TY - CHAP A1 - Bugert, Flemming A1 - Grabinger, Lisa A1 - Bittner, Dominik A1 - Hauser, Florian A1 - Nadimpalli, Vamsi Krishna A1 - Staufer, Susanne A1 - Mottok, Jürgen T1 - Towards Learning Style Prediction based on Personality T2 - Proceedings of the 5th European Conference on Software Engineering Education, Seeon/Germany, June 19 - 21, 2023 N2 - This paper assesses the relation between personality, demographics, and learning style. Hence, data is collected from 200 participants using 1) the BFI-10 to obtain the participant’s expression of personality traits according to the five-factor model, 2) the ILS to determine the participant’s learning style according to Felder and Silverman, and 3) a demographic questionnaire. From the obtained data, we train and evaluate a Bayesian network. Using Bayesian statistics, we show that age and gender slightly influence personality and that demographics as well as personality have at least a minor effect on learning styles. We also discuss the limitations and future work of the presented approach. Y1 - 2023 SN - 978-1-4503-9956-2 U6 - https://doi.org/10.1145/3593663.3593682 SP - 48 EP - 55 PB - ACM ER - TY - CHAP A1 - Bittner, Dominik A1 - Hauser, Florian A1 - Nadimpalli, Vamsi Krishna A1 - Grabinger, Lisa A1 - Staufer, Susanne A1 - Mottok, Jürgen T1 - Towards Eye Tracking based Learning Style Identification T2 - Proceedings of the 5th European Conference on Software Engineering Education, Seeon/Germany, June 19 - 21, 2023 N2 - The dropout rate at universities has been very high for years. Thereby, the inexperience and lack of knowledge of students in dealing with individual learning paths in various courses of study plays a decisive role. Adaptive learning management systems are suitable countermeasures, in which learners’ learning styles are classified using questionnaires or computationally intensive algorithms before a learning path is suggested accordingly. In this paper, a study design for student learning style classification using eye tracking is presented. Furthermore, qualitative and quantitative analyses clarify certain relationships between students’ eye movements and learning styles. With the help of classification based on eye tracking, the filling out of questionnaires or the integration of computationally or cost-intensive algorithms can be made redundant in the future. Y1 - 2023 SN - 978-1-4503-9956-2 U6 - https://doi.org/10.1145/3593663.3593680 SP - 138 EP - 147 PB - ACM ER - TY - CHAP A1 - Nadimpalli, Vamsi Krishna A1 - Bugert, Flemming A1 - Bittner, Dominik A1 - Hauser, Florian A1 - Grabinger, Lisa A1 - Staufer, Susanne A1 - Mottok, Jürgen ED - Gómez Chova, Luis ED - González Martínez, Chelo ED - Lees, Joanna T1 - Towards personalized learning paths in adaptive learning management systems: bayesian modelling of psychological theories T2 - Proceedings of the 16th annual International Conference of Education, Research and Innovation (ICERI 2023), Seville, Spain, 11/13/2023 - 11/14/2023 N2 - In educational research, non-personalized learning content increases learners' cognitive load, causing them to lower their performance and sometimes drop out of the course. Personalizing learning content with learners’ unique characteristics, like learning styles, personality traits, and learning strategies, is being suggested to improve learners’ success. Several theories exist for assessing learners’ unique characteristics. By the end of 2020, 71 learning style theories have been formulated, and research has shown that combining multiple learning style theories to recommend learning paths yields better results. As of the end of 2022, there is no single research that demonstrates a relationship between the Index of Learning Styles (ILS) based Felder-Silverman learning style model (FSLSM) dimensions, Big Five (BFI-10) based personality traits, and the Learning strategies in studying (LIST-K) based learning strategies factors for personalizing learning content. In this paper, an innovative approach is proposed to estimate the relationship between these theories and map the corresponding learning elements to create personalized learning paths. Respective questionnaires were distributed to 297 higher education students for data collection. A three-step approach was formulated to estimate the relationship between the models. First, a literature search was conducted to find existing studies. Then, an expert interview was carried out with a group of one software engineering education research professor, three doctoral students, and two master’s students. Finally, the correlations between the students' questionnaire responses were calculated. To achieve this, a Bayesian Network was built with expert knowledge from the three-step approach, and the weights were learned from collected data. The probability of individual FSLSM learning style dimensions was estimated for a new test sample. Based on the literature, the learning elements were mapped to the respective FSLSM learning style dimensions and were initiated as learning paths to the learners. The next steps are proposed to extend this framework and dynamically recommend learning paths in real time. In addition, the individual levels of learning style dimensions, personality traits, and learning strategies can be considered