Nonanchor equating presents a significant challenge in educational assessment when test forms lack common items, requiring innovative solutions to ensure score comparability across different test administrations. This study proposes a novel large language model-simulated nonequivalent groups with anchor test (LLM-SNGAT) method that leverages large language models (LLMs) to simulate test-taking samples and generate common item sets for equating purposes. The approach eliminates traditional dependencies on specialized test design and extensive demographic data collection by utilizing the inherent capabilities of LLMs to simulate diverse response patterns. We evaluated the method using Tucker and Levine equating approaches across multiple LLMs, including generative pre-trained transformer 4o (GPT-4o), O1-preview, and DeepSeek-R1. Results demonstrated the feasibility of the proposed approach, with the Tucker method showing superior performance and consistent improvements as common item coverage increased. Sensitivity analysis confirmed that model performance rankings remained consistent across varying prompt formulations. The study revealed characteristic that standard errors were smallest near the mean and became larger farther away from the mean, and identified optimal common item proportions of 30%-50% for stable equating performance. While current limitations include the capacity of LLMs to accurately simulate human cognitive and behavioral diversity, this proof-of-concept study provides preliminary evidence for the feasibility of the LLM-SNGAT methodology. The approach represents a paradigm shift from resource-intensive traditional methods to computationally driven solutions, offering promising prospects for addressing nonanchor equating challenges in the digital age.

Automated item generation (AIG) is a promising technology in educational assessments. The advent of large language models (LLMs) has further propelled AIG’s development, expanding the form and content of generated items. However, existing research has predominantly concentrated on question generation, neglecting critical processes such as pilot testing, item analysis, and revision, which are integral to the development of real-world assessments. This study designed and validated a multi-agent AIG framework that simulates teamwork dynamics. By designing profiles, plans, and actions for item developer agents, field operator agents, and data analyst agents, a multi-agent LLM-based AIG system was constructed. Using the RACE reading comprehension dataset, item developer agents generated items based on the reasoning subdivisions of the original items. 32 student agents with varying abilities subsequently answered these items. After collecting responses, data analyst agents employed classical test theory (CTT) and item response theory (IRT) to process the answers, culminating in a pilot test report presented to item developer agents for item revisions. The quality of generated items was evaluated using machine metrics, human assessment methods, and LLM-based pairwise comparisons. The results demonstrated that the proposed multi-agent approach significantly enhances the generation capabilities of the foundational model compared to single-agent baseline systems. This study highlighted the need for developing credible autonomous workflows for AIG and provided insights for future implementations of LLM-based multi-agent systems in educational assessments.

The emergence of advanced reasoning models like O1 and DeepSeek-R1 marks a significant advancement in LLM capabilities, enabling applications in complex tasks such as test equating in educational measurement. This study explored test equating without anchor items by leveraging four large language models-GPT-4o, O1-mini, O1-preview, and DeepSeek-R1-to simulate participant data for linking two distinct mathematics tests (X and Y). The two-parameter item response theory (IRT) model was used for equating. Results showed that advanced reasoning models outperformed general-purpose LLMs, with DeepSeek-R1 demonstrating smaller errors than O1-preview, and both outperforming GPT-4o and O1-mini. Increasing the size of linking groups reduced errors, and prompt characteristics such as gender roles influenced error magnitude. This study highlights the potential of advanced reasoning LLMs in enhancing test equating processes.

Critical thinking, as a crucial component of 21st-century core competencies, poses significant challenges for effective assessment in educational evaluation. This study proposes an automated assessment method for critical thinking in middle school Chinese language based on a Bidirectional Encoder Representations from Transformers—Convolutional Neural Network (BERT-CNN) hybrid model, achieving a multi-dimensional quantitative assessment of students’ critical thinking performance in writing through the synergistic effect of deep semantic encoding and local feature extraction. The research constructs an annotated dataset containing 4827 argumentative essays from three middle school grades, employing expert scoring across nine dimensions of the Paul–Elder framework, and designs three types of logic chain extraction algorithms: argument–evidence mapping, causal reasoning chains, and rebuttal–support structures. Experimental results demonstrate that the BERT-CNN hybrid model achieves a Pearson correlation coefficient of 0.872 in overall assessment tasks and an average F1 score of 0.770 in logic chain recognition tasks, outperforming the traditional baseline methods tested in our experiments. Ablation experiments confirm the hierarchical contributions of semantic features (31.2%), syntactic features (24.1%), and logical markers (18.9%), while revealing the model’s limitations in assessing higher-order cognitive dimensions. The findings provide a feasible technical solution for the intelligent assessment of critical thinking, offering significant theoretical value and practical implications for advancing educational evaluation reform and personalized instruction.

