Symposium
Augmented Reality for Learning and Instruction – Theory-based Research Beyond Media Comparisons
Chair(s): Jule M. Krüger (Universität Potsdam, Digitale Bildung), Kristin Altmeyer (Universität des Saarlandes, Empirische Bildungsforschung)
Discussant(s): Josef Buchner (Pädagogische Hochschule St.Gallen, Institut Digitale und Informatische Bildung)
Augmented reality (AR) is a form of information presentation that includes the combination of virtual and physical elements. The integration of these types of representations can be implemented in unique educational experiences to effectively elicit and support learning processes and lead to improved learning outcomes. With a technological focus, Azuma and colleagues (2001) identified three key features of AR technology as 1) the integration of real and virtual elements, 2) real-time interactivity, and 3) 3D registration. For instructional design, AR systems allow for the integration of instructionally relevant virtual data into physical, real-world environments. For example, instructional information can be embedded into corresponding real-world settings, and authentic 3D models can be embedded into instructional settings. From a learner-based perspective, AR thus allows for an integrated perception of virtual and physical elements for a contextualised experience in physical space, including interaction with combined physical-virtual materials (Krüger et al., 2019).
In recent years, research on AR for learning and instruction has increased a lot due to the technological developments in this area, such as devices for mobile AR. In general, positive effects on cognitive, affective, and behavioural learning outcomes have been identified of AR in comparison to other types of learning experiences and material (Chang et al., 2022; Malone et al., 2023). In a recent systematic review, Buchner and Kerres (2023) found that 80% of studies on AR in education published in the top twelve journals on educational technologies until 2020 use a media-comparison approach. However, this technology-centred approach has been widely criticized concerning its narrow view on instructional design, methodologically flawed research designs, and limited gain in scientific insights. Recently, it has been observed that there is a growing tendency to employ an expanding array of novel study designs to examine the underlying mechanisms that determine the effects of AR. A goal of this symposium is to bring together research from different labs that goes beyond the mere comparison of AR with other forms of information presentation. In line with the suggestions by Buchner and Kerres (2023), these alternative research approaches include value-added designs, which compare different versions of an AR-based learning experience, and learner-treatment-interaction designs, which examine the influence of learner characteristics on learning with AR.
Therefore, the general research question for this symposium does not focus on “if” AR can be used to effectively support learning and instruction but considers the “how” and the “when” of this question.
In the first contribution, a value-added study describes the evaluation of different AR application designs with multiple external representations in physics lab work based on the Cognitive Theory of Multimedia Learning and Cognitive Load Theory.
In the second contribution, a value-added study draws on situated learning to examine the influence of contextualising a learning experience in the medical field, without changing the AR application itself.
The third contribution reviews existing literature and defines a research gap concerning the consideration of learner differences as moderators in research on AR in education and thus the lack of learner-treatment-interaction study designs in the literature.
The fourth contribution explores learner characteristics in interaction with a value-added design on AR-based information placement on corresponding local trees in a setting in nature.
The contributions describe different approaches and insights that can bring the research field closer towards better understanding the how and when of learning and instruction in AR. The discussion of the symposium will focus on the opportunities and challenges in current research on AR in education, discussing the future directions of methods and design in and development of a framework for meaningful educational AR research.
Presentations of the Symposium
Augmented Reality for Visualizing Scientific Models in Physics Lab Work: The Role of (Multiple) Representations
Kristin Altmeyer1, Peter Edelsbrunner2, Barbara Gränz3, Sarah Hofer4, Christoph Hoyer5, Jochen Kuhn5, Zoya Kozlova4, Stefan Küchemann5, Andreas Lichtenberger3, Sarah Malone1, Roman Schmid3, Ralph Schumacher2, Bermann Steinmacher3, Elsbeth Stern2, Andreas Vaterlaus3, Max Warkentin5, Roland Brünken1
1Universität des Saarlandes, Empirische Bildungsforschung, 2ETH Zürich, Lehr- und Lernforschung, 3ETH Zürich, Physik und Ausbildung, 4LMU München, Lehr- und Lernforschung, 5LMU München, Didaktik der Physik
Theoretical Background
Visual-graphic representations of scientific models can support students by visualizing the invisible foundations of observations during scientific experimentation (Olympiou et al., 2013). However, these models are usually shown spatially and temporally separated from observed phenomena (e.g., on a worksheet handed out after lab work), which hampers conceptual understanding (Schroeder & Cenki, 2018). In accordance with the Cognitive Theory of Multimedia Learning (Mayer, 2014) and the Cognitive Load Theory (Sweller, 1998), Augmented Reality (AR) offers the potential to address this issue by enabling the real-time presentation of virtual models in close proximity and adaptive to corresponding real-world phenomena. This resulting contiguity of information can contribute to a reduction in learning-irrelevant cognitive load and improved learning outcomes (Buchner et al., 2021), particularly in the context of physics lab work (e.g., Altmeyer et al., 2020).
