depicted in Figure 2, an arguer
justifies a claim (C) by a fact (D) which both are linked
through a warrant (W), that explains the relationship
between the fact (D) and the claim (C) (Amhag, 2011,
p. 4). Additional optional elements can be added, such
as the qualifier (Q), which indicates the degree of
strength of the relationship through words such as
“probably” or “maybe” (Amhag, 2011, p. 4). Other
optional elements are the rebuttal (R), which
relativizes existing statements using words such as
“but” or “unless”; as well as the backing (B), which is
linked to the warrant (W) and states further implicit
motives and assumptions (Amhag, 2011, p. 4).
In addition to the TAP, there exist also simplified
argumentation models, which usually only consist of
a claim (C) and one or multiple premises (P) (see e.g.,
Stab & Gurevych, 2014; Wambsganss et al., 2020).
Figure 3 illustrates such a simplified model, where
two premises (in the TAP they were called data (D)
and warrant (W)) are combined to justify the claim
that “Marie Curie is mortal” (see Figure 3).
Our argumentation support tool Artist will rely on
the simplified model based on claims and premises
(see Figure 3) to focus on the most important aspects
and make the system as robust as possible. However,
for the overall learning design, we will use the TAP
(see Figure 2) as a reference framework to highlight
contents that our tool Artist cannot currently cover.
These contents such as the Backing (B) can then be
discussed verbally by the teacher during the learning
scenario to show students the current limitations of
our tool and to sensitize them to other important
aspects of argumentation.
Figure 3: Argument pattern with claim and premise(s).
Source: Own representation based on Toulmin (2003, p.
100) and Stab & Gurevych (2014, p. 1503).
2.2 Fostering Argumentation
Competency
Regarding a learning to argue approach, how can
argumentation competency be fostered? According to
Jonassen and Kim (2010, pp. 444-454), various
methods can be used for developing argumentation
competency in the classroom as well as in other
learning environments.
First, Jonassen and Kim (2010, p. 445) consider it
essential that students engage in meaningful, project-
based or problem-based learning tasks. In their view,
a good learning environment confronts students with
a puzzling claim or solution they must resolve.
Second, counterarguments should be created by
the students in order to better understand opposing
positions and adopt a less self-centred more holistic
perspective of a given topic (Jonassen & Kim, 2010,
p. 445).
Third, scaffolding elements can be used to stimulate
students’ thinking processes by asking topic relevant
questions (Jonassen & Kim, 2010, p. 446). The concept
of scaffolding goes back to the work of Wood et al.
(1976), who defined scaffolding as a “process that
enables a child or novice to solve a problem, carry out
a task or achieve a goal which would be beyond his
unassisted efforts” (Wood et al., 1976, p. 90).
Scaffolding can occur through different channels such
as hints, prompts, illustrations, or the provision of
feedback (Duffy & Azevedo, 2015). Graphical
argumentation aids are widely used (see e.g., Kirschner
et al., 2003) to visualize arguments to improve their
construction (Jonassen & Kim, 2010, p. 448).
In addition, since the 1990’s many software tools
have emerged that aim to support argumentation (see
e.g., Scheuer et al., 2010). Such tools often have the
capability to visualize arguments graphically and
point out missing connections (Scheuer et al., 2010,
p. 12). These tools can be used to support argument
analysis as well as argument generation (Scheuer et
al., 2010, p. 13). Depending on the use case, different
kind of feedback mechanisms such as immediate
system feedback, on-demand feedback, summative
system feedback or moderator-driven feedback may
be appropriate to support the learner (Scheuer et al.,
2010, p. 28).
Due to the advances of artificial intelligence (AI)
and natural language processing (NLP) it has become
possible to better analyze the writing quality of texts
(Crossley, 2020). In the context of academic writing,
new support tools have emerged (Rapp & Kauf, 2018;
Strobl et. al., 2019; Burkhard et al., 2022). For
example, the scientific writing assistant is able to
provide feedback on the overall students’ text
structure (Turunen, 2013). The tool can draw
attention to the fact that certain elements that occur in
the text are not mentioned in specific sections (e.g.,
the abstract); or that some passages (e.g., introduction
section) might be relatively too long or too short in
comparison to the rest of the text (Turunen, 2013).
