Digital quizzes are widely used in engineering higher education in diverse ways; however, their impact is frequently analysed as discrete assessment events rather than as pedagogically embedded learning opportunities. This study investigates the multifaceted landscape of how engineering education teachers and students perceive and use digital quizzes, engage with them, and perceive their pedagogical value as a form of learning support. The results are based on primary data collected through two surveys among the teachers and registered students in an engineering higher education institution; hence, 1291 responses from students and 314 from teachers were analysed. Overall, teachers see digital quizzes primarily as a tool for efficient assessment of core knowledge and supporting student engagement. However, teachers stressed the importance of well-designed questions. From the students' point of view, the overall message is that quizzes are most effective when they are frequent, low-stakes, allow multiple attempts, and provide meaningful feedback. The analysis outcomes collectively suggest that while digital quizzes can be beneficial in principle, they often fall short in practice due to design issues, superficiality, and lack of alignment with meaningful learning outcomes.

Generative Artificial Intelligence (GenAI) is reshaping Multimodal Learning Analytics (MMLA) by augmenting and extending the semantics of learning interactions. While recent systematic research indicates that GenAI is currently used primarily for individual learners’ tasks such as tutoring or co-programming, its application in collaborative learning environments remains largely underdeveloped and overlooked. This half-day workshop invites researchers to explore how Large Language Models (LLMs) and multimodal generative learning and teaching architectures can serve as additional sources of semantic sensors to decode meaning, motivation, coherence, and learning performance in group learning. By enhancing and extending simple behaviour tracking, we aim to envision systems and tools that foster meaningful reciprocal human-GenAI educational interaction, supporting teams in sustaining and further developing higher-order thinking skills such as problem-solving and critical thinking without falling into the trap of cognitive “replacement” or over-reliance.

This report contains a large potion of the result of the KTH e-learning surveys, administered inDecember 2024. One survey was sent out to teachers and one to students. The main purpose ofthe surveys is to gain insights of how teachers and students perceive the current digital learningenvironment and also explore around challenges, within a near future, concerning digital learningin higher education. The themes of those challenges in these surveys were generative AI and digitalassessment. The data gathered and the report will also be used to make informed decision for thework that is carried out within the E-learning object at KTH.

The aim of this report has been to present the result in a straightforward way, mainly with graphsand comments on those visualisation of the data. Thus, no deep statistical analysis will be presentedin this report. The survey was divided into the following parts: demography, digital education,generative AI, digital assessment. The data collection yielded responses from 314 teachers and 1291 students, corresponding to response rates of 19% and 6%, respectively.

The result from the digital education part indicates that teacher think KTH has succeeded well withdigital education in 2024 and that they are offered well-functioning digital infrastructures and toolsto conduct high-quality digital education. Furthermore, students think that the education at KTHuses digital opportunities in a good way and that the education is of high quality. Both groups alsoagree in general that KTH should continue developing digital education.

Students report that they are very familiar with AI chatbots and are also often using these tools fortheir studies. Teachers are less familiar, and 25% report that they use AI chatbots weekly or moreoften in their role as a teacher, whereas 37% have never used i. The wide availability of AI chatbotshas also led many students to change their study habits, and about two-thirds of the teachers havemade some changes to their teaching practice. The most commonly perceived usefulness and reported use of AI chatbots for teachers was help with creating materials for teaching or examination,whilst students perceive it as useful and use it in their studies to pose questions that they have, findinformation and summarize text. Students think that AI chatbots increase cheating, but few reportthat they have used AI chatbots in a way they were not allowed to. About one quarter of the teacherreported that they have made moderate or significant changes in their assessment practices.

The results from the part focused on digital assessment showed that slightly more than half of teachers (54%) had created at least one digital assessment during 2024, while 62% of students had takenat least one digital exam. The main reasons for using digital assessment were the ability to providetimely, frequent feedback and to save time and resources. Among students who prefer digital assessment, 63% cited quick feedback and many also found digital exams to be easier to complete andsubmit. Students who preferred paper-based exams emphasized feeling more comfortable (66%)and better able to focus (61%). Teachers who used digital assessment reported technical problemsduring exams as their main challenge, a concern echoed by students, who worried about log-inissues, lost internet connections, and needing to troubleshoot on their own.

The growing adoption of Large Language Models (LLMs) is transforming learning in higher education. While they provide flexible support, concerns remain about over-reliance, uncritical acceptance of outputs, and excessive cognitive offloading. To address these challenges, we foreground self-regulated learning for LLMs (SRL-for-LLM), defined as learners' ability to plan, monitor, and reflect on AI interactions to support learning. Guided by SRL theory, we introduce ChatWise, a proof-of-concept browser extension that combines conversational scaffolding with learning analytics to classify student prompts by SRL strategies and deliver adaptive metacognitive feedback in real time. A two-phase study included the design of ChatWise and a within-subject field study with STEM undergraduates. Log analyses using mixed-effects modeling showed that access to ChatWise more than doubled the likelihood of producing high-quality prompts. Findings further indicate improved prompt refinement, greater strategic awareness, and more reflective engagement with AI-supported learning. These results highlight the potential of SRL-aligned, analytics-driven scaffolding to support more effective student-LLM interactions.