Research Objectives
The objective of the present study was to investigate a newly-developed AR-supported experimental learning set-up in which electromagnetic phenomena are superimposed with virtual representations of vector field models (Donhauser et al., 2020). While extensive research has explored the presentation of (graphical) representations in traditional multimedia learning environments (e.g., Rau et al., 2015), the reasonable selection of AR-based representations in lab work settings requires further investigation. Therefore, the current study compares different virtual (multiple) representations in AR to examine their impact on cognitive load and conceptual knowledge acquisition. Indications for the design of effective AR learning environments will be derived.
Method
N=75 students (47% female, Mage=17.02, SD=1.22) participated in the study. In the first part of the study, all students were presented with a learning video, then they completed a representational competence test (covariate, 31 items adapted from Küchemann et al., 2021; α=.89/ ω=.90), watched a second learning video, and completed a pre-test of conceptual knowledge (14 items adapted from Küchemann et al., 2021; α=.90/ ω=.90). The second part of the study occurred within a two-day interval. Applying a between-subjects design, participants were randomly allocated to seven groups. The students conducted five experiments on the Lorentz force while utilizing AR-smartglasses. Depending on their respective group, they were presented with distinct virtual representations including vector fields, field lines, and a tripod, or with any combination of these representations. Finally, the participants completed a cognitive load questionnaire (Thees et al., 2021; 3 items per dimension, α/ω=.46-.96), a post-test of conceptual knowledge (α=.93/ ω=.94), and a usability questionnaire (Brooke, 1996).
Results
Usability across all AR conditions was rated as "good" (Bangor et al., 2009) with very minor and non-significant differences between groups (F(6, 70)=1.54, p>.05). Qualitative feedback indicates that students derived enjoyment from their AR-supported experiments, yet they highlighted opportunities for enhancing the design of virtual representations and reducing the complexity of AR-interaction. Regarding cognitive load (F(18, 184.33)=1.06, η²=.07) and also conceptual knowledge acquisition (F(7, 78)=1.66, η²g=.13), there were minimal and non-significant differences between the AR groups (p>.05). An exploratory analysis unveiled that groups employing the vector tripod either individually or in conjunction with other virtual representations demonstrated more substantial gains in conceptual knowledge acquisition, as opposed to those without access to the tripod (F(1, 84)=3.66, η²g=.08).
To summarize, the findings suggest that all representational iterations of the newly-developed AR-based experimental learning environment hold promise for supporting the acquisition of conceptual knowledge, with the vector tripod being the most helpful representation. Providing various instead of single virtual representations did not appear to foster or impede conceptual knowledge acquisition. Future research will investigate whether engaging with diverse representations promotes the development of representational competence. Planned subsequent research endeavors will also analyze AR-interactions and eye movements to refine the learning environment and gain further insights into AR-enhanced science education.
The Impact of Narrative and Physical Contextualisation on Situated Learning in Augmented Reality
Kevin Palzer1, Jule Krüger2, Daniel Bodemer1
1Universität Duisburg-Essen, Psychologische Forschungsmethoden – Medienbasierte Wissenskonstruktion, 2Universität Potsdam, Digitale Bildung
Theory
Situated learning is recognised as a valuable pedagogical approach for augmented reality (AR) environments (Garzón et al., 2020). Situated cognition, as described by Brown et al. (1989), proposes an inherent connection of learning and its context, distinguishing between school and application context, and Young (1993) suggests situating learning in real-world environments. In combining virtual elements and physical environments, AR can contextualise learning in authentic environments and bring authentic artefacts into instruction (Bower et al., 2014). By embedding learning in a corresponding physical context, AR-based contextuality can support the mental connection and integration of virtual and physical elements (Krüger et al., 2019). This integration can elicit context immersion (Kim, 2013). In order to further support Immersion, a narrative (Nilsson et al., 2016; Reid et al., 2005) and physical artefacts (Reid, 2008) can be used, where narrative and artefacts should be strongly coupled (Georgiou & Kyza, 2021). In the current study, we examine how these factors can induce perceived contextuality in AR and how they might impact learning processes and outcomes. We expect that adding a contextualising physical artefact and application-focused narrative leads to increased perceived contextuality, enjoyment and knowledge. Furthermore, we explore the impact of contextualisation on workload.