CSEDU 2023 - 15th International Conference on Computer Supported Education
106
The tool AcaWriter provides students with a
reflective report on their inserted text, highlighting
rhetorical moves that are usually used to construct
convincing texts (Knight et al., 2020; University of
Technology Sydney, 2019). For example, the tool
indicates if background information and previous
literature on the topic seems to be missing; or if the
topic is only treated in a very one-sided way (Knight
et al., 2020, p. 153).
Another tool that directly supports argumentation
is the application Argumentation Learning by the
research group around Wambsganss et al. (2020). In
a first step, students can insert their own texts. The
tool then analyzes the logical structure of the text by
identifying argument components (claims and
premises), as well as the relationships between pairs
of argumentative units (in the same logic as depicted
in Figure 3). Moreover, a set of summary quality
scores are assessed (readability, coherence and
persuasiveness). The results are presented in a
learning dashboard through (i) in-text highlighting of
the argument components; (ii) graph visualization of
the argumentation structure; and (iii) bar-chart
visualization of the three quality dimensions. In that
way, students receive immediate and personalized
feedback that supports them in iteratively improving
their argumentation if needed.
The study by Wambsganss et al. (2020) showed in
a laboratory experiment that students working with
the tool Argumentation Learning were able to write
“more convincing texts with better formal quality of
argumentation” compared to students using a
traditional discussion scripting approach based on
Stegmann et al. (2012) (Wambsganss et al., 2020, p.
1). In addition, design principles related to the design
and development of an argumentation feedback tool
(e.g., to provide the learning tool as a web-based
application, to provide the learning tool with a visual
argumentation and discourse feedback on written or
spoken information) have been worked out by
Wambsganss et al. (2020, p. 5).
Data-rich environments require a certain level of
data-literacy to realize their potential in and out of the
classroom (Wasson et al., 2016). At the same time,
when working with data-rich environments, students
can train their data-literacy competency. Data-literacy
can involve many different aspects such as the analysis
and interpretation of data, understanding problems
when using data or the critical reflection about data in
general (Bonikowska et al., 2019). Only if students can
interpret the data provided by the learning environment
(e.g., data visualization, feedback metrics),
argumentation support tools can develop their full
potential and increase learning gains.
3 RESEARCH
DESIGN & METHODS
In this paper, in the context of a 4-year project funded
by the Swiss National Science Foundation (SNSF),
we will build on the tool Argumentation Learning
created by the research group around Wambsganss et
al. (2020) with the goal of further adapting and
improving it to the context of an actual classroom.
Compared to a laboratory setting, other contextual
factors need to be considered such as the course
syllabus or adjusting the learning content to the level
of the students. The goal is to create a meaningful
teaching and learning scenario using the
argumentative writing tool Artist.
As methodological foundation for the design and
development of the teaching scenario, we follow the
educational design research (EDR) approach by
McKenney and Reeves (2018). EDR has two goals it
simultaneously tries to achieve: On the one hand,
EDR makes contributions to theory as it helps to
improve the theoretical understanding, which e.g., in
the form of guidelines can serve as a building block
for the design of future interventions. On the other
hand, EDR makes also contributions to practice, as it
addresses the problem at hand and provides maturing
interventions (McKenney & Reeves, 2018, p. 86).
EDR consists of three main processes: 1) analysis and
exploration, 2) design and construction, as well as 3)
evaluation and reflection (McKenney & Reeves,
2018, p. 77).
The first EDR process, 1) analysis and exploration
was covered in chapter 2, which introduced important
concepts related to argumentation competency for
academic writing and how it can be fostered. In
addition to that, features of state-of-the-art writing
tools in higher education were analyzed and compared,
whose findings have been published in a previous
paper (see Burkhard et al., 2022).