to improve the recommendations. Further, using probabilities for mapping learning elements to learning styles can increase the chance of initiating multiple learning paths for an individual learner. Y1 - 2023 U6 - https://doi.org/10.21125/iceri.2023.1144 SP - 4593 EP - 4603 PB - IATED ER - TY - CHAP A1 - Ezer, Timur A1 - Plößl, Moritz A1 - Grabinger, Lisa A1 - Bittner, Dominik A1 - Staufer, Susanne A1 - Nadimpalli, Vamsi Krishna A1 - Bugert, Flemming A1 - Hauser, Florian A1 - Mottok, Jürgen ED - Gómez Chova, Luis ED - González Martínez, Chelo ED - Lees, Joanna T1 - Deep learning for eye movement classification T2 - ICERI 2024 Proceedings, 17th annual International Conference of Education, Research and Innovation,11-13 November, 2024, Seville, Spain Y1 - 2024 U6 - https://doi.org/10.21125/iceri.2024.1028 SP - 4056 EP - 4065 PB - IATED ER - TY - CHAP A1 - Hauser, Florian A1 - Staufer, Susanne A1 - Grabinger, Lisa A1 - Röhrl, Simon A1 - Mottok, Jürgen ED - Gómez Chova, Luis ED - González Martínez, Chelo ED - Lees, Joanna T1 - On the analysis of student learning strategies: using the LIST-K questionnaire ro generate ai-based individualized learning paths T2 - Proceedings of the 16th annual International Conference of Education, Research and Innovation (ICERI 2023), Seville, Spain, 11/13/2023 - 11/14/2023 N2 - This paper presents the results of a data collection with the LIST-K questionnaire. This questionnaire measures students’ learning strategies and shows which strategies are particularly dominant or rather weak. Learning strategies have long been a major area of research in educational science and psychology. In these disciplines, learning strategies are understood as intentional behaviors and cognitive skills that learners employ to effectively complete learning tasks, by selecting, acquiring, organizing, and integrating information into their existing knowledge for long-term retention. The LIST-K, developed by Klingsieck in 2018, was chosen for accessing learning strategies due to its thematic suitability, widespread use, and test economy. It covers a total of four main categories (i.e., cognitive strategies, metacognitive strategies, management of internal resources, and management of external resources), each of which are subdivided into further subscales. With a total of 39 items answered via a 5-step Likert scale, the LIST-K can cover the topic relatively comprehensively and at the same time be completed in a reasonable amount of time of approximately 10 minutes. The LIST-K was used as part of a combined data collection along with other questionnaires on their personal data, their preferences regarding certain learning elements, their learning style (i.e. the ILS), and personality (i.e. the BFI-10). A total of 207 students from different study programs participated via an online survey created using the survey tool "LimeSurvey". Participation in the study was voluntary, anonymously, and in compliance with the GDPR. Overall, the results of the LIST-K show that students are willing to work intensively on relevant topics intensively and to perform beyond the requirements of the course seeking additional learning material. At the same time, however, it is apparent that the organization of their own learning process could still be improved. For example, students start repeating content too late (mean=2.70; SD=0.92) and do not set goals for themselves and do not create a learning plan (mean=3.19; SD=0.90). They also learn without a schedule (mean=2.23; SD=0.97) and miss opportunities to learn together with other students (mean=3.17; SD=0.94). The findings of the data collection will be used to create an AI-based adaptive learning management system that will create individualized learning paths for students in their respective courses. From the results of the LIST-K, it appears that the adaptive learning management system should primarily support organizational aspects of student learning. Even small impulses (an individual schedule of when to learn what or a hierarchical structuring of the learning material) could help students to complete their courses more successfully and improve their learning. Y1 - 2023 U6 - https://doi.org/10.21125/iceri.2023.1147 SP - 4611 EP - 4620 PB - IATED ER - TY - CHAP A1 - Bugert, Flemming A1 - Nadimpalli, Vamsi Krishna A1 - Bittner, Dominik A1 - Ezer, Timur A1 - Grabinger, Lisa A1 - Maier, Robert A1 - Röhrl, Simon A1 - Staufer, Susanne A1 - Hauser, Florian A1 - Mottok, Jürgen ED - Mottok, Jürgen ED - Hagel, Georg T1 - ML based Evaluation Methodology for Learning Path Recommender Systems T2 - Proceedings of the 6th European Conference on Software Engineering Education : ECSEE 2025, Seeon Germany, June 02-04, 2025 N2 - In education, recommender systems can provide students with personalized learning materials based on their preferences. When comparing various recommendation algorithms, the main question is, which algorithm provides the most suitable recommendations for each student. Answering this question requires a quantitative evaluation methodology (i.e. a concrete metric) for ranking the results of (even non-deterministic) recommender systems. While there is already literature on this topic, the uniqueness of our approach lies in the application of machine learning: we