Deep learning models for text classification have been increasingly used in intelligent tutoring systems and educational writing assistants. However, the scarcity of data in many educational settings, as well as certain imbalances in counts among the annotated labels of educational datasets, limits the generalizability and expressiveness of classification models. Recent research positions LLMs as promising solutions to mitigate the data scarcity issues in education. In this paper, we provide a systematic literature review of recent approaches based on LLMs for generating textual data and augmenting training datasets in the broad areas of natural language processing and educational technology research. We analyze how prior works have approached data augmentation and generation across multiple steps of the model training process, and present a taxonomy consisting of a five-stage pipeline. Each stage covers a set of possible options representing decisions in the data augmentation process. We then apply a subset of the identified methods to three educational datasets across different domains and source languages to measure the effectiveness of the suggested augmentation approaches in educational contexts, finding improvements in overall balanced accuracy across all three datasets. Based on our findings, we propose our pipeline as a conceptual framework for future researchers aiming to augment educational datasets for improving classification accuracy¹ .

¹The open-source code of our experiments, as well as the prompts used for the LLM and the detailed results of our experiments, can all be found on: https://github.com/epfl-ml4ed/data-aug-education

Jupyter is a web-based, interactive computing environment that supports many commonly-used programming languages. It has been widely adopted in the CS education community and is now rapidly expanding to other STEM disciplines due to the growing integration of programming in STEM education. However, unlike other educational platforms, there is currently no integrated way to capture, analyze, and visualize student interaction data in Jupyter notebooks. This means that teachers have limited to no visibility into student activity, preventing them from drawing insights from these data and providing timely interventions on the fly. In this paper, we present Jupyter Analytics, an end-to-end solution for teachers to collect, analyze, and visualize both synchronous and asynchronous learning activities in Jupyter. The Jupyter Analytics system consists of two JupyterLab extensions connected via a cloud-based backend. On the student side, we introduce the Jupyter Analytics Telemetry extension to anonymously capture students’ interaction activity with more structure and higher granularity than log data. On the teacher side, we introduce the Jupyter Analytics Dashboard extension, which visualizes real-time student data directly in the notebook interface. The Jupyter Analytics system was developed through an iterative co-design process with university instructors and teaching assistants, and has been implemented and tested in several university STEM courses. We report two use cases where Jupyter Analytics impacted teaching and learning in the context of exercise sessions, and discuss the potential value of our tools for CS education.

There are a variety of CSCL scripts that offer different levels of flexibility, ranging from high-structured guidance towards low-rigidity support. This study is grounded in the view that scripts should be designed with a certain degree of structuredness, and we aim to examine the presence of implicit scripts based on data collected from collaborative programming activities in authentic classroom settings. By extracting the task dependencies within Jupyter notebooks, we adopted Directed Acyclic Graphs (DAGs) to operationalize the concept of implicit scripts. Next, we analyzed the group interaction data and measured their collaboration strategy alignment with the implicit scripts. Notably, we found that introducing a planning phase significantly improved groups' adherence to the implicit scripts. Qualitative data were reported to triangulate our findings. We discuss pedagogical and technological implications for more general CSCL contexts.

Reflective writing is known as a useful method in learning sciences to improve the metacognitive skills of students. However, students struggle to structure their reflections properly, limiting the possible learning gains. Previous works in educational technologies literature have explored the paradigms of learning from worked and modelling examples, but (a) their application to the domain of reflective writing is rare, (b) such methods might not scale properly to large-scale classrooms, and (c) they do not necessarily take the learning needs of each student into account. In this work, we suggest two approaches of integrating AI-enabled support in digital systems designed around learning from worked and modelling examples paradigms, to provide personalized learning and feedback to students using large language models (LLMs). We evaluate Reflectium, our reflective writing assistant, show benefits of integrating AI support into the learning from examples modalities and compare the perception of the users and their interaction behaviour when using each version of our tool. Our work sheds light on the applicability of generative LLMs to different types of providing support using the learning from examples paradigm, in the domain of reflective writing.

Interactive simulations enhance science education and foster inquiry skills, but their open-ended nature can be cognitively overloading. While adaptive systems offer timely support, research on predicting conceptual understanding in these environments is limited. Most models are simulation-specific, leading to time-consuming and non-generalizable solutions. In this paper, we introduce a universal encoding that converts lower-level interaction data into higher-level features applicable across various open-ended learning environments (OELEs). This encoding aims to offer a general framework to model inquiry across environments and to alleviate challenges such as the “cold start” problem. Our findings demonstrate that models trained on the universal encoding perform comparably to or better than study-specific encodings across multiple contexts. Code is provided in https://github.com/epfl-ml4ed/universal-oele.

Generative AI technologies are increasingly being incorporated into creativity support tools. However, most generative AI tools rely on text-based prompting, requiring users to translate visual ideas into linguistic descriptions. This approach is misaligned with the sketch-driven workflows of creative professionals. To address this gap, we introduce SketchAI, a novel sketch-first interface for diffusion models that allows practitioners to use real-time sketching on a tablet computer to guide model outputs. Through a qualitative study with 29 fashion design apprentices, we explored the interface’s potential impacts on creative workflows. While some participants identified use cases where SketchAI streamlined routine tasks, others expressed concerns about its potential to undermine creative agency and exploration. These findings unearthed hidden complexity: while generative AI can support some aspects of creativity, its core capabilities may challenge the central identity of creative practitioners. While SketchAI does not resolve this problem, it does take a meaningful step towards reconciliation.