Building on this knowledge, the second EDR
process 2) design and exploration will be addressed in
chapter 4. In this chapter, the design of the teaching and
learning scenario with the argumentative writing
support tool Artist will be presented. To illustrate our
assumptions, structures, processes, and the expected
dependencies, we will use the conjecture mapping
technique by Sandoval (2014), which can be used to
conceptualize educational design research (see e.g.,
Moser et al., 2021; Boelens et al., 2020; Wozniak,
2015).
The third EDR process 3) evaluation and
reflection is discussed in chapter 5. In this chapter, the
take aways and lessons learned from an initial testing
of the designed learning scenario in seven academic
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map
107
writing classes (in total 80 first semester students) at
the University of St.Gallen will be described.
4 ARTIFACT DESCRIPTION:
LEARNING DESIGN TO
FOSTER ARGUMENTATION
COMPETENCY
Figure 4 shows the conjecture map of the designed
artifact, a learning design to foster argumentation
competency of first-semester university students.
From left to right, Figure 4 is arranged into the four
components high level conjectures (what are the
overall assumptions?), embodiment (what materials,
tasks and structures are needed for the learning
design?), mediating processes (how does
embodiment lead to observable interactions and
artifacts?) as well as outcomes (what are the desired
learning outcomes?).
4.1 High Level Conjectures
In the previous sections 1 and 2, we have already
described the high level conjectures necessary for our
scenario. Based on the overall goal to develop
argumentation competency for academic writing (I.),
we use a problem-based learning approach (II.) as
well as a computer-supported argumentation learning
tool (Artist) (III.) to foster learning. As important
design principles, we rely on the learning-to-argue-
approach (V.) (see Jonassen & Kim, 2010) as well as
on elements that have been found to characterize
good argumentation learning tools (VI.) (see Scheuer
et al., 2010; Wambsganss et al., 2020, p. 5).
4.2 Embodiment
Regarding the embodiment of the designed learning
scenario, students have access to the web-based
learning tool Artist (1). Figure 5 shows the user
interface of the learning tool. Among other things,
students can load predefined examples or generate
their own texts. By clicking a button, the
argumentative discourse structure of students' text is
mined (using pre-trained classifiers) and the scores
for the quality dimensions are calculated. Based on
the results of the argument analysis process, the
student receives visual feedback through a graphical
representation of their arguments. In addition,
students are provided with lecture slides (2) depicting
key argumentation concepts similar to the one in
Figure 3.
Figure 4: Conjecture map of the designed learning scenario.
CSEDU 2023 - 15th International Conference on Computer Supported Education
108
Figure 5: User interface of the learning tool Artist.
Note: The tool Artist is an adapted version based on the tool Argumentation Learning developed at the University of St.Gallen
by Wambsganss et al. (2020). The tool Artist was tested with German-speaking students in the German language. For the
purpose of illustration, Figure 5 is displayed in English.
Based on the provided lecture slides (2), the
teacher in a first step clarifies fundamental concepts
(3) such as the distinction between claims and
premises or the use of indicators to construct
arguments (e.g., through words like “because”,
“therefore” etc.). We consider this step important to
bring everyone on the same level and to avoid
conceptual misunderstandings. After that, students
are advised to use Artist. In the sense of a problem-
based learning approach (II), the students are
confronted with multiple broken examples that are
not working properly inside the Artist tool (4).
Students are given the task of correcting and
improving the incorrect examples by analyzing the
examples with the tool and adjusting them as they see
fit. While correcting the flawed examples, students
are required to back up both arguments as well as
counterarguments on the same topic in order to adopt
a less self-centred and more holistic perspective.
Figure 6 shows such a broken example (left side of
Figure 6), that had to be fixed (right side of Figure 6).
After students have become familiar with the
Artist tool by solving multiple predefined examples
on a given topic (4), the students are given the task of
creating their own conclusion on the topic and
justifying it (5). After that, a classroom discussion
between the teacher and the students takes place,
where the experiences made with Artist are critically
reflected and limitations pointed out (6). Overall, the
embodiment (see Figure 4) can be characterized by a
mixture of individual exercises with Artist as well as
classroom discussions (7), in which the teacher takes
on the role of an instructor and moderator of
classroom interaction (8).