deploy a likelihood based analysis via Hidden Markov Models named Aiakos. With this strategy, we aim to provide data-driven insights about accuracy and stability of recommendations towards a more reasonable selection of the appropriate recommender system. The training data for the Hidden Markov Models is collected from 80 students. Data from another 26 students is then used to discuss the behavior of our evaluation procedure considering a single recommendation as well as the results from 100 recommendations. Furthermore, the proposed concept allows to be applied to other domains as well. Y1 - 2025 SN - 9798400712821 U6 - https://doi.org/10.1145/3723010.3723022 SP - 40 EP - 48 PB - ACM ER - TY - CHAP A1 - Nadimpalli, Vamsi Krishna A1 - Bugert, Flemming A1 - Bittner, Dominik A1 - Staufer, Susanne A1 - Röhrl, Simon A1 - Hauser, Florian A1 - Ezer, Timur A1 - Grabinger, Lisa A1 - Maier, Robert A1 - Mottok, Jürgen T1 - Probabilistic Machine Learning for Simulating Complex Learner Profiles T2 - 2024 21st International Conference on Information Technology Based Higher Education and Training (ITHET), 06-08 November 2024, Paris, France N2 - Data is the foundation of all machine learning applications. In education science, especially for the learner characteristics that drive personalized learning, it is difficult to collect and often uncertain. It is challenging to model, train, evaluate, and analyze the underlying algorithms when developing AI-based systems and having small sample sizes. To address these problems, we present a synthetic data generator utilising probabilistic models. This generator can effectively model and simulate complex learner profiles. To achieve this, we collected extensive data on learning styles, learning strategies, personalities, and preferred learning paths from 593 students over several semesters at a higher education level. Then, Bayesian networks, Hidden Markov Models, and Markov Chains are used to model the relationships between learner profiles. Using the Bayesian information criterion, and cross-validation with log-likelihood scores, we compare various models to select the best fitting one for synthesizing the data. The synthetic data is then evaluated using statistical validation techniques. In addition, we developed a simulation module with the option to simulate learner profiles based on manual user-defined inputs. The data and code used in this work are available as open source1110.5281/zenodo.13768136 [Titel anhand dieser DOI in Citavi-Projekt übernehmen] contributing to open science and developers for customized simulated data. In the future, this data will refine the training, evaluation, analysis, and benchmarking of algorithms for personalized learning. KW - probabilistic models KW - synthetic data KW - learner characteristics KW - generative AI KW - higher education area Y1 - 2024 SN - 979-8-3315-1663-5 U6 - https://doi.org/10.1109/ITHET61869.2024.10837641 SN - 2473-2060 SP - 1 EP - 10 PB - IEEE ER - TY - CHAP A1 - Nadimpalli, Vamsi Krishna A1 - Staufer, Susanne A1 - Ezer, Timur A1 - Bugert, Flemming A1 - Bittner, Dominik A1 - Hauser, Florian A1 - Grabinger, Lisa A1 - Röhrl, Simon A1 - Maier, Robert A1 - Mottok, Jürgen ED - Gómez Chova, Luis ED - González Martínez, Chelo ED - Lees, Joanna T1 - Predicting learner characteristics using machine learning T2 - ICERI 2024 Proceedings, 17th annual International Conference of Education, Research and Innovation,11-13 November, 2024, Seville, Spain N2 - In education science research, data collection is challenging due to difficulty identifying students at the higher education level, privacy concerns, and varying levels of student engagement. Importantly, psychological questionnaires can be lengthy, leading to incomplete responses. We conducted repeated studies, and over time, the focus of the research adapted, introducing new materials and consequently leading to missing learner characteristics in some datasets. In this research, the issue of incomplete learner characteristics is addressed using data from three different studies: winter term 2022/2023 (n=297), summer term 2023 (n=274), and winter term 2023/2024 (n=25). These studies collected various learner characteristics, such as learning styles, personalities, learning strategies, and learning element preferences. However, learning element preferences and learning strategies were missing in the winter term of 2022, and the summer term of 2023 respectively. To analyze the data and predict these missing features, statistical analysis, and machine learning techniques were employed. Then, these models are rigorously evaluated using cross-validation and performance metrics like accuracy, precision, recall, and F1-score. Our findings provide insights into the relationships between learners' learning styles, personalities, learning strategies, and learning element preferences. This offers valuable implications for the design and implementation of educational interventions, like learning path recommendations. The results imply that machine learning models can predict missing learner characteristics, thus addressing the problem of incomplete data in educational research Y1 - 2024 U6 - https://doi.org/10.21125/iceri.2024.1454 SP - 6005 EP - 6014 PB - IATED ER - TY - THES A1 - Grabinger, Lisa