Towards adaptive support for self‐regulated learning of causal relations: Evaluating four Dutch word vector models

British Journal of Educational Technology, Год журнала: 2024, Номер 55(4), С. 1354 - 1375

Опубликована: Янв. 16, 2024

Advances in computational language models increasingly enable adaptive support for self‐regulated learning (SRL) digital environments (DLEs; eg, via automated feedback). However, the accuracy of those is a common concern educational stakeholders (eg, policymakers, researchers, teachers and learners themselves). We compared four Dutch (ie, spaCy medium, large, FastText ConceptNet NumberBatch) context secondary school students' causal relations from expository texts, scaffolded by diagram completion. Since machine relies on human‐labelled data best results, we used dataset with 10,193 answers, compiled over decade research using completion intervention to enhance monitoring their text comprehension. The were combination popular classifiers logistic regression, random forests, vector neural networks) evaluate performance automatically scoring diagrams terms correctness events sequence structure). Five metrics studied, namely accuracy, precision, recall, F 1 area under curve receiver operating characteristic (ROC‐AUC). medium model combined network classifier achieved five metrics, while NumberBatch worked sequence. These evaluation results provide criterion adoption adaptively SRL DLEs. Practitioner notes What already known about this topic Accurate prerequisite effective self‐regulation. Students struggle accurately monitor comprehension texts. Completing improves but there room further improvement. Automatic could be during diagramming. paper adds Comparison word automatic diagrams. above solutions. Evaluation estimating semantic similarity between student answers. Implications practice and/or policy High‐quality (em)power diagramming scoring. evaluated solutions can embedded (DLEs). Criteria increased saliency (in)correct answers might help improve accuracy.

Язык: Английский

---

Exploring the use of metacognitive monitoring cues following a diagram completion intervention

Advances in Health Sciences Education, Год журнала: 2024, Номер 29(4), С. 1323 - 1351

Опубликована: Янв. 29, 2024

Abstract Studying texts constitutes a significant part of student learning in health professions education. Key to from text is the ability effectively monitor one’s own cognitive performance and take appropriate regulatory steps for improvement. Inferential cues generated during experience typically guide this monitoring process. It has been shown that interventions assist learners using comprehension improve their accuracy. One such intervention having complete diagram. Little known, however, about how use shape judgments. In addition, previous research not examined difference cue between categories learners, as good poor monitors. This study explored types patterns used by participants after being subjected diagram completion task prior prediction (PoP). Participants’ thought processes were studied means think-aloud method subsequent PoP. Results suggest relying on comprehension-specific may lead better Poor monitors relied multiple failed available appropriately. They gave more incorrect responses made commission errors diagram, which likely led overconfidence. Good monitors, other hand, utilized are predictive seemed have However, they tended be cautious judgement, probably them underestimate themselves. These observations contribute current understanding effectiveness cue-prompt provide direction future enhancing

Язык: Английский

---

Evaluating Sentence‐BERT‐powered learning analytics for automated assessment of students' causal diagrams

Journal of Computer Assisted Learning, Год журнала: 2024, Номер 40(6), С. 2667 - 2680

Опубликована: Апрель 23, 2024

Abstract Background When learning causal relations, completing diagrams enhances students' comprehension judgements to some extent. To potentially boost this effect, advances in natural language processing (NLP) enable real‐time formative feedback based on the automated assessment of diagrams, which can involve correctness both responses and their position chain. However, responsible adoption effectiveness diagram depend its reliability. Objectives In study, we compare two Dutch pre‐trained models (i.e., RobBERT BERTje) combination with machine‐learning classifiers—Support Vector Machine (SVM) Neural Networks (NN), terms different indicators We also contrast techniques semantic similarity machine learning) for estimating correct a student response Methods For training evaluation models, capitalize human‐labelled dataset containing 2900+ completed by 700+ secondary school students, accumulated from previous diagramming experiments. Results Conclusions predicting responses, 86% accuracy Cohen's κ 0.69 were reached, combinations using SVM being roughly three‐times faster (important applications) than NN counterparts. 92% 0.89 reached. Implications Taken together, these figures equip educational designers decision‐making when NLP‐powered analytics are warranted relation learning; thereby enabling learners reducing teachers' workload.

Язык: Английский

---

Enhancing Student Reflections with Natural Language Processing based Scaffolding: A Quasi-Experimental Study in a Large Lecture Course

Computers and Education Artificial Intelligence, Год журнала: 2025, Номер unknown, С. 100397 - 100397

Опубликована: Март 1, 2025

Язык: Английский

---

Causal diagramming to improve students' monitoring accuracy and text comprehension: Effects of diagram standards and self‐scoring instructions

Applied Cognitive Psychology, Год журнала: 2024, Номер 38(1)

Опубликована: Янв. 1, 2024

Abstract Students' monitoring of their text comprehension must be accurate for self‐regulated learning to effective. Completing causal diagrams after reading (i.e., diagramming) already improves students' accuracy some extent. We investigated whether providing secondary school students with a standard correctly completed) diagram and self‐scoring instructions would further improve in delayed (Experiment 1; n = 98) or immediate 2; 177) diagramming design. Self‐scoring did not compared the control condition(s) either experiment. Presumably, self‐scored even without do so. In contrast findings from prior research standards, an explorative analysis suggests that produce differences accuracy. Immediate diagramming, however, led better than may therefore preferable over under certain conditions.

Язык: Английский

---

Evaluating computer science students reading comprehension of educational multimedia-enhanced text using scalable eye-tracking methodology

Smart Learning Environments, Год журнала: 2024, Номер 11(1)

Опубликована: Июнь 21, 2024

Abstract In this research, a mixed-method approach was employed to conduct large-scale eye-tracking measurements, traditionally associated with high costs and extensive time commitments. Utilizing consumer-grade webcams in conjunction open-source software, data collected from an expansive cohort of students, thereby demonstrating the scalability cost-effectiveness innovative methodology. The primary objective research discern disparities reading behaviour when students were presented standard text accompanied by illustrations, compared same highlighted key terms. participants, comprised first-year university completed questionnaire introductory test ascertain their knowledge level. Subsequently, they segregated into two groups participated sessions, during which ocular movements recorded. amassed underwent both qualitative analyses, facilitated visualizations, quantitative analysis, employing statistical measures on results. Notably, no significant difference observed gaze patterns or results between experimental control groups. However, divergence identified high-achieving those experiencing difficulties, as evidenced averaged composite heatmaps generated data. findings underscore pivotal points. Firstly, feasibility conducting experiments is demonstrated. Traditional studies field often employ small population samples due financial constraints methods that utilize specialized hardware. contrast, our methodology scalable, relying low-end hardware enabling record personal devices. Secondly, while may not provide substantial benefits for fine-tuning already optimized readability, it could serve valuable tool identifying assisting learners who are struggling. This holds potential revolutionize interpretation within educational settings.

Язык: Английский

---

Introduction to special issue on facilitating self-regulated learning with scaffolds: Recent advances and future directions

Metacognition and Learning, Год журнала: 2023, Номер 18(3), С. 623 - 629

Опубликована: Ноя. 6, 2023

Abstract It is important for learners to engage in self-regulated learning (SRL), as it predicts academic achievement a wide range of disciplines. However, SRL can be difficult enact. Therefore, scaffolds have been designed support SRL. In our introductory article this special issue on facilitating with scaffolds, we present framework categorize different place the contributions framework, highlights from contributions, and conclude discussion designing facilitate

Язык: Английский

---

Voorbij keuzestress

Опубликована: Янв. 1, 2024

---