4.3 Mediating Processes
As mediating processes (see Figure 4), we can
observe the students' use of the Artist dashboard (a.),
how they experiment with strategies to fix the broken
examples (b. & e.) as well as the generation of their
own texts inside the Artist tool (c.) to improve their
text quality (g.). In addition, the teacher can observe
student motivation (c.) (e.g., during classroom
discussions) as well as student autonomy (d.) (e.g.,
measured by how many times students need
assistance while working with Artist).
4.4 Outcomes
Regarding the outcomes (see Figure 4), on the one
hand, we expect that students have improved
argumentation competency. Because students are
guided in their argument creation by the Artist tool,
we expect them to create arguments with better
quality (i.). In addition, as during the classroom
discussions experiences made with the tool are
critically reflected and argumentative concepts
investigated, we expect that students will have an
improved content knowledge about argumentation
(ii.). On the other hand, students may also have
improved data-literacy due to the participation in the
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map
109
Figure 6: Problem-based learning environment: fixing broken examples.
Note: The tool Artist was tested with German-speaking students in the German language. For the purpose of illustration,
Figure 6 is displayed in English.
learning scenario. By working with the tool as well as
participating in the classroom discussions, students
get a better understanding of the capabilities and
limitations of text analysis (iii.) and improve their
understanding on how to interpret data (iv.) (e.g.,
understanding data visualization and feedback
metrics in general, applying them to their own text).
5 EVALUATION & REFLECTION
As part of an initial EDR design cycle, the learning
design described in the previous section was tested in
the fall term of 2022 at the University of St. Gallen
with seven academic writing classes. The learning
design was tested with German-speaking students in
the German language. Since a total of 80 students
participated in the learning design, this corresponds
to a class size of 10-15 students per class.
In a first phase, the teacher clarified fundamental
argumentation concepts. For this purpose, lecture
slides were used. Students learned about the
difference between claims and premises as well as
indicators (e.g., “because”, “as a result”) to construct
arguments. Since students in the first semester are
very heterogeneous in terms of their prior knowledge,
this approach seemed meaningful to establish a
common ground (e.g., regarding the terminology used
to describe arguments). During this phase of around
five to ten minutes, students seemed motivated and
had only few comprehension questions.
In a second phase, the teacher shared the link to
Artist. The students were given the task to solve
within Artist the predefined examples about the topic
of “public surveillance”. In the process, students had
to correct and solve one example related to the pro-
arguments and one example related to the contra
arguments. After that, students were given the task by
the teacher to generate with Artist their own
conclusion about the topic and to justify it. During
this phase of around ten minutes, the teacher walked
around the classroom and observed the students'
CSEDU 2023 - 15th International Conference on Computer Supported Education
110
behavior. Because all examples (as well as sample
solutions if needed) were directly available in Artist,
students could work independently and had only few,
if any, questions.
In a third phase, students were asked to complete
a short survey to stimulate reflection on the tool usage
and to get formative feedback for future improvement
of the Artist tool. In three open ended questions,
students were asked 1) what they liked about the tool,
2) what features should be improved, and 3) to state
reasons why they would or would not use the tool
Artist as a support to write their own seminar paper.
Overall, the received feedback was mixed. 25 out
of 80 students (31%) found the tool useful and would
like to use it for writing their own seminar paper.
These students thought the tool provided valuable
feedback. 15 out of 80 students (19%) would not use
the tool. These students mostly stated that they found
the tool confusing in general, or that they were able
to construct better arguments themselves in a more
time efficient way without using the tool. 40 out of 80
students (50%) were still undecided and answered
with maybe or probably. These students often thought
that the tool could be valuable to get a second opinion,
but that they sometimes had difficulties to understand
why the tool made certain recommendations. Due to
that fact, they found it difficult to fully rely on it.
