Reflective learning enhances understanding, especially when instructors promptly address difficulties raised in student reflections. Automated doubt detection can reduce time for instructors, yet existing classification approaches take substantial time for manual annotation and model training. This paper investigates whether large and small language models (LLMs, SLMs) can automate doubt detection without time-consuming training. Using a dataset of anonymized student reflections, we evaluate zero-shot, few-shot prompting, and multi-step reasoning against prior supervised classification baselines. We show that LLMs (GPT-4o, Claude-4, Gemini-2.5) surpass earlier F1 scores without prompting, while prompting further improves their performance. However, using proprietary LLMs can raise cost and privacy concerns. We also show that selected SLMs (Mistral, Qwen) outperform baselines while addressing these concerns. We extend the analysis with a category-level error study, showing that explicit doubts are detected more reliably, while tentative doubts (softened by cautious language), learning challenge (arising from difficulties in applying concepts), and masked doubts (concealed by positive or polite phrasing) are missed more often. These findings highlight both the promise and limitations of language models for doubt detection and the need to ensure that cautious or polite learners who may not express their doubts explicitly are recognized and supported in learning analytics systems.

This study investigates ChatGPT-4o's ability to answer multi-modal assessment exercises in computer science (CS) courses. While the use of large language models (LLMs) to answer text-based exercises are extensively researched, their ability to answer exercises involving artifacts of other modalities remains underexplored. To close this gap, we evaluate ChatGPT-4o's answers to 120 multi-modal CS exercises in programming, software design, human-computer interaction, statistical analysis, process analysis, and simulation. The multi-modal artifacts in these exercises include class diagrams, sequence diagrams, user interface images, analytical charts, workflow diagrams and object-flow diagrams. Our comparisons to the expected answers of these exercises show that ChatGPT-4o performs well for exercises with class and sequence diagrams possibly due to the availability of more data for training. The potential for misuse by students highlights these exercises are better suited for closed-book exams or as scaffolding activities. ChatGPT-4o answers better for those multi-modal exercises designed to assess students at the lower levels of Bloom's taxonomy than the higher levels. This discrepancy is possibly due to ChatGPT-4o's lack of understanding underlying design concepts and limited ability to generate new multi-modal artifacts, making exercises requiring higher order of cognitive thinking suitable for take-home assignment. We hope the insights from this study provide a foundation to develop effective multi-modal assessments.

This study explores the integration of a Large Language Model (LLM) based chatbot, PromptTutor, into flipped classrooms (FC) for undergraduate Computer Science (CS) education. PromptTutor is designed to provide personalized, immediate feedback to support student learning in FC by incorporating reflective learning and scaffolding strategies. The traditional FC typically lacks this immediate feedback during the pre-class learning phase, risking decreased student motivation according to existing literature. This study examines if students improve in learning outcomes and motivation after using PromptTutor. Through a controlled crossover experiment with 50 students, the study demonstrates statistically significant improvements in students' quiz performance and motivation compared to traditional FC. Our work underscores the potential of LLM-based tools in addressing FC challenges, offering actionable insights for educators and institutional leaders in technology-enhanced learning environments.

In the age of rapid internet expansion, social media platforms like Twitter have become crucial for sharing information, expressing emotions, and revealing intentions during crisis situations. They offer crisis responders a means to assess public sentiment, attitudes, intentions, and emotional shifts by monitoring crisis-related tweets. To enhance sentiment and emotion classification, we adopt a transformer-based multi-task learning (MTL) approach with attention mechanism, enabling simultaneous handling of both tasks, and capitalizing on task interdependencies. Incorporating attention mechanism allows the model to concentrate on important words that strongly convey sentiment and emotion. We compare three baseline models, and our findings show that BERTweet outperforms the standard BERT model and exhibits similar performance to RoBERTa in crisis tweets. Furthermore, we employ natural language processing techniques to extract key subject entities (e.g., police, victims) and leverage the publicly available commonsense knowledge model, COMET-ATOMIC 2020, to identify their intentions in given crisis scenarios. Evaluation of COMET-ATOMIC 2020 on subject-based intent prediction in crisis tweets reveals that BART was superior to GPT2-XL model, providing crisis responders with vital information for better decision making. Notably, the integration of sentiment and emotion classification, identification of attention words and subject-based intent prediction represents a novel methodology, not previously applied in the context of crisis scenarios.

Having students write short self-reflections at the end of each weekly session enables them to reflect on what they have learnt in the session and topics they find challenging. Analysing these self-reflections provides instructors with insights on how to address the missing conceptions and misconceptions of the students and appropriately plan and deliver the next session. Currently, manual methods adopted to analyse these student reflections are time consuming and tedious. This paper proposes a solution model that uses content mining and NLP techniques to automate the analysis of short self-reflections. We evaluate the solution model by studying its implementation in an undergraduate Information Systems course through a comparison of three different content mining techniques namely LDA–bigrams, GSDMM-bigrams, and Word2Vec based Clustering models. The evaluation involves both qualitative and quantitative methods. The results show that the proposed techniques are useful in discovering insights from the self-reflections, though the performance varied across the three methods. We provide insights into comparisons of the perspectives, which are useful to instructors.

Traditionally, teaching is usually one directional where the instructor imparts knowledge and there is minimal interaction between learners and instructor. With the focus on learner-centered pedagogy, it can be a challenge to provide timely and relevant guidance to individual learners according to their levels of understanding. One of the options available is to collect reflections from learners after each lesson to extract relevant feedback so that doubts or questions can be addressed in a timely manner. In this paper, we derived an approach to automate the identification of doubts from students’ informal reflections through features analysis, word representation and machine learning. Using reflections as a feedback mechanism and aligning it to the weekly course content can pave the way to a promising approach for learner-centered teaching and personalized learning.