The interactive mechanism between social and concept networks is a key question in connectivist learning, which explains the impact of interaction on cognitive development and knowledge generation. The successful practice of "Internet Change Education: Dialogue between Theory and Practice", the first cMOOC in China, provides data support and research context for exploring the learning rules. This study proposed an evaluation framework of individual social network status in cMOOCs and a model for concept network characteristics analysis in cMOOCs. Analysis from 10,598 interactive data points of cMOOC1.0 using complex network analysis, topic clustering algorithm and correlation analysis found that there was a correlation between a learner's individual network status and concept network characteristics, and there are two kinds of special connectivist learners in cMOOC learning: someone with higher network status but lower knowledge output, and someone with higher knowledge output but weaker network status. Furthermore, we put forward three practical suggestions: (1) facilitators should focus on guiding learners' expression and connection to important individuals; (2) according to different correlation trends, facilitators can flexibly tailor activity design strategies, such as providing additional wayfinding support for learners with low centrality; and (3) carry out adaptive learning evaluations and design diversified certificates.

Scientific inquiry ability is closely related to the process of hands-on inquiry practice. However, its assessment is often separated from this practice due to the limitation of technical basis and labor cost. The development of multimodal data analysis provides a new opportunity to realize automated assessment based on hands-on practice. Therefore, this study aims to explore whether and how we can use automated multimodal data analysis approaches to measure the scientific inquiry ability of students during the hands-on inquiry practice. In a scientific inquiry activity called "Explore the Moon," designed for 472 fourth graders, we collected textual, tabular, and video data. Aiming to analyze and evaluate the data, we first designed a modal conversion method based on the multimodal pre-trained model LLaVA-7B and a text scoring method integrating keyword matching, one-way nearness, and Jaccard similarity. Then, to bridge the computing ability with the scoring criteria from science teachers, we constructed a structured representation language and verified the human–machine consistency of automated scoring. Finally, we used a multidimensional item response theory (IRT) model to validate the assessment's overall quality and analyze the participants' scientific inquiry ability. The proposed data analysis method has high man–machine consistency, and the results of IRT analysis present reasonable item characteristics. In summary, we constructed a low-cost and scalable multimodal assessment approach based on scientific inquiry activities, providing methodological support for science teachers to carry out formative evaluation of students' scientific inquiry activities in the daily inquiry-based learning environment.

Manually scoring and revising student essays has long been a time-consuming task for educators. With the rise of natural language processing techniques, automated essay scoring (AES) and automated essay revising (AER) have emerged to alleviate this burden. However, current AES and AER models require large amounts of training data and lack generalizability, which makes them hard to implement in daily teaching activities. Moreover, online sites offering AES and AER services charge high fees and have security issues uploading student content. In light of these challenges, and recognizing the advancements in large language models (LLMs), we aim to fill these research gaps by analyzing the performance of open-source LLMs when accomplishing AES and AER tasks. Using a human-scored essay dataset (n = 600) collected in an online assessment, we implemented zero-shot, few-shot, and p-tuning AES methods based on the LLMs and conducted a human-machine consistency check. We conducted a similarity test and a score difference test for the results of AER with LLMs support. The human-machine consistency check result shows that the performance of open-source LLMs with a 10B parameter size in the AES task is close to that of some deep learning baseline models, and it can be improved by integrating the comment with the score into the shot or training continuous prompts. The similarity test and score difference test results show that open-source LLMs can effectively accomplish the AER task, improving the quality of the essays while ensuring that the revision results are similar to the original essays. This study reveals a practical path to cost-effectively, time-efficiently, and content-safely assisting teachers with student essay scoring and revising using open-source LLMs.

BackgroundThere is a strong association between interactions and cognitive engagement, which is crucial for constructing new cognition and knowledge. Although interactions and cognitive engagement have attracted extensive attention in online learning environments, few studies have revealed the evolution of cognitive engagement with interaction levels. ObjectivesThe study aims to automatically identify learners' interactions and cognitive engagement and then analyse the evolution of learners' cognitive engagement with interaction levels and during different stages of online learning. MethodsThe participants of the study were learners who participated in an online open course. Their text data from discussion forums on five learning themes were collected. Data were analysed using text mining and ENA. ResultsLearners' cognitive engagement in online learning was related to interaction levels. As learners' online interaction levels changed from surface to deep, cognitive engagement levels changed from low to high. With the continuous occurrence of deep interactions, cognitive feedback became more complex. At the social-emotional interaction level, although learners' cognitive engagement levels began to change from low to high, complex cognitive feedback was still insufficient. In addition, the analysis of the evolution of cognitive engagement during different stages of online learning showed that learners' patterns of cognitive engagement changed significantly as the learning process continued, from initially dynamic and complex to a stable development pattern. ImplicationsThe results of the study are of theoretical significance and practical guidance for further understanding the relationship between online interaction levels and cognitive engagement as well as the process of online collaborative knowledge exploration, construction, and even connectivity.