After the survey, a classroom discussion took
place, where the experiences made with Artist were
critically reflected and limitations pointed out by the
teacher. On the one hand, working with Artist can be
valuable for first-year students, as it helps to
understand the basic concepts of argumentation such
as claims and premises. It also encourages thinking
about meaningful text structure as well as the use of
indicators to make an argument explicit. On the other
hand, argumentation is often much more complex as
it involves more than just claims and premises (see
Toulmin’s argument pattern in Figure 2). For
example, arguments often involve further backing
(B), implicit motives underlying the premises and
claims (Amhag, 2011, p. 4). Such implicit motives are
often not mentioned, or one is not even explicitly
aware of them. Particularly first-year students have
trouble understanding their own positionality and
thus their implicit assumptions about the world
(Holmes, 2020). Writing positionality statements
with students may be helpful to make implicit
assumptions about the world explicit (Robinson &
Wilson, 2022, pp. 10-16) to adopt a less self-centric,
more holistic argumentation perspective. Such
content is currently not included in Artist, but could
be added to the learning environment in a next step.
As an additional limitation of the tool, the applied
machine learning approach used by the tool to create
the feedback recommendations was mentioned to the
students. Since machine learning approaches today
are often a black-box, it is difficult or even impossible
to interpret why certain recommendations were made
(Zornoza, 2020). Even though it is explained within
Artist how the displayed metrics are (roughly)
calculated and therefore attempted to create a certain
transparency, not every recommendation made by the
tool is comprehensible down to the last detail due to
the applied machine learning approach. Therefore,
students must critically question the
recommendations they receive by the tool and
strengthen in this way their data-literacy competency.
Figure 7: The three different phases of the learning
scenario.
Overall, we believe that the testing of the learning
design as part of a first EDR design cycle has been
mostly successful. However, our designed learning
scenario is subject to several limitations. First, the
learning design (and its tool Artist) is still in a
development phase and has therefore been tested with
only seven classes whose teachers possessed a certain
affinity for technology and were already familiar with
the interface of the Artist tool. To test external
validity, a larger sample size would be desirable. In
handling Artist, additional new teachers may need to
be instructed. Second, students only worked with
Artist for a relatively short period of time of around
ten to fifteen minutes because the tool was used in the
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map
111
context of a normal classroom lesson to learn basic
argumentation competency. In a next step, it would
be interesting to investigate the use over a longer time
period and with longer texts to see if Artist is not only
helpful for learning basic argumentation competency
but can support students also in their daily text writing
(e.g., for writing a seminar paper). Third, no control
group design was used, making it difficult to draw a
definitive conclusion about the learning outcomes
achieved. However, consistent with the EDR
approach and the goal of obtaining formative
feedback as part of an initial design cycle, this
limitation was deliberately accepted.
In a next step, the goal will be to integrate more
elements of the learning scenario into the Artist tool.
For example, the clarification of fundamental
argumentation concepts (3), undertaken by a teacher
in our learning scenario, could be outsourced directly
to Artist as part of an enhanced onboarding process.
In the sense of a self-learning environment, this
would allow students to work more independently
with Artist.
5 CONCLUSION & OUTLOOK
This paper investigated how learning designs with
argumentation learning support systems can be
developed to increase argumentation competency of
first-year university students. Building on literature
about argumentation competency and how it can be
fostered (see section 2), the conjecture mapping
technique of Sandoval (2014) was used, to illustrate
our assumptions as well as the expected conjectures.
The designed learning scenario has the dual goal of
fostering argumentation competency as well as data-
literacy of students. Although the preliminary
feedback received from the classes is promising,
further iterative EDR cycles of development are
needed to improve our learning design and to evaluate
it for its learning effects.
From a theoretical perspective, the conjectures we
have derived about the learning design can be a
starting point for further research and discussion, as
they highlight the complexity and multiple demands
of technology applications in real classrooms
(compared to a laboratory setting). From a practical
perspective, the paper can serve as a guideline for
other researchers who want to implement similar
projects and explore the potential of the technology in
more detail. It further contributes to a better
understanding on how argumentation learning can be
designed and implemented in the context of academic
writing.
The designed learning scenario with Artist shows
that writing tools can be used to support and relieve
the teacher in the classroom. While using digital tools
for education does not mean that fewer teachers are
needed (Dillenbourg, 2016), the role of the teacher
may evolve and change.
