Please refer to these proceedings as: Khaleel Asyraaf Mat Sanusi, Bibeg Hang Limbu, Jan Schneider, Daniele Di Mitri, & Roland Klemke: Proceedings of the 2nd International Workshop on Multimodal Immersive Learning Systems. At the 17th European Conference on Technology Enhanced Learning - Addressing Global Challenges and Quality Education. Toulouse, France. September 12-16, 2022. CEUR-WS.org/Vol-3247, ISSN 1613-0073. ©2022 for the individual papers by the papers’ authors. Copying is permitted for private and academic purposes. Re-publication of material from this volume requires permission from the copyright owners. Address of the first editor: Khaleel Asyraaf Mat Sanusi Cologne Game Lab - Cologne University of Applied Sciences (TH Cologne) Schanzenstr. 28, 51063 Cologne, Germany ks@colognegamelab.de Preface The first international interdisciplinary workshop on MILeS (Multimodal Immersive Learning Systems) was organised in the context of the MILKI-PSY (Multimodal Immersive Learning with Artificial Intelligence for Psychomotor Skills) project in 2021 to develop AI-enhanced, data-intensive, MILeS for psychomotor skills development. A year later, the MILeS workshop continues its tradition of excellence, bringing together researchers at the cutting edge in domains such as AI & data science, educational science, sports science etc., by organising the second edition of the MILeS workshop on 13th September 2022 in Toulouse, France. Multimodal immersive technologies such as augmented and virtual reality, along with rapidly evolving technologies such as AI and sensors, are redefining human-computer interaction and, consequently, the domain of technology-enhanced learning. Decades ago, the seminal work. “Bloom's 2 sigma problem”, led to the inception of the potentialities of “educational technology” for personalised learning. This was immediately followed by an influx of researchers from all domains interested in education exploring various technologies in search of this promised land and the holy grail. In recent years, Ericsson with his seminal work on “deliberate practice”, recapitulated much of Bloom's conclusions, giving us an epiphany that not much has changed or been achieved in education. This is ever more so true for psychomotor skills domains that focus on performance outcomes in the real world. Obviously, no skill lies exclusively in one of the three domains of learning but rather in a conceptual three-dimensional space where each domain represents a dimension, as Limbu et al. (2022) highlighted during the workshop. Limbu et al. (2022) shed light on the need for a new framework to view education, which moves away from the traditional view of learning that emphasises the brain as the core learning entity, towards a more embodied and contextual view of learning. MILeS are noteworthy precisely as they promise to situate learning in authentic contexts. As Aristotle argued, how learning should always be an active process. However, facilitating situated learning in authentic contexts requires a shift away from traditional cognitive instructional designs. Cárdenas et al. (2022) state that learning psychomotor skills constitutes not only the physical aspects but also the technical and mental aspects and that additional considerations must be given to the feedback design and timing in MILeS. Similarly, Di Mitri et al. (2022) hammer this point by defining “augmented” feedback, stating that feedback must be timely & actionable within the context. MILeS offer a plethora of “theoretical” affordances for contextualising psychomotor learning. Currently, in its infancy, much of these affordances are yet to be realised, lying behind walls of theoretical and technological complications which need to be ascended. In this light, the workshop also focused on the design and development of infrastructures that will help MILeS researchers conquer them. Slupczynski et al. (2022) envisioned a cloud infrastructure for collecting streams of multimodal data and to train machine learning models in the cloud. This infrastructure could potentially break the bubbles of individual researchers who can rely on data large amounts of data from other researchers. make use of existing trained models, and contribute towards improving the models. An example of such a model is proposed by Paaßen et al. (2022), who propose a model to match the teacher's demonstration of each motion to the student’s attempts and to identify differences between demonstration and attempt, which