Learning and Instruction最新文献_第5页

Effects of a mathematical language intervention on mathematical development in preschool dual language learners 数学语言干预对学龄前双语学习者数学发展的影响

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-29 DOI: 10.1016/j.learninstruc.2025.102269

Eylül Turan , Davina Van den Broek , Elke Peters , Bert De Smedt , Joke Torbeyns

Background

Intervention studies suggested that instruction in mathematical language enhances preschoolers’ early numerical competencies, with mixed effects. This research focused on single language learners (SLLs) and rarely included dual language learners (DLLs). Because DLLs have been found to have limited knowledge of quantitative language, we do not know whether interventions on mathematical language are equally or even more effective for DLLs.

Aims

This study investigated the effectiveness of a teacher-delivered quantitative language intervention on preschoolers' quantitative language and numerical competencies. We determined whether children's language status (i.e., DLLs versus SLLs) influenced potential intervention effects.

Sample

Participants were 154 children (aged 3- to 5-year-old).

Methods

Schools were randomly assigned to a quantitative language intervention or an active control condition. Children in the intervention condition received three picture books read by their teacher with embedded quantitative language content. The active control group received similar books without quantitative language content. Classroom teachers had to read each book 3 times over 3 weeks. Children were tested before and after the intervention on quantitative language and numerical competencies.

Results

The intervention had positive effects on DLLs' quantitative language, η_p² = .049. We observed a positive effect on some SLLs' quantitative language, but only when children who performed at maximum on the pretest were discarded. There were no significant effects of the intervention on children's numerical competencies.

Conclusion

Findings indicate the efficacy of quantitative language interventions on mathematical language skills. More work is needed to understand how to best promote children's and DLLs' numerical competencies.

背景：干预研究表明，数学语言教学可以提高学龄前儿童的早期数字能力，但效果好坏参半。这项研究主要集中在单语学习者（sll）身上，很少涉及双语学习者（dll）。由于发现dll对定量语言的了解有限，我们不知道对数学语言的干预是否对dll同样有效，甚至更有效。目的本研究探讨了教师提供的定量语言干预对学龄前儿童定量语言和数字能力的影响。我们确定了儿童的语言状态（即，非语言语言与非语言语言）是否会影响潜在的干预效果。样本参与者为154名儿童（3- 5岁）。方法将学校随机分为定量语言干预组和主动对照组。干预组的孩子收到了三本绘本，绘本由老师朗读，绘本中嵌入了定量语言内容。积极对照组接受类似的书籍，但没有定量的语言内容。课堂教师必须在三周内将每本书读三遍。在干预前后对儿童进行定量语言和数字能力测试。结果干预对患者定量语言能力有显著的正向影响，χ 2 = 0.049。我们观察到对一些特殊语言学习者的定量语言有积极的影响，但只有当在前测中表现最好的孩子被丢弃时。干预对儿童的数字能力没有显著影响。结论定量语言干预对数学语言技能的影响是有效的。我们需要做更多的工作来了解如何最好地促进儿童和dll的数字能力。

{"title":"Effects of a mathematical language intervention on mathematical development in preschool dual language learners","authors":"Eylül Turan , Davina Van den Broek , Elke Peters , Bert De Smedt , Joke Torbeyns","doi":"10.1016/j.learninstruc.2025.102269","DOIUrl":"10.1016/j.learninstruc.2025.102269","url":null,"abstract":"<div><h3>Background</h3><div>Intervention studies suggested that instruction in mathematical language enhances preschoolers’ early numerical competencies, with mixed effects. This research focused on single language learners (SLLs) and rarely included dual language learners (DLLs). Because DLLs have been found to have limited knowledge of quantitative language, we do not know whether interventions on mathematical language are equally or even more effective for DLLs.</div></div><div><h3>Aims</h3><div>This study investigated the effectiveness of a teacher-delivered quantitative language intervention on preschoolers' quantitative language and numerical competencies. We determined whether children's language status (i.e., DLLs versus SLLs) influenced potential intervention effects.</div></div><div><h3>Sample</h3><div>Participants were 154 children (aged 3- to 5-year-old).</div></div><div><h3>Methods</h3><div>Schools were randomly assigned to a quantitative language intervention or an active control condition. Children in the intervention condition received three picture books read by their teacher with embedded quantitative language content. The active control group received similar books without quantitative language content. Classroom teachers had to read each book 3 times over 3 weeks. Children were tested before and after the intervention on quantitative language and numerical competencies.</div></div><div><h3>Results</h3><div>The intervention had positive effects on DLLs' quantitative language, η<sub>p</sub><sup>2</sup> = .049. We observed a positive effect on some SLLs' quantitative language, but only when children who performed at maximum on the pretest were discarded. There were no significant effects of the intervention on children's numerical competencies.</div></div><div><h3>Conclusion</h3><div>Findings indicate the efficacy of quantitative language interventions on mathematical language skills. More work is needed to understand how to best promote children's and DLLs' numerical competencies.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102269"},"PeriodicalIF":4.9,"publicationDate":"2025-11-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145684391","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Maintaining the outcomes of meaningful learning: Where are we and where do we need to go? An introduction to the special issue on lasting learningx☆ 保持有意义的学习成果：我们在哪里，我们需要去哪里？《持久学习》专刊简介

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-28 DOI: 10.1016/j.learninstruc.2025.102277

Julian Roelle , Tobias Richter

引用次数: 0

Utilizing AI and machine learning in vocal training programs: A contemporary approach 在声乐训练项目中利用人工智能和机器学习：一种当代方法

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-27 DOI: 10.1016/j.learninstruc.2025.102278

Di Wu

Background

The integration of machine learning methods into vocal training has emerged as a transformative approach in music education. Artificial intelligence (AI) is increasingly utilized to enhance vocal instruction, offering personalized and adaptive learning experiences. However, the effectiveness of AI-based methods compared to traditional approaches requires further exploration.