如今,教育大数据的应用进入了瓶颈期,机械、单一、截断式的数据已无法满足当下教育发展的需要.而与一般性的教育大数据相比,教育领域中学生队列资源具有连续、丰富、可解释等优势,有望成为破解发展瓶颈的重要举措.基于此,文章首先梳理了学生队列资源的概念与价值潜能,然后基于医疗领域队列建设思路,总结包含教育目标选取、暴露因素定义、研究对象设置、随访计划制定和数据分析存储的队列资源建设整体框架,最后指出队列资源建设存在的隐私保护、追踪成本和样本覆盖面等诸多挑战,并提出提高安全意识、发挥技术优势和开展更大规模队列建设等策略.文章通过研究,旨在为推动教育领域学生队列资源建设与发展提供参考,弥补现有教育数据缺陷,发挥数据在教育领域中的作用.

成人终身学习是我国教育现代化的重要增长极,也是教育改革的重大实践命题.知识观关乎教育实践的根本性和基础性认识问题,而传统知识观在知识价值、知识形态、知识生产与传播、知识组织上的局限,已成为制约成人终身学习创新发展的主要原因.从人类知识观发展脉络来看,其经历了强调知识的客观属性、社会属性以及网络属性的三个阶段.强调知识网络属性的新知识观,其核心要义和创新价值集中体现在:知识从精加工的符号化信息回归为全部的人类智慧,其体现出的知识全谱系特征为破解成人终身学习在标准化资源供给与个性化学习需要上的矛盾提供了新思路;互联网环境中涌现出网络化知识新形态,以及伴随而生的联通主义学习为成人终身学习提供了新途径;知识生产与传播过程相融合大幅加速了知识的更新与流动,其孕育出的在线社区型课程能有效满足成人动态变化的学习需求;互联网环境中知识组织方式从以学科为线索转向以问题为线索,为破解高度综合社会实践与成人终身学习分科培养间的矛盾提供了抓手.构建服务全民终身学习的教育体系首先要树立新的知识观,并以此为成人终身学习领域的创新发展提供新动能.

交互是联通主义学习的核心和取得成功的关键.现有交互水平评估主要依靠人工编码实现,存在时间滞后性和耗时耗力等局限,难以实时反映交互水平状况.因此,本研究以国内首门联通型慕课"互联网+教育:理论与实践的对话Ⅱ"为案例,运用人工智能领域中的深度学习算法,通过模型构建、模型检验和模型应用三个阶段实现了对联通慕课参与者交互水平的自动化评估.研究发现,在联通主义学习中:1)学习者成为交互主体,多元主体驱动课程交互;2)寻径和意会是决定联通主义学习成效的关键.本研究一方面检验和发展了原有联通主义交互规律研究,另一方面实现了联通主义学习情境中参与者交互水平的自动化评估,为联通主义学习评价和课程评价提供了工具支撑,探索了人工智能技术在改进教育研究和实践方面的价值和可能性.

教育哲学是关于教育最根本问题的认识,是教育实践的"道",决定了教育改革发展的方向.进入新发展阶段,终身学习在促进社会可持续发展上的价值日益凸显,其学习空间逐渐向互联网迁移,学习者的学习需求更加多元化,终身学习的新趋势呼唤教育哲学创新.从教育本体论层面来看,终身学习情境下的教育本质演变为联通,强调通过联通促进个体与世界间多个层次的互动联系.从终身学习的认识论来看,知识演变为网络化形态,呈现全谱系、动态网络化、群智汇聚和问题导向等新特征,推动终身学习组织体系、教学观和学习观的变革.从方法论层面来看,终身学习情境下的教育系统日益复杂、多元和开放,教育研究应当树立教育生态观,注重采取数据密集型研究范式和设计性研究范式.终身学习领域必须高度重视"道"的突破,加强哲学基础研究,为学习型社会建设提供更加强有力的思想武器.