Although writing and argumentation tools can
support us in our writing and even are able to create
whole texts for us (such as chatGPT), argumentation
competency will in our view remain of critical
importance. Only if we understand what determines
good arguments and can critically reflect on them, we
will be able to make sense of the recommendations of
such tools and adapt them to our needs. The GPT-3
language model (underlying model of ChatGPT)
provides developers with a playground for prototypes
to create training systems like Artist. However, once
GPT-3 is used for an extended period of time, usage
fees apply. The vision for the use of AI in academic
writing could be to build an ecosystem of available
tools for students and teachers in a digitally protected
educational space. However, it is still an open
question whether it makes sense to work with and
build upon GPT-3 or to continue to use and develop
smaller, open-source language models for this
targeted purpose of argumentation.
ACKNOWLEDGEMENTS
We would like to thank the Swiss National Science
Foundation (SNSF) for the grant to support our
project Next Generation of Digital Support for
Fostering Students’ Academic Writing Skills: A
Learning Support System based on Machine Learning
(ML), a collaboration project between the University
of St.Gallen in Switzerland and the Mahidol
University in Thailand.
We further would like to thank the research team
around Thiemo Wambsganss, Christina Niklaus,
Matthias Cetto, Matthias Söllner, Siegfried
Handschuh and Jan Marco Leimeister, who created
the tool Argumentation Learning on which our tool
Artist is based on.
REFERENCES
Amhag, L. (2011). Students’ argument patterns in
asynchronous dialogues for learning. Research
highlights in technology and teacher education, 3, 137-
144.
Baker, M. J. (1999). Argumentation and constructive
interaction. Foundations of argumentative text
CSEDU 2023 - 15th International Conference on Computer Supported Education
112
processing, 5, 179-202. University of Amsterdam
Press.
Becker-Mrotzek, M., & Schindler, K. (2007). Texte
schreiben. Kölner Beiträge zur Sprachdidaktik, 7-26.
Gilles & Francke Verlag.
Boelens, R., De Wever, B., & McKenney, S. (2020).
Conjecture mapping to support vocationally educated
adult learners in open-ended tasks. Journal of the
Learning Sciences, 29(3), 430-470. https://doi.org/
10.1080/10508406.2020.1759605
Bonikowska, A., Sanmartin, C., & Frenette, M. (2019).
Data literacy: What it is and how to measure it in the
public service. Statistics Canada, Analytical Studies
Branch. https://www150.statcan.gc.ca/n1/pub/11-633-
x/11-633-x2019003-eng.htm
Burkhard, M., Seufert, S., Panjaburee, P., Pichitpornchai,
C., & Niklaus, C. (2022). Micro-and Macro-Level
Features of NLP-Based Writing Tools in Higher
Education. Proceedings of the 30th International
Conference on Computers in Education (ICCE 2022).
Kuala Lumpur, Malaysia.
ChatGPT Pro. (2023, January 23). ChatGPT: The Most
Advanced AI Chatbot in 2022. https://chatgpt.pro/
Crossley, S. A. (2020). Linguistic features in writing quality
and development: An overview. Journal of Writing
Research, 11(3), 415-443. https://doi.org/10.17239/
jowr-2020.11.03.01
Dillenbourg, P. (2016). The evolution of research on digital
education. International Journal of Artificial
Intelligence in Education, 26(2), 544-560.
https://doi.org/10.1007/s40593-016-0106-z
Duffy, M. C., & Azevedo, R. (2015). Motivation matters:
Interactions between achievement goals and agent
scaffolding for self-regulated learning within an
intelligent tutoring system. Computers in Human
Behavior, 52, 338-348. https://doi.org/10.1016/j.chb.
2015.05.041
Gubelmann, R., Niklaus, C., & Handschuh, S. (2022). A
Philosophically-Informed Contribution to the
Generalization Problem of Neural Natural Language
Inference: Shallow Heuristics, Bias, and the Varieties
of Inference. In Proceedings of the 3rd Natural Logic
Meets Machine Learning Workshop (NALOMA III), 38-
50, Galway, Ireland. Association for Computational
Linguistics..