Aims

This study aims to evaluate the impact of AI-driven vocal training programs on the development of key vocal skills, accessibility, adaptability, and personalized learning. The research seeks to determine whether AI-based instruction leads to measurable improvements in vocal performance.

Sample

The study analyzed a diverse group of vocal students (N = 120) with varying levels of experience. Participants were divided into two groups: one receiving AI-assisted training and the other following traditional vocal instruction methods.

Methods

A comparative analysis was conducted using quantitative assessments of vocal skill development. Key parameters, including articulation, diction, breath control, and intonation accuracy, were evaluated. AI-based training programs were also assessed for their adaptability and effectiveness using standardized scoring metrics.

Results

Findings indicate a statistically significant improvement in vocal training outcomes with AI-based methods (p < 0.001). Notable advantages were observed in accessibility (t = 11.18), adaptability (t = 10.51), and personalized learning (t = 9.45). AI training programs received high effectiveness scores (8.7–9.5/10) and adaptability ratings (7.9–9.7/10). However, limitations included reduced live performance interaction (9/10) and challenges in handling emotional nuances (8/10).

Conclusions

AI-driven vocal training offers significant benefits, enhancing accessibility, skill development, and learning efficiency. Despite some limitations, AI presents a promising tool for modernizing vocal education and accelerating student progress.

将机器学习方法整合到声乐训练中已经成为音乐教育的一种变革性方法。人工智能（AI）越来越多地用于增强声乐教学，提供个性化和自适应的学习体验。然而，与传统方法相比，基于人工智能的方法的有效性需要进一步探索。本研究旨在评估人工智能驱动的声乐训练项目对关键声乐技能发展、可及性、适应性和个性化学习的影响。这项研究旨在确定基于人工智能的教学是否会导致声音表现的显著改善。该研究分析了一组不同的声乐学生（N = 120），他们有不同程度的经验。参与者被分为两组：一组接受人工智能辅助训练，另一组接受传统的声乐教学方法。方法采用嗓音技能发展的定量评价方法进行对比分析。评估关键参数，包括发音、措辞、呼吸控制和语调准确性。基于人工智能的培训项目还使用标准化评分指标评估了它们的适应性和有效性。研究结果表明，基于人工智能的方法在语音训练结果上有统计学上显著的改善（p < 0.001）。在可及性（t = 11.18）、适应性（t = 10.51）和个性化学习（t = 9.45）方面均有显著优势。人工智能训练项目获得了较高的有效性评分（8.7-9.5/10）和适应性评分（7.9-9.7/10）。然而，限制包括减少现场表演互动（9/10）和处理情感细微差别的挑战（8/10）。结论人工智能驱动的声乐训练具有显著的效果，提高了可及性、技能发展和学习效率。尽管存在一些局限性，但人工智能是一种很有前途的工具，可以使声乐教育现代化，加快学生的进步。

{"title":"Utilizing AI and machine learning in vocal training programs: A contemporary approach","authors":"Di Wu","doi":"10.1016/j.learninstruc.2025.102278","DOIUrl":"10.1016/j.learninstruc.2025.102278","url":null,"abstract":"<div><h3>Background</h3><div>The integration of machine learning methods into vocal training has emerged as a transformative approach in music education. Artificial intelligence (AI) is increasingly utilized to enhance vocal instruction, offering personalized and adaptive learning experiences. However, the effectiveness of AI-based methods compared to traditional approaches requires further exploration.</div></div><div><h3>Aims</h3><div>This study aims to evaluate the impact of AI-driven vocal training programs on the development of key vocal skills, accessibility, adaptability, and personalized learning. The research seeks to determine whether AI-based instruction leads to measurable improvements in vocal performance.</div></div><div><h3>Sample</h3><div>The study analyzed a diverse group of vocal students (N = 120) with varying levels of experience. Participants were divided into two groups: one receiving AI-assisted training and the other following traditional vocal instruction methods.</div></div><div><h3>Methods</h3><div>A comparative analysis was conducted using quantitative assessments of vocal skill development. Key parameters, including articulation, diction, breath control, and intonation accuracy, were evaluated. AI-based training programs were also assessed for their adaptability and effectiveness using standardized scoring metrics.</div></div><div><h3>Results</h3><div>Findings indicate a statistically significant improvement in vocal training outcomes with AI-based methods (p < 0.001). Notable advantages were observed in accessibility (t = 11.18), adaptability (t = 10.51), and personalized learning (t = 9.45). AI training programs received high effectiveness scores (8.7–9.5/10) and adaptability ratings (7.9–9.7/10). However, limitations included reduced live performance interaction (9/10) and challenges in handling emotional nuances (8/10).</div></div><div><h3>Conclusions</h3><div>AI-driven vocal training offers significant benefits, enhancing accessibility, skill development, and learning efficiency. Despite some limitations, AI presents a promising tool for modernizing vocal education and accelerating student progress.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102278"},"PeriodicalIF":4.9,"publicationDate":"2025-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

In their hands: Multimodal learning analytics as reflective-practice resources empowering mathematics teachers’ professional development 在他们手中：多模态学习分析作为反思性实践资源，促进数学教师的专业发展

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-26 DOI: 10.1016/j.learninstruc.2025.102272

Alik Palatnik , Catalina Lomos , Dor Abrahamson

Background

Educational researchers have been using multimodal data (e.g., eye-tracking) to study teachers’ instructional practices. Yet, by and large, these technologies have not been handed over to teachers as resources for self-reflection.

Aims

This exploratory study aimed to develop and evaluate novel methodologies for supporting teachers in using multimodal data, specifically eye-tracking-overlaid multiple-point-of-view videos, as resources for deepening their understanding of learning and teaching.

Sample and methods

Five pre-service mathematics teachers (PSMT) participated in this study. First, they collaboratively constructed a body-scale polyhedron while wearing eye-tracking devices. In a subsequent structured reflection session, the participants collectively analyzed eye-tracking-overlaid multiple-point-of-view recordings of their construction activity. Drawing on Goodwin's Co-Operative Action theory, the researchers then qualitatively micro-analyzed the PSMTs' individual and interactive multimodal behaviors in both activities.