is a key step in addressing the core issue of transitioning from multimodal sensor data to feedback in MILeS (Romano et al., 2022). During the workshop, several authors presented their preliminary use case studies, taking research on MILeS a step further towards the holy grail, i.e. personalised independent learning of (psychomotor) skills. Geisen et al. (2022) presented their study with a MILeS application for training golf in which she provided visual feedforward to improve the putting performance of golfers. Similarly, Mat Sanusi et al. (2022) proposed a MILeS with an intelligent and personalised feedback system that adapts based on the type of mistakes displayed by the learner. The application of MILeS is, of course, not just limited to the context of learning sports skills. It can also be used to foster human-robot collaboration in contexts such as assembly lines, as proposed by Keller et al. (2022), where it is important to teach the learner not only about the collaborative task but also about the mental model of the robot. Evidently, the MILeS workshop continues to be the Mecca for researchers working on MILeS to get together, learn, share and discuss ideas, and continue to make true the possibility of a truly personalised & situated (psychomotor) skills learning. It was the greatest pleasure to put and see the sMILeS on researchers' faces, despite the MILeS that they have yet to walk, and a 500 more. Acknowledgements The MILeS 2022 chairs would like to thank the authors for their submissions and the EC-TEL workshop chairs for their advice and guidance during the MILeS workshop. The MILeS chairs also served as Program Committee members that provided high-quality reviews for the received submissions. The following project supported the organisation of the MILeS 2022 workshop: Multimodal Immersive Learning with Artificial Intelligence for Psychomotor Skills (MILKI-PSY), funded under the grant number: 16DHB4013 by the Federal Ministry of Education and Research (BMBF). Cologne, September 13th, 2022 On behalf of the editors, Khaleel Asyraaf Mat Sanusi, Bibeg Limbu, & Jan Schneider Program Committee Members Khaleel Asyraaf Mat Sanusi Cologne University of Applied Sciences (TH Köln) Daniel Majonica Cologne University of Applied Sciences (TH Köln) Anja Richert Cologne University of Applied Sciences (TH Köln) Valérie Varney Cologne University of Applied Sciences (TH Köln) Tobias Keller Cologne University of Applied Sciences (TH Köln) Roland Klemke Cologne University of Applied Sciences (TH Köln) and Open University of the Netherlands (OUNL) Jan Schneider Leibniz Institute for Human Development and Educational Information (DIPF) Daniele Di Mitri Leibniz Institute for Human Development and Educational Information (DIPF) George-Petru Ciordas-Hertel Leibniz Institute for Human Development and Educational Information (DIPF) Fernando P. Cardenas-Hernandez Leibniz Institute for Human Development and Educational Information (DIPF) Gianluca Romano Leibniz Institute for Human Development and Educational Information (DIPF) Miloš Kravčík German Research Center for Artificial Intelligence (DFKI) Benjamin Paaßen German Research Center for Artificial Intelligence (DFKI) Ralf Klamma RWTH Aachen Michal Slupczynski RWTH Aachen Stefanie Klatt German Sport University Cologne (DSHS) Mai Geisen German Sport University Cologne (DSHS) Tobias Baumgartner German Sport University Cologne (DSHS) Nina Riedl German Sport University Cologne (DSHS) Bibeg Limbu Center for Education and Learning (LDE-CEL), Technische Universiteit Delft Table of Contents Preface 3 Acknowledgements 4 Program Committee Members 5 MILKI-PSY Cloud: MLOps-based Multimodal Sensor Stream Processing Pipeline for Learning Analytics in Psychomotor Education 8 We can teach more than we can tell: combining Deliberate Practice, Embodied Cognition, and Multimodal Learning. 15 Few-shot Key Pose Detection for Learning of Psychomotor Skills 22 Considerations in Feedback and Periodization for the Multimodal Learning Experience of Running via Wearable Devices 28 IMPECT-Sports: Using an Immersive Learning System to Facilitate the Psychomotor Skills Acquisition Process 34 XR golf putt trainer: User Opinions on an Innovative Real-time Feedback Tool 40 Reflecting on the Actionable Components of a Model for Augmented Feedback 45 Meaningful Feedback from Wearable Sensor Data to Train Psychomotor Skills 51 Prerequisite Knowledge of Learning Environments in Human-Robot Collaboration for dyadic teams 55