Heilweil, R. (2022, December 7). AI is finally good at stuff,
and that’s a problem. Vox. https://www.vox.com/
recode/2022/12/7/23498694/ai-artificial-intelligence-
chat-gpt-openai
Holmes, A. G. D. (2020). Researcher Positionality--A
Consideration of Its Influence and Place in Qualitative
Research--A New Researcher Guide. Shanlax
International Journal of Education, 8(4), 1-10.
Jeong, H., Hmelo-Silver, C. E., & Jo, K. (2019). Ten years
of Computer-Supported Collaborative Learning: A
meta-analysis of CSCL in STEM education during
2005–2014. Educational research review, 28, 100284.
https://doi.org/10.1016/j.edurev.2019.100284
Jonassen, D. H., & Kim, B. (2010). Arguing to learn and
learning to argue: Design justifications and guidelines.
Educational Technology Research and Development,
58(4), 439-457. https://doi.org/10.1007/s11423-009-
9143-8
Kirschner, P. A., Buckingham-Shum, S. J., & Carr, C. S.
(2003). Visualizing argumentation: Software tools for
collaborative and educational sense-making. Springer.
Knight, S., Shibani, A., Abel, S., Gibson, A., Ryan, P.,
Sutton, N., Wight, R., Lucas, C., Sándor, Á., Kitto, K.,
Liu, M., Vijay Mogarkar, R., & Buckingham Shum, S.
(2020). Acawriter: A learning analytics tool for
formative feedback on academic writing. Journal of
Writing Research, 12(1), 141-186. https://doi.org/10.17
239/jowr-2020.12.01.06
Kruse, O., & Chitez, M. (2014). Schreibkompetenz im
Studium: Komponenten, Modelle, Messung. In
Dreyfürst, S. & Sennewald, N. (Eds.). Grundlagentexte
zur Theorie, Didaktik und Beratung, 107-126. Verlag
Barbara Budrich.
Marche, S. (2022, December 6). The College Essay Is
Dead. The Atlantic. https://www.theatlantic.com/
technology/archive/2022/12/chatgpt-ai-writing-college
-student-essays/672371/
McKenney, S., & Reeves, T. (2018). Conducting
educational design research (2nd ed.) Routledge:
Oxon.
Moser, L., Guggemos, J., & Seufert, S. (2021). Improving
a MOOC to foster information literacy by means of a
conjecture map. International Journal of Learning
Technology, 16(1), 65-86. https://doi.org/10.1504/
IJLT.2021.115470
Rapanta, C., Garcia-Mila, M., & Gilabert, S. (2013). What
is meant by argumentative competence? An integrative
review of methods of analysis and assessment in
education. Review of Educational Research, 83(4), 483-
520. https://doi.org/10.3102/0034654313487606
Rapp, C., & Kauf, P. (2018). Scaling Academic Writing
Instruction: Evaluation of a Scaffolding Tool (Thesis
Writer). International Journal of Artificial Intelligence
in Education, 28, 590–615. https://doi.org/10.1007/
s40593-017-0162-z
Robinson, O., & Wilson, A. (2022). Practicing and
Presenting Social Research. The University of British
Columbia, Open UBC. https://pressbooks.bc
campus.ca/undergradresearch/
Sandoval, W. (2014). Conjecture Mapping: An Approach
to Systematic Educational Design Research, Journal of
the Learning Sciences, 23(1), 18-36.
https://doi.org/10.1080/10508406.2013.778204
Scheuer, O., Loll, F., Pinkwart, N., & McLaren, B. M.
(2010). Computer-supported argumentation: A review
of the state of the art. International Journal of
Computer-supported collaborative learning, 5(1), 43-
102. https://doi.org/10.1007/s11412-009-9080-x
Seufert, S., Burkhard, M., & Handschuh, S. (2021).
Fostering Students' Academic Writing Skills: Feedback
Model for an AI-enabled Support Environment. In T.