Method
A laboratory study was conducted in an experimental 2×2 between-subjects design with narrative (educational vs. application) and physical contextualisation (physical artefact present vs. absent) as factors in a value-added study (see Buchner & Kerres, 2023). N = 40 participants between 18 and 32 years (M = 23.50, SD = 3.52), 24 female and 15 male, were evenly distributed into the four conditions. The topic of the human digestive system was introduced with the narrative describing either an educational (university course) or an application context (medical internship). The AR application on a Microsoft Hololens 2 showed a virtual 3D model of the digestive system, either on its own (physical artefact absent) or projected onto a person (physical artefact present). After the learning phase, participants answered the NASA TLX for workload (Hart, 2006), a scale on intrinsic motivation with a subscale on enjoyment (Wilde et al., 2009) and the ARcis Questionnaire for contextuality (Krüger & Bodemer, 2022). Afterwards, a knowledge test on naming the digestive system components was administered.
Results
The data were analysed with 2×2 ANOVAs. Concerning perceived contextuality, we only found a main effect of physical artefact, with higher contextuality for present artefact, F(1,36) = 14.88, p < .001, ω² = 0.26. Regarding enjoyment, we only found a main effect of narrative, showing increased enjoyment in the application narrative, F(1,36) = 4.60, p = .039, ω² = 0.08. For knowledge, we found a main effect of narrative, F(1,36) = 5.15, p = .029, ω² = 0.09, and an interaction effect, F(1,36) = 4.48, p = .041, ω² = 0.08: highest knowledge in the application narrative with no artefact, and lowest knowledge in the education narrative with no artefact. The explorative analyses of workload (NASA TLX) only showed a main effect of physical contextualisation on frustration, with increased frustration for present artefact, F(1,36) = 4.70, p = .037, ω² = 0.09.
Discussion
We found that a physical artefact in AR facilitated perceived contextuality, in accordance with the ARcis framework (Krüger et al., 2019). Using an application narrative increased enjoyment and knowledge, but the effect on knowledge was diminished by the physical artefact. This might be explained by the increase in learners' frustration, which was unexpected, but may be attributable to the presence of another person as a social stressor leading to increased anxiety and arousal (Kushnir, 1986). More research is necessary, but this study describes first indications of the impact of contextualisation in AR on affective and cognitive variables.
AR Glasses in the Classroom with no Human Learner Behind? A Systematic Literature Review
Zoya Kozlova, Sarah Hofer
LMU München, Lehr- und Lernforschung
Theoretical Background
According to Steffen (2019), most of the activities which augmented reality (AR) enables are either impossible to carry out in the physical world due to the physical laws (e.g., seeing the entire solar system while sitting in the classroom) or too dangerous or inconvenient to perform (e.g., asking a novice surgeon to do a training surgery on a real patient). These two key affordances make AR well-suited for use in STEM subjects teaching, e.g., conducting dangerous experiments in chemistry or visualizing normally invisible magnetic fields. At the same time, affordances as a term of ecological psychology are seen more as a concept defining relations between the environment and the organism (Chemero, 2003). Namely, the environment affords behavior to the organism, so affordances are relations between the abilities of organisms and features of the environment. Affordances of AR then can only be fully exploited when being attributed to a certain organism with specific characteristics.
Research Question
This brought us to evaluate to which extent AR affordances consider the learner as the organism in recent studies. In other words, we seek to explore to which extent the research into the use of AR in STEM education considers the learner characteristics and learning outcomes.
Method
In a systematic review, we derived 667 studies from the two databases Scopus and Web of Science. In our search we used the terms “augmented reality”, “augmenting reality”, and “mixed reality”, coupled with “learning”, “education”, “training”, “teaching”, and “instruction” for the period of 2013-2022. In the first selection phase, we excluded studies done in non-STEM subjects (e.g., art). In the second phase, we identified the studies that examine learner characteristics together with the learning outcomes. We categorized the studies based on the knowledge outcome (declarative and procedural knowledge and the associated cognitive processes – reproduce, transfer, produce), individual characteristics, e.g., spatial ability, prior knowledge, and the AR affordances, e.g., delivering learning content in a 3D perspective, visualizing the invisible etc. In the end, only 24 studies satisfying our inclusion criteria were distinguished.