Results

The technology enabled the PSMTs to highlight, code, and elaborate on nuanced aspects of individuals’ perceptual and social behaviors during the construction activity. These insights led to a spontaneous generalization of pedagogical implications for their own prospective facilitation of classroom group work.

Conclusions

Supplementing PSMTs' video-based reflection on their own activity with multiple-point-of-view gaze data from their mobile eye trackers enriches the PSMTs’ microgenetic analysis of their own collaborative multimodal learning process, in turn potentially leading more generally to pedagogical inferences for their own prospective professional practice.

教育研究者一直在使用多模态数据（如眼动追踪）来研究教师的教学实践。然而，总的来说，这些技术并没有作为自我反思的资源交给教师。本探索性研究旨在开发和评估新的方法，以支持教师使用多模态数据，特别是眼动追踪覆盖的多视角视频，作为加深他们对学习和教学理解的资源。样本与方法5名职前数学教师参与本研究。首先，他们在佩戴眼球追踪设备的同时合作构建了一个身体尺度的多面体。在随后的结构化反思环节中，参与者集体分析了眼动追踪覆盖的多视角建筑活动记录。根据古德温的合作行为理论，研究人员定性地微观分析了psmt在两种活动中的个体和互动多模态行为。结果该技术使psmt能够突出、编码和详细说明个体在构建活动中的感知和社会行为的细微方面。这些见解导致了他们自己对课堂小组工作的前瞻性促进的教学含义的自发概括。结论：用移动眼动仪的多视角注视数据补充psmt对自己活动的基于视频的反思，丰富了psmt对自己的协作多模式学习过程的微遗传学分析，进而可能为他们自己未来的专业实践提供更普遍的教学推断。

{"title":"In their hands: Multimodal learning analytics as reflective-practice resources empowering mathematics teachers’ professional development","authors":"Alik Palatnik , Catalina Lomos , Dor Abrahamson","doi":"10.1016/j.learninstruc.2025.102272","DOIUrl":"10.1016/j.learninstruc.2025.102272","url":null,"abstract":"<div><h3>Background</h3><div>Educational researchers have been using multimodal data (e.g., eye-tracking) to study teachers’ instructional practices. Yet, by and large, these technologies have not been handed over to teachers as resources for self-reflection.</div></div><div><h3>Aims</h3><div>This exploratory study aimed to develop and evaluate novel methodologies for supporting teachers in using multimodal data, specifically eye-tracking-overlaid multiple-point-of-view videos, as resources for deepening their understanding of learning and teaching.</div></div><div><h3>Sample and methods</h3><div>Five pre-service mathematics teachers (PSMT) participated in this study. First, they collaboratively constructed a body-scale polyhedron while wearing eye-tracking devices. In a subsequent structured reflection session, the participants collectively analyzed eye-tracking-overlaid multiple-point-of-view recordings of their construction activity. Drawing on Goodwin's Co-Operative Action theory, the researchers then qualitatively micro-analyzed the PSMTs' individual and interactive multimodal behaviors in both activities.</div></div><div><h3>Results</h3><div>The technology enabled the PSMTs to highlight, code, and elaborate on nuanced aspects of individuals’ perceptual and social behaviors during the construction activity. These insights led to a spontaneous generalization of pedagogical implications for their own prospective facilitation of classroom group work.</div></div><div><h3>Conclusions</h3><div>Supplementing PSMTs' video-based reflection on their own activity with multiple-point-of-view gaze data from their mobile eye trackers enriches the PSMTs’ microgenetic analysis of their own collaborative multimodal learning process, in turn potentially leading more generally to pedagogical inferences for their own prospective professional practice.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102272"},"PeriodicalIF":4.9,"publicationDate":"2025-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Experimental evidence on the effects of preservice teachers’ growth and fixed mindsets on teaching self-efficacy and anticipated instructional practices 职前教师成长和固定心态对教学自我效能感和预期教学实践影响的实验证据

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-22 DOI: 10.1016/j.learninstruc.2025.102266

Patricia Schwiering , Anke Heyder

Aims

Growth versus fixed mindsets refer to beliefs about the malleability (growth mindset) or stability (fixed mindset) of individual characteristics. While prior research has focused on students' mindsets, experimental studies targeting teachers' mindsets about student abilities are scarce. We investigated the effects of a brief growth-mindset intervention on preservice teachers' school career counseling, expectations, anticipated teaching self-efficacy, emotions, stress, and teaching practices.

Methods

In a preregistered experiment, we randomly assigned 306 preservice teachers to the intervention group or control group. The intervention group reflected on their mission as teachers which fostered growth mindsets in earlier research. Both groups then reported their mindset, read the vignette of a fictitious low-achieving student, reported their expectations, wrote a school career counseling letter to the student, and reported their anticipated teaching self-efficacy, emotions, stress, and teaching practices.

Results

The intervention group reported a stronger growth mindset than the control group, higher teaching self-efficacy expectations, a higher likelihood of using instructional practices that foster cognitive stimulation and autonomy, and a lower likelihood of using performance practices. Structural equation modeling showed that affecting preservice teachers' mindsets changed their anticipated teaching self-efficacy and practices. Most school career counseling texts were growth mindset oriented. The intervention and control group did not differ in their texts’ overall mindset orientation, though the intervention group used more sentences and offered descriptively more help.

Conclusion

Fostering growth mindsets in preservice teachers affected their teaching self-efficacy, anticipated instructional practices, and–to some extent–the way they counsel low-achieving students.

