Open Access

This study investigated the initiation of digitally supported learning activities and personal and institutional factors associated with them in different higher education courses, based on the C♭-model. The C♭-model is a theoretical framework that systematizes contextual factors, which influence students‘ learning activities as the most important facilitator of students’ learning success. Using a self-assessment instrument with anchored scenarios in a sample of 1625 higher education teachers, we were able to identify three levels at which higher education teachers initiated digital learning activities: a low level (powerpointers), a moderate level (clickerers), and a high level (digital pros). The findings also support the relevance of the contextual factors specified in the C♭-model for initiating a high level of digital learning activities, namely digitalization policy and commitment of university administration, institutional equipment, technical and educational support, self-assessed basic digital skills, and self-assessed technology-related teaching skills. All of these factors explain a substantial amount of variance in the level of initiated digital learning activities. We conclude that a comprehensive approach rather than isolated measures might contribute to successful teaching and learning in higher education.

Background Artificial intelligence, particularly natural language processing (NLP), enables automating the formative assessment of written task solutions to provide adaptive feedback automatically. A laboratory study found that, compared with static feedback (an expert solution), adaptive feedback automated through artificial neural networks enhanced preservice teachers' diagnostic reasoning in a digital case-based simulation. However, the effectiveness of the simulation with the different feedback types and the generalizability to field settings remained unclear. Objectives We tested the generalizability of the previous findings and the effectiveness of a single simulation session with either feedback type in an experimental field study. Methods In regular online courses, 332 preservice teachers at five German universities participated in one of three randomly assigned groups: (1) a simulation group with NLP-based adaptive feedback, (2) a simulation group with static feedback and (3) a no-simulation control group. We analysed the effect of the simulation with the two feedback types on participants' judgement accuracy and justification quality. Results and Conclusions Compared with static feedback, adaptive feedback significantly enhanced justification quality but not judgement accuracy. Only the simulation with adaptive feedback significantly benefited learners' justification quality over the no-simulation control group, while no significant differences in judgement accuracy were found. Our field experiment replicated the findings of the laboratory study. Only a simulation session with adaptive feedback, unlike static feedback, seems to enhance learners' justification quality but not judgement accuracy. Under field conditions, learners require adaptive support in simulations and can benefit from NLP-based adaptive feedback using artificial neural networks.

Scientific reasoning and argumentation (SRA) skills are crucial in higher education, yet comparing studies on these skills remains challenging due to the scarcity of well-developed SRA-tests with robust psychometric properties. In this paper, the case-based ASSESSRA approach is proposed to evaluate university students’ SRA-skills, focusing specifically on the skills evidence evaluation and drawing conclusions. A prototype constructed using this approach in an educational context demonstrated reliability within an expert panel (n = 9; ICC = .81). In a subsequent study, the validity of the ASSESSRA approach was examined with 207 students, a partial-credit-model exhibited an acceptable fit, demonstrating no significant outfit and excellent distribution of ability parameters and Thurstonian thresholds. The ASSESSRA-prototype, coupled with provided guidelines, offers a versatile framework for developing comparable SRA-tests across diverse domains. This approach not only addresses the current gap in SRA assessment instruments but also holds promise for enhancing the understanding and promotion of SRA-skills in higher education.

During the COVID-19 pandemic, emergency online learning impeded the pursuit of in-person activities that usually foster successful socialization in higher education. To investigate the effects of online learning on socialization, we asked two exploratory research questions: (1) How and to what extent does the level of socialization change during the first online semester? and (2) To what extent does level of change predict course dropout and academic performance? In our case study, using a sample of new students at a large German university, we ran an autoregressive three-factorial model of socialization (role, relationships, organization) with three measurements taken during the new students’ first semester, which was the second semester in which emergency online learning took place. Our results show that the relationships component of socialization did not increase over the semester, while the role and organization components increased. Furthermore, our results support a negative effect of the organization component of socialization on course dropout and a positive effect of the relationship component of socialization on academic performance.

Many complex discourse-level tasks can aid domain experts in their work but require costly expert annotations for data creation. To speed up and ease annotations, we investigate the viability of automatically generated annotation suggestions for such tasks. As an example, we choose a task that is particularly hard for both humans and machines: the segmentation and classification of epistemic activities in diagnostic reasoning texts. We create and publish a new dataset covering two domains and carefully analyse the suggested annotations. We find that suggestions have positive effects on annotation speed and performance, while not introducing noteworthy biases. Envisioning suggestion models that improve with newly annotated texts, we contrast methods for continuous model adjustment and suggest the most effective setup for suggestions in future expert tasks.

Diagnostic reasoning is a key component of many professions. To improve students' diagnostic reasoning skills, educational psychologists analyse and give feedback on epistemic activities used by these students while diagnosing, in particular, hypothesis generation, evidence generation, evidence evaluation, and drawing conclusions. However, this manual analysis is highly time-consuming. We aim to enable the large-scale adoption of diagnostic reasoning analysis and feedback by automating the epistemic activity identification. We create the first corpus for this task, comprising diagnostic reasoning self-explanations of students from two domains annotated with epistemic activities. Based on insights from the corpus creation and the task's characteristics, we discuss three challenges for the automatic identification of epistemic activities using AI methods: the correct identification of epistemic activity spans, the reliable distinction of similar epistemic activities, and the detection of overlapping epistemic activities. We propose a separate performance metric for each challenge and thus provide an evaluation framework for future research. Indeed, our evaluation of various state-of-the-art recurrent neural network architectures reveals that current techniques fail to address some of these challenges.