学生综合素质评价是完善人才培养和选拔机制的重要方面,已成为推动新时代基础教育改革的关键.在教育信息化的背景下,学生综合素质评价呈现出在内容层面将信息化素养作为评价要点、在形式层面通过信息化平台支持评价创新、在方法层面注重运用智能技术促进流程变革的发展样态.然而,技术赋能的学生综合素质评价也面临着评价理念异化偏离、技术开发能力薄弱、基础资源建设失衡、数据开放难以实现、运行机制不够完善等现实挑战.在教育信息化向数字化转型的战略行动中,为充分发挥信息技术赋能学生综合素质评价优势,应当从以下方面着力:一是秉持科学的教育评价理念,形成体现人文价值的评价文化;二是持续完善智能技术应用,推进学生综合素质评价全流程变革;三是夯实基础性支撑,提升薄弱地区信息化评价实施能力;四是构建数据开放共享协同机制,完善个体隐私保护措施;五是出台保障性政策与技术标准,推动评价活动科学开展.

信息技术为新时代表现性评价变革提供了重要机遇,促进信息技术与表现性融合创新是深化教育评价改革和推进素质教育发展的必然要求.该文着眼素质教育对高阶思维能力评价的现实需求,在分析当前表现性评价实践问题的基础上,提出了基于信息技术的表现性评价,其本质是在教育信息化背景下,以证据中心设计理论为指引,充分利用信息技术变革表现性评价全流程.从创新测评任务呈现、拓展测评数据采集、优化测评数据分析和促进测评活动实施等维度剖析了基于信息技术的表现性评价的作用点.未来需要从下列方面来推进基于信息技术的表现性评价:树牢改革理念,正确认识表现性评价的内容、形式与方法;优化技术服务,持续改进表现性评价的可用性、易用性与实用性;完善数据标准,稳步提高表现性评价的规范性、统一性和效率性;建强人才队伍,大力推动表现性评价的专业性、科学性和可靠性.

知识观是对知识本质的认识,是教育的认识论,也是教育工作的出发点和根本性问题,决定了课程教学内容乃至教育组织体系和服务模式.知识是人类对认知客体的认知和表征,受到媒介技术的影响,技术变革也推动着知识观的进步.以媒介技术发展为线索,知识的"技术"发展史可划分为口传时期、手写文字时期、印刷文字时期、电子传播时期、数字传播时期五个阶段,其在知识的表征形态、存储方式、传播载体、主要类型、传播方式上均有差异.在数字传播时期,基于互联网的信息空间使得知识内涵扩展、知识数量激增,随之出现的网络化知识呈现出分布式存储、动态变化、综合性碎片化、知识生产即传播等特点.知识的本质属性从静态的线性知识演变为动态的网络化知识.网络化知识呈现出复杂网络演化的规律,对其表征与抽取是揭示网络知识演化规律的关键.网络化知识可以汇聚人类全部智慧,通过"涌现"实现对认知客体的整体识别,能够及时汇聚并传播最新前沿成果,在完整性、复杂性、动态性三方面呈现"回归"知识本源的趋势.网络化知识的出现将引发教育内容和组织方式变革,推动学校与社会日益融合,并呼唤人才培养目标的转型.

近期,我国高度重视和大力推进"互联网+教育"工作,不仅明确了利用互联网等新一代信息技术推动教育变革的核心思路和五大着力方向,并且强调了学术创新的重要性.前期教育信息化经验也启示我们,理论与实践两张皮现象严重阻碍了教育改革创新的进程.作者在研读国家政策文件、反思实践问题的基础上,进一步细化了基本理论、技术环境、治理政策三大科研导向下的12个学术创新方向:基本理论层包括哲学基础、知识进化、学习规律、教学模式四个方向;技术环境层包括平台服务、数据模型、资源优化、终端影响四个方向;治理政策层包括治理模式、供给机制、业态发展、改革成效四个方向.希望文章能够帮助教育领域研究者明晰"互联网+教育"研究方向与着力点,发挥学术创新对保障"互联网+教育"工作稳步推进、突破改革困境、引领实践创新的助益效果.

"互联网+教育"是一个不断发展的创新领域,是教育高质量发展的必然选择.本体涵盖了某个领域语义网络的关键概念,可以客观反映特定领域的发展重点.研究以构建"互联网+教育"领域本体为主线,以剖析"互联网+教育"领域结构为重点,实现了"互联网+教育"领域本体可视化与结构系统化.研究发现,"互联网+教育"领域具有"两级宏观—两级微观"的体系结构,宏观层面分为基础支撑、发展任务、动力引擎、保障体系四个维度,微观层面包括26个一级结构和53个二级结构.并进一步从夯实基础支撑、落实发展任务、强化动力引擎、完善保障体系等方面提出了推进"互联网+教育"发展的具体建议,以期为促进"互联网+教育"领域健康发展奠定理论基础.