成长与固定心态是指对个人特征的可塑性（成长心态）或稳定性（固定心态）的信念。虽然先前的研究主要集中在学生的心态上，但针对教师对学生能力的心态的实验研究很少。研究了成长心态干预对职前教师学校职业咨询、期望、预期教学自我效能感、情绪、压力和教学实践的影响。方法采用预注册实验，将306名职前教师随机分为干预组和对照组。干预组反思了他们作为教师的使命，在早期的研究中培养了成长心态。然后，两组都报告了他们的心态，阅读了一个虚构的成绩差的学生的小短文，报告了他们的期望，给学生写了一封学校职业咨询信，报告了他们对教学自我效能感、情绪、压力和教学实践的预期。结果干预组学生成长心态较对照组强，教学自我效能预期较高，采用促进认知刺激和自主的教学实践的可能性较高，而采用绩效实践的可能性较低。结构方程模型显示，影响职前教师心态会改变其预期教学自我效能感和教学实践。大多数学校职业咨询文本以成长心态为导向。干预组和对照组在文本的整体思维取向上没有差异，尽管干预组使用了更多的句子，并提供了更多的描述性帮助。结论培养职前教师成长型思维会影响其教学自我效能感、预期教学实践，并在一定程度上影响其辅导低年级学生的方式。

{"title":"Experimental evidence on the effects of preservice teachers’ growth and fixed mindsets on teaching self-efficacy and anticipated instructional practices","authors":"Patricia Schwiering , Anke Heyder","doi":"10.1016/j.learninstruc.2025.102266","DOIUrl":"10.1016/j.learninstruc.2025.102266","url":null,"abstract":"<div><h3>Aims</h3><div>Growth versus fixed mindsets refer to beliefs about the malleability (<em>growth mindset</em>) or stability (<em>fixed mindset</em>) of individual characteristics. While prior research has focused on students' mindsets, experimental studies targeting teachers' mindsets about student abilities are scarce. We investigated the effects of a brief growth-mindset intervention on preservice teachers' school career counseling, expectations, anticipated teaching self-efficacy, emotions, stress, and teaching practices.</div></div><div><h3>Methods</h3><div>In a preregistered experiment, we randomly assigned 306 preservice teachers to the intervention group or control group. The intervention group reflected on their mission as teachers which fostered growth mindsets in earlier research. Both groups then reported their mindset, read the vignette of a fictitious low-achieving student, reported their expectations, wrote a school career counseling letter to the student, and reported their anticipated teaching self-efficacy, emotions, stress, and teaching practices.</div></div><div><h3>Results</h3><div>The intervention group reported a stronger growth mindset than the control group, higher teaching self-efficacy expectations, a higher likelihood of using instructional practices that foster cognitive stimulation and autonomy, and a lower likelihood of using performance practices. Structural equation modeling showed that affecting preservice teachers' mindsets changed their anticipated teaching self-efficacy and practices. Most school career counseling texts were growth mindset oriented. The intervention and control group did not differ in their texts’ overall mindset orientation, though the intervention group used more sentences and offered descriptively more help.</div></div><div><h3>Conclusion</h3><div>Fostering growth mindsets in preservice teachers affected their teaching self-efficacy, anticipated instructional practices, and–to some extent–the way they counsel low-achieving students.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102266"},"PeriodicalIF":4.9,"publicationDate":"2025-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145616715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Applying multimodal learning analytics to naturalistic recordings of clinical simulations: Towards an accurate and scalable pipeline for automated feedback generation 将多模态学习分析应用于临床模拟的自然记录：为自动反馈生成精确和可扩展的管道

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-22 DOI: 10.1016/j.learninstruc.2025.102267

Vitaliy Popov , Steve Nguyen , Xavier Ochoa

Background

Educational audiovisual recordings, spanning domains from teacher training to clinical simulations, often remain underutilized due to the intensive human labor required for comprehensive analysis and timely feedback.

Aims

This study addresses this challenge by developing and evaluating a multimodal learning analytics (MmLA) pipeline that integrates state-of-the-art visual, speech, and language models to automatically capture complex communication skills.

Sample

244 medical and social work students practicing breaking bad news in a team-based simulated clinical scenario.

Methods

We developed and evaluated the MmLA pipeline, which integrates speech extraction, gaze tracking, and semantic analysis using a Large Language Model (LLM). We empirically evaluated the pipeline by measuring the accuracy of each feature-extraction stage, including diarized transcript generation, visual-attention pattern tracking, and LLM-based semantic evaluation of discourse, against human assessors, and by comparing the final automated predictions with human-generated scores.

Results

The developed MmLA pipeline effectively extracts and fuses features from naturalistic videos of standardized patient simulations, approaching human-level accuracy, though diarization remains challenging with a moderate 29.9 % SER (lower is better). The predictive model using behavioral, verbal, and semantic features achieved 81.82 % accuracy in assessing interaction comfort, with eye contact emerging as the most influential predictor.

Conclusions

By leveraging the most common multimodal data sources, such as video and audio, the study provides a scalable methodological blueprint that can be adapted to diverse educational settings to foster communication skills.

从教师培训到临床模拟的教学视听记录，由于需要大量的人力来进行全面分析和及时反馈，因此往往没有得到充分利用。本研究通过开发和评估多模态学习分析（MmLA）管道来解决这一挑战，该管道集成了最先进的视觉、语音和语言模型，以自动捕获复杂的沟通技巧。244名医学和社会工作专业的学生在一个以团队为基础的模拟临床场景中练习打破坏消息。方法我们开发并评估了MmLA管道，该管道使用大型语言模型（LLM）集成了语音提取，凝视跟踪和语义分析。我们通过测量每个特征提取阶段（包括日记化转录生成、视觉注意模式跟踪和基于llm的话语语义评估）的准确性来对管道进行实证评估，并将最终的自动预测与人工生成的分数进行比较。开发的MmLA管道有效地从标准化患者模拟的自然视频中提取和融合特征，接近人类水平的精度，尽管分割仍然具有挑战性，SER为29.9%（越低越好）。使用行为、语言和语义特征的预测模型在评估交互舒适度方面达到了81.82%的准确率，其中眼神接触是最具影响力的预测因素。通过利用最常见的多模式数据源，如视频和音频，该研究提供了一个可扩展的方法蓝图，可以适应不同的教育环境，以培养沟通技能。