Bastiaens (Ed.), Proceedings of Innovate Learning
Summit 2021, 49-58. Online, United States: Association
for the Advancement of Computing in Education
Computer Supported Argumentation Learning: Design of a Learning Scenario in Academic Writing by Means of a Conjecture Map
113
(AACE). https://www.learntechlib.org/primary/p/220
270/
Seufert, S., & Spiroudis, E. (2017). Wissenschaftliche
Schreibkompetenzen in der Studieneingangsphase
fördern: Erfahrungen mit einem Rückmelde- und
Bewertungsraster (Rubric). Zeitschrift Schreiben, 1-9.
Sharples, M. (2022, May 17). New AI tools that can write
student essays require educators to rethink teaching
and assessment. The London School of Economics and
Political Science (LSE). https://blogs.lse.ac.uk/
impactofsocialsciences/2022/05/17/new-ai-tools-that-
can-write-student-essays-require-educators-to-rethink-
teaching-and-assessment/
Stab, C., & Gurevych, I. (2014). Annotating argument
components and relations in persuasive essays.
Proceedings of COLING 2014, the 25th international
conference on computational linguistics: Technical
papers, 1501-1510.
Stegmann, K., Wecker, C., Weinberger, A., & Fischer, F.
(2012). Collaborative argumentation and cognitive
elaboration in a computer-supported collaborative
learning environment. Instructional Science, 40(2),
297-323. https://doi.org/10.1007/s11251-011-9174-5
Strobl, C., Ailhaud, E., Benetos, K., Devitt, A., Kruse, O.,
Proske, A., & Rapp, C. (2019). Digital support for
academic writing: A review of technologies and
pedagogies. Computers & Education, 131, 33–48.
https://doi.org/10.1016/j.compedu.2018.12.005
Toulmin, S. E. (2003). The uses of argument, Updated
Edition. Cambridge University Press.
Turunen, T. P. (2013). Introduction to Scientific Writing
Assistant (SWAN)- Tool for Evaluating the Quality of
Scientific Manuscripts [Master thesis, School of
Computing, University of Eastern Finland]. SWAN.
https://cs.joensuu.fi/swan/publications.html
University of Technology Sydney. (2019, May 11). Using
the AcaWriter platform [Video]. YouTube.
https://www.youtube.com/watch?v=CHNLZrc1YWk
&t=16s
Vedral, J., & Ederer-Fick, E. (2015). Schreibforschung und
Schreibdidaktik: Modelle und Theorien
wissenschaftlichen Schreibens. In Egger, R.,
Wustmann, C., & Karber, A. (Eds.). Forschungs-
geleitete Lehre in einem Massenstudium: Bedingungen
und Möglichkeiten in den Erziehungs- und
Bildungswissenschaften, 217-238. Springer Verlag.
Walton, D. N. (1989). Informal logic: A handbook for
critical argument. Cambridge University Press.
Wasson, B., Hansen, C. & Netteland, G. (2016). Data
literacy and use for learning when using learning
analytics for learners. CEUR Workshop Proceedings,
1596, 38–41. http://hdl.han-dle.net/11250/2429460
Wambsganss, T., Niklaus, C., Cetto, M., Söllner, M.,
Handschuh, S., & Leimeister, J. M. (2020). AL: an
adaptive learning support system for argumentation
skills. Proceedings of the 2020 CHI Conference on
Human Factors in Computing Systems, 1-14.
https://doi.org/10.1145/3313831.3376732
Wood, D., Bruner, J. S., & Ross, G. (1976). The role of
tutoring in problem solving. Journal of Child
Psychology and Psychiatry, and Allied Disciplines,
17(2), 89–100. https://doi.org/10.1111/j.1469-
7610.1976.tb00381.x.
Wozniak, H. (2015). Conjecture mapping to optimize the
educational design research process. Australasian
Journal of Educational Technology, 31(5).
https://doi.org/10.14742/ajet.2505
Zornoza, J. (2020, April 15). Explainable Artificial
Intelligence: Discover one of the biggest trends in
Machine Learning and AI. Towards data science.
https://towardsdatascience.com/explainable-artificial-
intelligence-14944563cc79
CSEDU 2023 - 15th International Conference on Computer Supported Education
114