T1 - Data Science in der empirischen Eyetracking-Forschung N2 - Empirische Forschung stellt Anfänger vor eine Vielzahl von Herausforderungen. Besonders herausfordernd gestaltet sich die (statistische) Datenanalyse -- eine Aufgabe, die bislang Vorkenntnisse aus zwei verschiedenen Bereichen erfordert: Experimentalforschung und Programmierung. Es muss bekannt sein, was zu tun ist, aber auch wie dies technisch umgesetzt werden kann. Die vorliegende Arbeit befasst sich mit der Frage, wie angehende Forschende bei dieser Aufgabe bestmöglich unterstützt werden können. In einem ersten Schritt wird hierfür die Ausgangslage erfasst. Mit einer systematischen Literaturrecherche wird identifiziert, welche Analyseverfahren in der Praxis Anwendung finden; für diese Verfahren werden dann jeweils der theoretische Hintergrund und die praktische Anwendung aufgearbeitet. Zudem werden in unsystematischer Recherche die Möglichkeiten ausgelotet, die bestehende Software-Systeme bieten. Basierend darauf wird dann eine Webanwendung konzipiert bzw. realisiert, die gängige Analysen auf eine Weise ermöglicht, die sowohl für Statistik- als auch für Programmieranfänger geeignet ist. Das entstandene System wird in einem dritten Schritt durch heuristische Diskussion und empirische Nutzerstudien bestätigt. Angemerkt sei, dass alle Untersuchungen auf die empirische Eyetracking-Forschung in der Disziplin Software Engineering zugeschnitten, aber nicht darauf beschränkt sind. KW - Data Science KW - Eyetracking KW - Statistik Y1 - 2025 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:898-opus4-84797 ER - TY - CHAP A1 - Staufer, Susanne A1 - Bugert, Flemming A1 - Hauser, Florian A1 - Grabinger, Lisa A1 - Ezer, Timur A1 - Nadimpalli, Vamsi Krishna A1 - Bittner, Dominik A1 - Röhrl, Simon A1 - Mottok, Jürgen ED - Gómez Chova, Luis ED - González Martínez, Chelo ED - Lees, Joanna T1 - Tyche algorithm: Markov models for generating learning paths in learning management systems T2 - INTED2024 Proceedings: 18th International Technology, Education and Development Conference, Valencia, Spain. 4-6 March, 2024 N2 - In the intricate tapestry of the cosmos, where celestial threads weave stories of destiny, an enigmatic deity stands at the intersection of chance and fate — Tyche, the goddess of fortune. In science and probability, coincidence plays a distinctive role in Bayesian Networks (BNs) and Markov Models (MMs). This paper introduces the Tyche algorithm named after the goddess of fortune. The Tyche algorithm is a Markov model designed to generate learning paths in Learning Management Systems (LMSs). A learning path is a type of individualization that personalises the order of learning elements within an LMS course. Learning elements are fundamental components within an LMS course, depicting the learning content in diverse ways. In our case, the learning path tailoring is based on the learner’s learning style according to Felder-Silverman Learning Style Model (FSLSM) – an indicator for the ideal pathway and learning element for the learner’s optimum learning. The Tyche algorithm offers the advantage to provide students the most suitable learning path. Tyche is a MM structure with various matrices containing transition probabilities depending on the learning style. Nine categories of learning elements defined in a previous survey form the basis of the node structure of the MM. For a generic approach, a survey was designed to obtain the transition probabilities depending on the individual learning style. The survey with more than 100 German students participated is processed with the tool LimeSurvey. Students are asked about their learning style using the Index of Learning Styles (ILS) questionnaire according to Felder-Silverman and about the percentage probabilities of learning elements to get their individual sequence as learning path. The percentages are queried in two different ways. Firstly, the students were asked to provide a sequence of learning elements within each position of the sequence filled with probabilities for all learning elements. Secondly, a learning element is given and the students are asked to indicate which learning element they would work on next. The first way of asking for the probabilities is used to find the start node in the MM, whereas the second approach forms the probability matrices between the nodes within the MM. As result of the survey, the Tyche algorithm presents generic transition probabilities. It improves the learning process of individuals only by asking for their learning style: it generates individual learning paths through the learning elements within an LMS based on the MM explained above by solely getting the answers of the ILS questionnaire as input. In the future, other questionnaires such as BFI-10 for personality traits or LIST-K for learning strategies may offer a more comprehensive input. However, the next step is to evaluate Tyche with about 25 students in a software engineering lecture. This is planned for the year 2024. KW - Markov model KW - learning style KW - learning elements KW - learning management system KW - higher education area Y1 - 2024 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:898-opus4-71148 SN - 978-84-09-59215-9 SN - 2340-1079 SP - 4195 EP - 4205 PB - IATED Academy ER -