{"title":"Applying multimodal learning analytics to naturalistic recordings of clinical simulations: Towards an accurate and scalable pipeline for automated feedback generation","authors":"Vitaliy Popov , Steve Nguyen , Xavier Ochoa","doi":"10.1016/j.learninstruc.2025.102267","DOIUrl":"10.1016/j.learninstruc.2025.102267","url":null,"abstract":"<div><h3>Background</h3><div>Educational audiovisual recordings, spanning domains from teacher training to clinical simulations, often remain underutilized due to the intensive human labor required for comprehensive analysis and timely feedback.</div></div><div><h3>Aims</h3><div>This study addresses this challenge by developing and evaluating a multimodal learning analytics (MmLA) pipeline that integrates state-of-the-art visual, speech, and language models to automatically capture complex communication skills.</div></div><div><h3>Sample</h3><div>244 medical and social work students practicing breaking bad news in a team-based simulated clinical scenario.</div></div><div><h3>Methods</h3><div>We developed and evaluated the MmLA pipeline, which integrates speech extraction, gaze tracking, and semantic analysis using a Large Language Model (LLM). We empirically evaluated the pipeline by measuring the accuracy of each feature-extraction stage, including diarized transcript generation, visual-attention pattern tracking, and LLM-based semantic evaluation of discourse, against human assessors, and by comparing the final automated predictions with human-generated scores.</div></div><div><h3>Results</h3><div>The developed MmLA pipeline effectively extracts and fuses features from naturalistic videos of standardized patient simulations, approaching human-level accuracy, though diarization remains challenging with a moderate 29.9 % SER (lower is better). The predictive model using behavioral, verbal, and semantic features achieved 81.82 % accuracy in assessing interaction comfort, with eye contact emerging as the most influential predictor.</div></div><div><h3>Conclusions</h3><div>By leveraging the most common multimodal data sources, such as video and audio, the study provides a scalable methodological blueprint that can be adapted to diverse educational settings to foster communication skills.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102267"},"PeriodicalIF":4.9,"publicationDate":"2025-11-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571195","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

STEM class momentary investigation: Teaching and learning engagement STEM课堂瞬间调查：教与学的参与

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-21 DOI: 10.1016/j.learninstruc.2025.102270

Ssu-Kuang Chen, Sunny S.J. Lin

Background

In STEM project-based learning (PjBL) classrooms, hands-on and inquiry-based activities foster learning and attention. Given the dynamic and time-constrained nature of these classes, understanding student learning engagement is crucial.

Aims

We aimed to identify teaching approaches through moment-to-moment classroom observations and examine how STEM self-efficacy, prior knowledge, gender, and teaching approaches influenced learning engagement.

Sample

Observations included 78 classes with 312 students from two elementary schools, seven middle schools, and one high school in Taiwan. Surveys were collected from 179 students (58 % males) across 49 classes.

Methods

Student engagement and teaching behaviors were recorded every minute. Cluster analysis classified teacher behaviors into distinct teaching approaches, and the proportion of class minutes devoted to each cluster was calculated. Multilevel linear models were used to test how class-level teaching approaches and student-level factors predicted five engagement types: passive, active, constructive, interactive, and irrelevant behaviors.

Results

Two teaching approaches emerged: teacher-centered and student-centered. Student-centered activities were associated with more constructive, fewer passive behaviors. Prior knowledge positively influenced constructive and interactive engagement, while reducing irrelevant behaviors. Student-centered activities weakened the positive association between prior knowledge and active engagement, while strengthening the relationship between self-efficacy and interactive engagement. Females were more passive, whereas males were more constructive.

Conclusions

Student-centered activities, prior knowledge, and male gender contributed to higher-order engagement in STEM-PjBL classes. Student-centered teaching moderated learning processes by weakening the link between prior knowledge and active engagement while strengthening the connection between self-efficacy and interactive engagement. Practical suggestions were offered based on these findings.

在基于STEM项目的学习（PjBL）课堂中，动手和探究性活动促进了学习和注意力。考虑到这些课程的动态性和时间有限性，了解学生的学习参与度至关重要。我们旨在通过实时的课堂观察来确定教学方法，并研究STEM自我效能感、先验知识、性别和教学方法如何影响学习参与。样本观察包括台湾地区两所小学、七所初中及一所高中的78个班级共312名学生。调查收集了来自49个班级的179名学生（58%为男性）。方法每分钟记录学生参与情况和教学行为。聚类分析将教师行为划分为不同的教学方法，并计算每个类的课堂时间占比。使用多层线性模型来测试课堂水平教学方法和学生水平因素如何预测五种参与类型：被动，主动，建设性，互动和无关行为。结果形成了以教师为中心和以学生为中心的两种教学模式。以学生为中心的活动更有建设性，消极行为更少。先验知识积极影响建设性和互动性参与，同时减少无关行为。以学生为中心的活动削弱了先验知识与主动投入的正相关关系，而强化了自我效能感与互动投入的关系。女性更被动，而男性更有建设性。结论以学生为中心的活动、先验知识和男性性别对STEM-PjBL课堂的高阶参与度有影响。以学生为中心的教学弱化了先验知识与主动参与之间的联系，强化了自我效能感与互动参与之间的联系，从而调节了学习过程。在此基础上提出了切实可行的建议。