Results
Apart from the small proportion of the studies addressing individual characteristics (less than 5%), another interesting finding suggests that spatial ability and prior knowledge are the most explored individual characteristics in AR research (9 studies for spatial ability and 2 for prior knowledge). In terms of the learning outcomes, only 21% of the studies tackled procedural knowledge outcomes. The main focus was on reproduction and transfer of declarative knowledge (88% for reproduction and 54% for transfer). The most used AR affordance was presenting learning content in a 3D perspective (21 out of 24 studies).
Discussion
At present, AR technology seems to be seen as a universal tool, suitable for all learners, without sufficient consideration of their unique individual differences. The existing research on individual differences in AR sparkles numerous questions that are yet to be fully explored. For instance, it is still questionable, whether or not AR has a compensatory effect for the students with lower spatial ability or whether higher spatial ability serve as an enhancer in AR learning environment. Other important learner characteristics, such as working memory capacity, have hardly been investigated in AR learning. The findings of this review help to define the future research agenda for AR.
Augmented reality in nature: an exploratory study on the placement of learning content and the impact of learner characteristics
Jule Krüger
Universität Potsdam, Digitale Bildung
Theory
In augmented reality (AR), virtual and physical elements can be combined and presented integrated into one view, which enables the contextualisation of interactive and spatial virtual representations in authentic physical environments without changing the physical surroundings themselves. For example, a natural environment can be enriched by virtual information designed for the purpose of achieving specific learning objectives. This goal-oriented embedding of virtual information in an associated physical environment leverages the AR-specific property of contextuality (see Krüger et al., 2019). This placement of information within a real-world context can play a role in authentic experience, enjoyment, and learning (Bower et al., 2014; Harley et al., 2016; Kamarainen et al., 2013), influencing immersion, motivation, and learning processes (Georgiou & Kyza, 2021; Sylaiou et al., 2010; Weerasinghe et al., 2022). The current study further explores the role of the exact positioning of virtual information in an AR application in nature on cognitive and affective factors, behaviour, and learning outcomes. Placement of instructional information closer to corresponding physical objects is expected to lead to improved cognitive load, presence, positive emotions, knowledge, and different behaviours. Potential moderations of learner characteristics are explored.
Method
In a two-group design, the study took place outside in nature. N = 18 participants, 18 to 33 years old (M = 22.67, SD = 4.35), 16 female, 3 male, were distributed evenly (near vs. far condition). A prior knowledge test and task expectancy questions (Wigfield & Eccles, 2000) were administered. In the learning phase, participants retrieved marker-based information on plants in the environment with a tablet-based AR application, with markers either anchored to corresponding plants (near) or placed further away (far). Cognitive load was measured with a questionnaire (Klepsch et al., 2017), presence with the AR immersion scale (Georgiou & Kyza, 2017), positive emotions with the short Achievement Emotions Questionnaire (Bieleke et al., 2021), and behaviour with self-designed questions. Afterwards, a knowledge test was administered, and the participants were interviewed on their experience.
Results
A mixed 2×2 ANOVA on prior and resulting tree name knowledge showed a positive main effect of time, F(1,16) = 45.90, p < .001, and an interaction effect, F(1,16) = 6.88, p = .018, with a larger increase for near than far placement. Far compared to near placement led learners to look into the environment less often, t(15.52) = 2.48, p = .025, which is also in accordance with the interview data, describing that participants with near placement compared the tablet-based information and the plants. Concerning other variables, no effects of group but significant moderation effects were detected. Only for low and not high prior knowledge learners germane cognitive load, t(14) = 2.50, p = .025, and pride, t(14) = 2.59, p = .021, increased for near placement. Further, only for low and not high task expectancy learners, germane cognitive load, t(14) = 2.30, p = .037, presence, t(14) = 4.14, p = .001, and pride, t(14) = 2.76, p = .015, increased for near placement.
Discussion
The data suggest that proximity of the plants led to a larger increase in tree-specific knowledge and caused implicit bottom-up or stimulus-driven attentional control (see Egeth & Yantis, 1997) towards the plants. Learning scenarios in which virtual and physical elements are to be integrated into a coherent mental model might profit from this. The effect on other factors were contingent on learners’ characteristics, with more positive effects when prior knowledge and expectancy were low. The results in this study support the necessity to examine specific design decisions and their interaction with learner characteristics in AR. Future studies with bigger samples can build on the results of this exploratory study.
|
