Together with colleagues from Finland, South Korea, Czech Republic, Australia, and the US, we critically examine nine persistent challenges that continue to undermine the fairness, transparency, and effectiveness of AI in education (acting as the elephant in the room within AI in education, learning analytics, educational data mining, learning sciences, and educational psychology communities). Our study argues that hybrid human-AI methods offer a promising path towards developing trustworthy and responsible AI systems for education.

Link to the paper: https://arxiv.org/abs/2504.16148

April, 2025 

In collaboration with colleagues from Finland, South Korea, the Czech Republic, and the US, we proposed the idea of responsible educational data mining through hybrid human-AI intelligence. We compared different families of AI—symbolic, sub-symbolic, and neural-symbolic—in predicting 7th-grade mathematics national test performance using data from a 3rd-grade assessment conducted in Estonian schools. Our findings reveal that the hybrid neural-symbolic method not only exhibited stronger predictive power but also incorporated a more integrated set of factors in its decision-making, including motivation, (meta)cognition, and learned knowledge. In contrast, symbolic and sub-symbolic approaches emphasized different dimensions: symbolic methods primarily relied on cognitive and motivational factors, while sub-symbolic methods focused on cognitive features, learned knowledge, and gender—both largely neglecting metacognitive components. These contrasting outcomes highlight the importance of a holistic evaluation of AI methods in educational data mining and learning analytics—moving beyond predictive accuracy alone. Most importantly, they underscore the potential of hybrid, human-centred neural-symbolic approaches to address the limitations of current AI models and promote responsible AI for educational data mining and learning analytics.

Link to the paper: https://arxiv.org/abs/2504.00615

April, 2025 

We conducted a large-scale quasi-experimental study with around 400 students in over 15 Estonian schools to test differentiated learning materials in 8th grade math. At a seminar at Tallinn University, we shared our findings with participating teachers and a representative from the Opiq digital learning platform. Together, we discussed future studies and co-development. This collaboration ensures that our research is both grounded in practice and ready for real classrooms.

April, 2025 

Together with other track chairs from the University of Central Florida (USA), Linnaeus University (Sweden), University of Jyväskylä (Finland), and Max Planck Institute for Software Systems (Germany), we successfully organized the third edition of the AI for Education track at the 40th ACM Symposium on Applied Computing, held this year in Italy.

We received submissions from researchers across 17 countries. Following ACM SAC guidelines, the acceptance rate was around 25%.

Thanks to all authors, reviewers, and participants for contributing to another solid edition of the AIED track.

April, 2025 

Danial delivered a public talk to colleagues at the Faculty of Education and Psychology at the University of Jyväskylä in Finland. The hour-long talk introduced the Edu-HAI project, its activities, and findings, while also addressing some of the existing challenges related to responsible and trustworthy AI for education.

February, 2025 

Phil Winne (Professor Emeritus in the Faculty of Education at Simon Fraser University) has joined the Edu-HAI project.

January, 2025 

Martijn Meeter, advisory board member of the Edu-HAI project, participated in our seminar, reviewing and discussing our 2024 activities.

January, 2025 

Our team members have conducted a quasi-experimental study examining the effects of math-differentiated materials, as well as a Delphi study involving Estonian teachers to explore pedagogical tools and interventions for supporting self-regulated learning in digital environments.

January, 2025 

Our team has conducted a systematic review examining the measurement of (meta)emotion, (meta)motivation, and (meta)cognition using digital trace data in K-12 self-regulated learning. The paper will be presented at the ACM SAC Conference in April 4, 2025. The paper will appear in the SAC'25 Proceedings.

January, 2025 

Our team has proposed an SRL model for studying learning in digital environments. The paper was presented at the 17th annual International Conference of Education, Research and Innovation in Seville, Spain, by Eve Kikas and will appear in the ICERI2024 Proceedings. 

November, 2024 

On September 12, 2024, Kaja Toomla attended the 3rd International Symposium on Digital Transformation, hosted by Linnaeus University in Växjö, Sweden. At the event, Kaja presented her accepted abstract on her PhD research titled "A Generalizable Framework for Tracing and Supporting Self-Regulated Learning in K-12."

September, 2024 

As part of the Edu-HAI project, Ekaterina Krivich has conducted research on the design and development of dynamic Bayesian networks for learner modelling. In collaboration with other team members, she designed a dynamic Bayesian network model specifically for middle school mathematics and trained its parameters using synthetic data from student interactions with a digital learning system. This model can infer and early predict students' knowledge of various concepts at current and future time steps.

After validating the model, Ekaterina developed an interactive prototype that allows users to interact with the model, adjust evidence, view inferences and predictions, and visualise the influence of different concepts on one another. 

Explore the prototype here: https://dbn.firstlaw.io/

August, 2024

The Edu-HAI research group recently gathered for a productive three-day camp in Pühajärve. They reviewed the project's first six months, collaborated on drafting an article based on their findings, and planned the next data collection. 

August, 2024

Marek Druzdzel has participated in Edu-HAI project seminar in Tallinn University, discussing probabilistic graphical models.

June, 2024

We presented the EDU-HAI project at the traditional spring seminar of the School of Digital Technologies. This annual event serves as a platform to reflect on the academic year's achievements, highlight key projects and milestones, and foster new opportunities for collaboration across different Schools within the University.

May, 2024