{"title":"STEM class momentary investigation: Teaching and learning engagement","authors":"Ssu-Kuang Chen, Sunny S.J. Lin","doi":"10.1016/j.learninstruc.2025.102270","DOIUrl":"10.1016/j.learninstruc.2025.102270","url":null,"abstract":"<div><h3>Background</h3><div>In STEM project-based learning (PjBL) classrooms, hands-on and inquiry-based activities foster learning and attention. Given the dynamic and time-constrained nature of these classes, understanding student learning engagement is crucial.</div></div><div><h3>Aims</h3><div>We aimed to identify teaching approaches through moment-to-moment classroom observations and examine how STEM self-efficacy, prior knowledge, gender, and teaching approaches influenced learning engagement.</div></div><div><h3>Sample</h3><div>Observations included 78 classes with 312 students from two elementary schools, seven middle schools, and one high school in Taiwan. Surveys were collected from 179 students (58 % males) across 49 classes.</div></div><div><h3>Methods</h3><div>Student engagement and teaching behaviors were recorded every minute. Cluster analysis classified teacher behaviors into distinct teaching approaches, and the proportion of class minutes devoted to each cluster was calculated. Multilevel linear models were used to test how class-level teaching approaches and student-level factors predicted five engagement types: passive, active, constructive, interactive, and irrelevant behaviors.</div></div><div><h3>Results</h3><div>Two teaching approaches emerged: teacher-centered and student-centered. Student-centered activities were associated with more constructive, fewer passive behaviors. Prior knowledge positively influenced constructive and interactive engagement, while reducing irrelevant behaviors. Student-centered activities weakened the positive association between prior knowledge and active engagement, while strengthening the relationship between self-efficacy and interactive engagement. Females were more passive, whereas males were more constructive.</div></div><div><h3>Conclusions</h3><div>Student-centered activities, prior knowledge, and male gender contributed to higher-order engagement in STEM-PjBL classes. Student-centered teaching moderated learning processes by weakening the link between prior knowledge and active engagement while strengthening the connection between self-efficacy and interactive engagement. Practical suggestions were offered based on these findings.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"102 ","pages":"Article 102270"},"PeriodicalIF":4.9,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145555348","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Effects of instructional design, instructional preferences, and cognitive load on problem solving and knowledge acquisition in a computer-based office simulation 基于计算机的办公室模拟教学设计、教学偏好和认知负荷对问题解决和知识获取的影响

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-18 DOI: 10.1016/j.learninstruc.2025.102255

Sabrina Ludwig , Andreas Rausch , Michelle Taub

Background

Instructional designs based on problem solving and self-regulation have been extensively studied in the context of computer-based learning environments. However, the question of how problem-solving phases should be embedded in instructional designs remains. Several studies, especially in STEM, suggest that ‘direct instruction followed by problem-solving (DI-PS)’ benefits procedural knowledge acquisition while the instructional design ‘problem-solving followed by direct instruction’ (PS-DI) benefits conceptual knowledge.

Aims

This study aimed to investigate the effects of DI-PS and PS-DI, learners’ instructional preferences, and cognitive load on problem-solving and knowledge acquisition in a computer-based office simulation.

Sample(s)

Eighty-one German undergraduate business education students participated in a pre-post-test experimental study, randomly assigned to either the DI-PS or PS-DI condition. The students worked on a supplier selection task within a computer-based office simulation.

Methods

Knowledge tests and questionnaires were employed to measure knowledge acquisition, cognitive load, and learning preferences. Log data analyses as well as variance and regression analyses were conducted.

Results

Results showed that PS-DI led to higher conceptual knowledge gains, whereas DI-PS tended to yield slightly higher problem-solving performance. Both groups improved in conceptual and procedural knowledge, yet no statistically significant differences between the sequencing conditions emerged. The PS-DI group reported higher intrinsic cognitive load during the problem-solving phase. Moreover, the DI-PS group reported higher retrospective satisfaction with the instructional design than the PS-DI group, aligning with existing literature on instructional preferences.

Conclusions

These findings provide insights for improving learning designs, personalised recommendations, and adaptive support in computer-based learning environments to enhance learners' performance. Further research is needed to examine long-term learning and transfer.

基于问题解决和自我调节的教学设计在基于计算机的学习环境中得到了广泛的研究。然而，如何解决问题的阶段应该嵌入教学设计的问题仍然存在。一些研究，特别是在STEM领域，表明“直接指导后解决问题”（DI-PS）有利于程序性知识的获取，而“解决问题后直接指导”（PS-DI）的教学设计有利于概念性知识的获取。目的本研究旨在探讨基于计算机的办公模拟中，学生的教学偏好和认知负荷对学生解决问题和知识习得的影响。81名德国商科本科学生参加了一项测试前-测试后的实验研究，随机分为DI-PS组和PS-DI组。学生们在一个基于计算机的办公室模拟环境中进行供应商选择任务。方法采用知识测验和问卷调查的方法，考察学生的知识获取、认知负荷和学习偏好。对测井数据进行分析，并进行方差分析和回归分析。结果结果表明，PS-DI能获得更高的概念知识，而DI-PS则倾向于获得略高的问题解决能力。两组在概念性和程序性知识方面都有所提高，但排序条件之间没有统计学上的显著差异。PS-DI组在解决问题阶段报告了更高的内在认知负荷。此外，DI-PS组对教学设计的回顾性满意度高于PS-DI组，这与现有的教学偏好文献一致。这些发现为在基于计算机的学习环境中改进学习设计、个性化建议和适应性支持以提高学习者的表现提供了见解。需要进一步的研究来检验长期学习和迁移。

{"title":"Effects of instructional design, instructional preferences, and cognitive load on problem solving and knowledge acquisition in a computer-based office simulation","authors":"Sabrina Ludwig , Andreas Rausch , Michelle Taub","doi":"10.1016/j.learninstruc.2025.102255","DOIUrl":"10.1016/j.learninstruc.2025.102255","url":null,"abstract":"<div><h3>Background</h3><div>Instructional designs based on problem solving and self-regulation have been extensively studied in the context of computer-based learning environments. However, the question of how problem-solving phases should be embedded in instructional designs remains. Several studies, especially in STEM, suggest that ‘direct instruction followed by problem-solving (DI-PS)’ benefits procedural knowledge acquisition while the instructional design ‘problem-solving followed by direct instruction’ (PS-DI) benefits conceptual knowledge.</div></div><div><h3>Aims</h3><div>This study aimed to investigate the effects of DI-PS and PS-DI, learners’ instructional preferences, and cognitive load on problem-solving and knowledge acquisition in a computer-based office simulation.</div></div><div><h3>Sample(s)</h3><div>Eighty-one German undergraduate business education students participated in a pre-post-test experimental study, randomly assigned to either the DI-PS or PS-DI condition. The students worked on a supplier selection task within a computer-based office simulation.</div></div><div><h3>Methods</h3><div>Knowledge tests and questionnaires were employed to measure knowledge acquisition, cognitive load, and learning preferences. Log data analyses as well as variance and regression analyses were conducted.</div></div><div><h3>Results</h3><div>Results showed that PS-DI led to higher conceptual knowledge gains, whereas DI-PS tended to yield slightly higher problem-solving performance. Both groups improved in conceptual and procedural knowledge, yet no statistically significant differences between the sequencing conditions emerged. The PS-DI group reported higher intrinsic cognitive load during the problem-solving phase. Moreover, the DI-PS group reported higher retrospective satisfaction with the instructional design than the PS-DI group, aligning with existing literature on instructional preferences.</div></div><div><h3>Conclusions</h3><div>These findings provide insights for improving learning designs, personalised recommendations, and adaptive support in computer-based learning environments to enhance learners' performance. Further research is needed to examine long-term learning and transfer.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"101 ","pages":"Article 102255"},"PeriodicalIF":4.9,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579108","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Who perseveres in school? Understanding the developmental trajectories, predictors, and consequences of students’ perseverance 谁在学校里坚持不懈？了解学生毅力的发展轨迹、预测因素和后果

