The 25th International Conference on Artificial Intelligence in Education (AIED 2024), held from July 8-12 in Recife, Brazil, was a significant event for the Highly Informative Learning Analytics Research Programme. This year marked the first Brazilian-German cooperation in this field, supported by the Alexander Humboldt Foundation, the DIPF in Frankfurt and IPN in Kiel under the ALICE project. Two workshop papers presented at the conference showcased innovative approaches to automatically analyze concept maps, promising to automate the way educators assess and understand the student-created context.
#AIED24 #LearningAnalytics #ConceptMaps #AIinEducation #EducationalTechnology #MachineLearning #CulturalDiversity #RealTimeFeedback #EdTech #AI #Education #CrossCulturalCollaboration
Paper 1: The Influence of Diverse Educational Contexts on Concept Map Structures
Authors: Laís P. Van Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, and Hendrik Drachsler
Presented at: 7th International Workshop on Culturally-Aware Tutoring Systems (CATS 2024)
Concept maps are essential tools in educational settings, helping educators grasp students’ knowledge construction. However, the open nature of concept maps means their structures can vary widely among students and contexts. This paper presents a qualitative assessment of 317 concept maps from Germany and six from Brazil, involving students from secondary to Master’s levels, with varying familiarity with concept maps and in different settings (individual vs. group).
The maps were classified into three structural types: network, spoke, or chain. A Decision Tree model was trained to classify the map structures using the German data. When tested on the Brazilian dataset, the model’s performance decreased, suggesting that cultural and contextual differences significantly influence concept map construction. This finding emphasizes the need to adapt the teaching and use of concept maps to diverse educational contexts.
The CATS 2024 workshop highlighted the importance of considering cultural influences in educational practices. Research indicates that culture impacts cognition, motivation, and emotions, which naturally attracted the AIED community’s interest. The workshop engaged researchers in discussions on designing educational systems that account for cultural and contextual variations, a critical step in creating more effective and inclusive educational technologies.
Paper 2: Concept Map Assessment Through Structure Classification
Authors: Laís P. V. Vossen, Isabela Gasparini, Elaine H. T. Oliveira, Berrit Czinczel, Ute Harms, Lukas Menzel, Sebastian Gombert, Knut Neumann, and Hendrik Drachsler
Presented at: Workshop on Automated Evaluation of Learning and Assessment Content
Concept maps’ wide applicability in almost all domains makes them valuable tools for conceptual learning and understanding students’ conceptual understanding of a domain. This paper focused on classifying concept map structures into three types: spoke, network, and chain. By examining 317 distinct concept maps, the study used statistical data to train multiclass classification models, achieving impressive accuracy with a Decision Tree model.
The paper shows the potential of concept maps to provide highly informative feedback to students. Such immediate insights into students’ understanding can significantly enhance the learning process, allowing educators to tailor their teaching strategies more effectively.
One of the key arguments emerging from this paper is the power of interdisciplinary and cross-cultural collaboration within concept maps. Integrating perspectives from diverse cultural and educational contexts enriches the research process and leads to more robust and adaptable educational technologies. By combining the expertise of researchers from Brazil and Germany, these studies have demonstrated the importance of considering cultural diversity in developing and applying AI in education. This collaborative approach ensures that educational tools are more inclusive and effective across different contexts, ultimately enhancing learning outcomes for a broader range of students.
Conclusion
The presentations at AIED 2024 showcased steps ahead in the automated analysis of concept maps, these studies not only provide deeper insights into how educational contexts influence concept map structures but also pave the way for innovative assessment tools that can deliver highly informative feedback to learners. As the intersection of artificial intelligence and education continues to evolve, such collaborative efforts and technological advancements will be crucial in shaping the future of learning analytics and educational assessment.