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-17 DOI: 10.1016/j.learninstruc.2025.102249

Jiajing Li , Ronnel B. King

Background

Perseverance is a critical facilitator of learning. However, prior research on perseverance has predominantly adopted a domain-general trait perspective, drawing on variable-centered approaches. The role of domain-specific perseverance in the context of schooling has been relatively neglected. Furthermore, the existence of distinct perseverance trajectories among students, their predictors, and consequences remain underexplored.

Aims

This preregistered longitudinal study examined how students’ perseverance changed over time, whether students varied in their perseverance trajectories, and the predictors (social support and personal goals) and consequences (academic achievement) of these trajectories.

Sample and analyses

The study followed 5593 secondary students from 16 schools over three years. Both variable-centered (latent growth curve modeling) and person-centered (growth mixture modeling) approaches were adopted.

Results

The variable-centered analysis indicated no overall change in perseverance. However, the person-centered analysis revealed a nuanced pattern with four distinct trajectories: “low decreasing” (11 %), “low stable” (33 %), “moderate decreasing” (46 %), and “high increasing” (10 %). The most adaptive trajectory was the “high increasing” group, marked by initially high perseverance and subsequent improvements. Teacher support and mastery goals were the strongest predictors of profile membership. Individuals in the “high increasing” profile had the highest academic achievement, while those in the “low decreasing” profile had the lowest.

Conclusion

This study builds on prior research on perseverance, which has primarily drawn on a trait perspective. The findings highlight that perseverance evolves over time, but these changes are not uniform across individuals. Providing teacher support and helping students set mastery goals might be crucial pathways to enhance perseverance.

毅力是学习的关键促进因素。然而，以往对毅力的研究主要采用领域-一般特质视角，采用以变量为中心的方法。在学校教育背景下，特定领域毅力的作用相对被忽视了。此外，学生之间存在明显的毅力轨迹，其预测因素和后果仍未得到充分探讨。目的：这项预先登记的纵向研究考察了学生的毅力如何随着时间的推移而变化，学生的毅力轨迹是否不同，以及这些轨迹的预测因素（社会支持和个人目标）和后果（学业成就）。样本和分析这项研究在三年的时间里跟踪调查了来自16所学校的5593名中学生。采用以变量为中心（潜在生长曲线模型）和以人为中心（生长混合模型）两种方法。结果以变量为中心的分析显示，毅力没有发生总体变化。然而，以人为中心的分析揭示了一个微妙的模式，有四个不同的轨迹：“低下降”（11%），“低稳定”（33%），“中等下降”（46%）和“高增加”（10%）。最具适应性的轨迹是“高增长”组，其特征是最初的高毅力和随后的改进。教师支持和掌握目标是档案成员的最强预测因子。“高增长”个体的学业成绩最高，而“低下降”个体的学业成绩最低。结论本研究建立在先前关于毅力的研究基础上，这些研究主要是从特质角度出发的。研究结果强调，毅力会随着时间的推移而变化，但这些变化在个体之间并不一致。提供教师支持和帮助学生设定精通目标可能是增强毅力的关键途径。

{"title":"Who perseveres in school? Understanding the developmental trajectories, predictors, and consequences of students’ perseverance","authors":"Jiajing Li , Ronnel B. King","doi":"10.1016/j.learninstruc.2025.102249","DOIUrl":"10.1016/j.learninstruc.2025.102249","url":null,"abstract":"<div><h3>Background</h3><div>Perseverance is a critical facilitator of learning. However, prior research on perseverance has predominantly adopted a domain-general trait perspective, drawing on variable-centered approaches. The role of domain-specific perseverance in the context of schooling has been relatively neglected. Furthermore, the existence of distinct perseverance trajectories among students, their predictors, and consequences remain underexplored.</div></div><div><h3>Aims</h3><div>This preregistered longitudinal study examined how students’ perseverance changed over time, whether students varied in their perseverance trajectories, and the predictors (social support and personal goals) and consequences (academic achievement) of these trajectories.</div></div><div><h3>Sample and analyses</h3><div>The study followed 5593 secondary students from 16 schools over three years. Both variable-centered (latent growth curve modeling) and person-centered (growth mixture modeling) approaches were adopted.</div></div><div><h3>Results</h3><div>The variable-centered analysis indicated no overall change in perseverance. However, the person-centered analysis revealed a nuanced pattern with four distinct trajectories: “low decreasing” (11 %), “low stable” (33 %), “moderate decreasing” (46 %), and “high increasing” (10 %). The most adaptive trajectory was the “high increasing” group, marked by initially high perseverance and subsequent improvements. Teacher support and mastery goals were the strongest predictors of profile membership. Individuals in the “high increasing” profile had the highest academic achievement, while those in the “low decreasing” profile had the lowest.</div></div><div><h3>Conclusion</h3><div>This study builds on prior research on perseverance, which has primarily drawn on a trait perspective. The findings highlight that perseverance evolves over time, but these changes are not uniform across individuals. Providing teacher support and helping students set mastery goals might be crucial pathways to enhance perseverance.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"101 ","pages":"Article 102249"},"PeriodicalIF":4.9,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145579109","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Validating automated assessments of teaching effectiveness using multimodal data 使用多模态数据验证教学效果的自动评估

IF 4.9 1区教育学 Q1 EDUCATION & EDUCATIONAL RESEARCH

Learning and Instruction

Pub Date : 2025-11-15 DOI: 10.1016/j.learninstruc.2025.102264

Tim Fütterer , Ruikun Hou , Babette Bühler , Efe Bozkir , Courtney Bell , Enkelejda Kasneci , Peter Gerjets , Ulrich Trautwein

Background

For enhancing student learning in classrooms, high-quality teaching is essential. Research highlighted core dimensions of effective teaching, including classroom management, student support, and cognitive activation. However, traditional methods of assessing teaching effectiveness dimensions (e.g., student surveys) have limitations, including rating biases and resource intensiveness.

Aims

To overcome these challenges, we explored machine learning (ML) algorithms for the automated assessment of teaching effectiveness.

Sample

The study analyzed multimodal data—such as video, audio, and transcripts—from the Global Teaching Insights study, which included video recordings and transcripts from 46 teachers and 1,132 students in Germany.

Method

Scores for 18 teaching effectiveness subdimensions from three core dimensions were automatically generated by training attention-based ML models on multimodal features extracted from pretrained encoders. These ML-generated scores were compared with scores provided by human experts. A content validity study was conducted, where human experts evaluated the plausibility of ML-generated scores against human-generated scores. Structural equation models were used to assess the relationship between teaching effectiveness subdimensions and students’ tested achievement.

Results

ML-generated scores were more reliable for some subdimensions (e.g., nature of discourse), and they were also plausible and content valid. ML-generated scores achieved higher absolute accuracy than human scores in 11 of 18 subdimensions. Limitations include reliance on human ratings as ground truth and inconsistent predictive validity, underscoring the need for refined models to generate actionable insights, such as real-time feedback systems.

Conclusions

The findings provide valuable insights for the development of automated feedback, enhancing the practical application of teaching effectiveness assessments.

为了提高学生在课堂上的学习，高质量的教学是必不可少的。研究强调了有效教学的核心维度，包括课堂管理、学生支持和认知激活。然而，评估教学有效性维度的传统方法（如学生调查）存在局限性，包括评级偏差和资源集约化。为了克服这些挑战，我们探索了用于教学有效性自动评估的机器学习（ML）算法。该研究分析了来自全球教学洞察研究的多模式数据，如视频、音频和成绩单，其中包括来自德国46名教师和1132名学生的视频记录和成绩单。方法通过训练基于注意力的机器学习模型对预训练编码器提取的多模态特征进行训练，自动生成三个核心维度的18个子维度的教学效果得分。这些机器生成的分数与人类专家提供的分数进行了比较。进行了内容效度研究，其中人类专家评估了ml生成分数与人类生成分数的合理性。采用结构方程模型评估教学效能子维度与学生测试成绩之间的关系。结果sml生成的分数在某些子维度（例如，话语的性质）上更可靠，而且它们也是可信的和内容有效的。ml生成的分数在18个子维度中的11个子维度上的绝对准确性高于人类分数。局限性包括依赖于人类的评价作为基础事实和不一致的预测有效性，强调需要改进模型来产生可操作的见解，例如实时反馈系统。结论本研究结果为自动化反馈的发展，加强教学效果评估的实际应用提供了有价值的见解。

{"title":"Validating automated assessments of teaching effectiveness using multimodal data","authors":"Tim Fütterer , Ruikun Hou , Babette Bühler , Efe Bozkir , Courtney Bell , Enkelejda Kasneci , Peter Gerjets , Ulrich Trautwein","doi":"10.1016/j.learninstruc.2025.102264","DOIUrl":"10.1016/j.learninstruc.2025.102264","url":null,"abstract":"<div><h3>Background</h3><div>For enhancing student learning in classrooms, high-quality teaching is essential. Research highlighted core dimensions of effective teaching, including classroom management, student support, and cognitive activation. However, traditional methods of assessing teaching effectiveness dimensions (e.g., student surveys) have limitations, including rating biases and resource intensiveness.</div></div><div><h3>Aims</h3><div>To overcome these challenges, we explored machine learning (ML) algorithms for the automated assessment of teaching effectiveness.</div></div><div><h3>Sample</h3><div>The study analyzed multimodal data—such as video, audio, and transcripts—from the Global Teaching Insights study, which included video recordings and transcripts from 46 teachers and 1,132 students in Germany.</div></div><div><h3>Method</h3><div>Scores for 18 teaching effectiveness subdimensions from three core dimensions were automatically generated by training attention-based ML models on multimodal features extracted from pretrained encoders. These ML-generated scores were compared with scores provided by human experts. A content validity study was conducted, where human experts evaluated the plausibility of ML-generated scores against human-generated scores. Structural equation models were used to assess the relationship between teaching effectiveness subdimensions and students’ tested achievement.</div></div><div><h3>Results</h3><div>ML-generated scores were more reliable for some subdimensions (e.g., nature of discourse), and they were also plausible and content valid. ML-generated scores achieved higher absolute accuracy than human scores in 11 of 18 subdimensions. Limitations include reliance on human ratings as ground truth and inconsistent predictive validity, underscoring the need for refined models to generate actionable insights, such as real-time feedback systems.</div></div><div><h3>Conclusions</h3><div>The findings provide valuable insights for the development of automated feedback, enhancing the practical application of teaching effectiveness assessments.</div></div>","PeriodicalId":48357,"journal":{"name":"Learning and Instruction","volume":"101 ","pages":"Article 102264"},"PeriodicalIF":4.9,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145525475","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"教育学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0