Content area
In-service teacher professional development (TPD) is essential for improving teacher quality and student outcomes. Effective professional development equips teachers to actively engage in problem-solving and meaning construction. However, current online TPD often lacks tailored support, structured analysis, communication, and feedback, limiting teachers’ ability to engage in deep knowledge-building. Generative Artificial Intelligence (GenAI), exemplified by models like ChatGPT, has attracted significant attention for its potential in education, particularly in offering personalized feedback and fostering deep cognitive engagement. This study examines a large language model developed in China to investigate its impact on in-service teachers’ knowledge-building processes. Through analysis of frequency and epistemic network, this study demonstrates that GenAI significantly enhances in-service teachers’ information analysis and critical thinking. It also promotes greater attention to information processing, evaluation, and knowledge transfer during the knowledge-building process, although it performs less effectively in fostering social interaction and collaboration. The study further reveals that GenAI’s impact on knowledge building varies across learning tasks, with its support being particularly significant in higher-order, complex tasks. Building on these findings, the study offers recommendations for professional development for teachers.
Introduction
In-service teacher learning, often referred to as teacher professional development (TPD), enables teachers to acquire new skills and improve their professional practice (Desimone et al., 2002; Kalinowski et al., 2020). It is recognized as a key factor in enhancing teacher quality and improving school and student outcomes (Opfer & Pedder, 2011). As such, the professional development of in-service teachers has become a major focus in educational research. Compared to traditional face-to-face instruction, online learning offers greater flexibility as it is not bound by time or location, allowing learners to engage at their own pace (Boling & Martin, 2005). This flexibility makes online learning a preferred choice for in-service teachers (Parsons et al., 2019; Powell & Bodur, 2019). Although online learning provides in-service teachers with the opportunity to acquire knowledge anytime and anywhere, their learning is often self-directed and characterized by passive information reception. Furthermore, it is challenging for online learning platforms or service providers to provide personalized feedback to each learner, leaving them to face difficulties such as the complexity of learning content, inaccurate knowledge sharing (Shin et al., 2018), and uncertainty about how to improve or refine their understanding (Chen et al., 2021). These limitations hinder in-service teachers’ ability to engage in meaningful interactions and knowledge sharing, making it even harder for them to develop skills in information analysis, integration, and application. In conclusion, the above-mentioned problem limited the development of learners’ specific ability, which is known as knowledge building (Yücel & Usluel, 2016). Knowledge building is crucial for teachers’ professional development, as it involves exchanging and constructing knowledge through interaction (Cress & Kimmerle, 2013). This process encompasses aspects like information integration, comprehension, and application, contributing to both individual learning and productive collaboration (Yücel & Usluel, 2016). In-service teachers, balancing roles as both learners and educators, face the challenge of transferring abstract professional knowledge into practical pedagogical knowledge (Kennedy, 2005). Current teacher professional development lacks effective strategies or guidance to bridge this gap, resulting in low levels of knowledge building and poor outcomes. Teachers’ instructional practices, however, directly influence the development of students’ knowledge and competencies (Uiterwijk-Luijk et al., 2019). Therefore, effective training in these areas is essential. Such training may transform classrooms into communities of inquiry, fostering knowledge-building environments that support students in developing higher-order competencies (Bereiter & Scardamalia, 2014).
In recent years, the rapid development of Generative Artificial Intelligence (GenAI) has brought new technical support and practical possibilities for facilitating deep knowledge building among in-service learners. GenAI is capable of personalized content recommendation and knowledge generation based on user needs, offering meaningful feedback to support learners in independent or collaborative learning (Cooper, 2023). For instance, GenAI can provide timely feedback on student work, offer personalized learning experiences, generate lesson plans, and facilitate language acquisition (Dai et al., 2023; Kohnke et al., 2023). These features create interactive and adaptive learning environments. In addition, GenAI tools can assist learners in problem-solving by generating discussion prompts and suggestions that encourage critical thinking (Ray, 2023). They provide personalized feedback and step-by-step guidance in the teaching and learning environment, enabling learners to build knowledge systematically and engage in deeper cognitive processes. GenAI-assisted, represented by ChatGPT, provides new ideas for facilitating high-level knowledge building.
The application of GenAI in education holds promise, particularly the variant represented by ChatGPT, which demonstrates significant potential in fostering high-level knowledge building. However, its role in facilitating teachers’ professional development, as well as its mechanisms and strategies for promoting effective knowledge building, remain underexplored. Based on this, this study employs a large language model in China as a case study to conduct an empirical investigation with 143 in-service learners enrolled in a Learning Science course at University B. It explores the potential of GenAI to promote high-level knowledge building among learners and examines how GenAI influences learners’ knowledge building across various stages of learning. The ultimate goal is to provide a scientific basis and practical suggestions for the effective application of GenAI technology in classroom teaching, thereby facilitating its integration into the field of education.
The structure of this paper is organized as follows: The second section presents a literature review, systematically synthesizing existing research findings in related fields and establishing the foundation for research question formulation. The third section outlines the research design, detailing the selection of the research subject, the procedural design of research activities, and the methodologies for data collection and analysis. The fourth section analyzes and discusses the study results, addressing the research questions and elucidating the significance of the findings. The fifth section summarizes the key conclusions of the study, offers practical recommendations, and discusses the study’s limitations along with potential avenues for future research.
Literature review
Strategies for facilitating in-service teachers’ online deep knowledge building
Teacher professional development is achieved through a range of activities designed to meet teachers’ diverse needs and preferences, enhancing their knowledge, skills, and instructional practices (Huang et al., 2024). This approach is considered an effective way to promote teacher development (Avalos, 2011; King, 2014; Sancar et al., 2021). In terms of teacher development and professional training, online learning has become the preferred mode of learning for in-service teachers as it is more flexible compared to traditional face-to-face instruction. Studies have generally found no difference in teacher perceptions and learning effectiveness between in-person teacher professional development and online teacher professional development (Fishman et al., 2013; Tømte & Gjerustad, 2020; Yoon et al., 2020). Online teacher professional development allows learners to self-pace their learning without the constraints of time and location (Compen et al., 2019; Dede et al., 2009; Reeves & Pedulla, 2011), thereby supporting the improvement of teachers’ knowledge, skills, and instructional practices (Parsons et al., 2019; Powell & Bodur, 2019). However, despite these advantages, online learning also presents certain issues and challenges. Traditional teaching methods may hinder personalized support and immediate feedback (Xu et al., 2024). The self-directed nature of online learning and passive information receipt make it difficult for teachers to provide tailored feedback to each learner. Learners may face problems such as inaccurate knowledge sharing due to the complexity of the learning content (Shin et al., 2018), uncertainty about how to improve and refine their knowledge (Chen et al., 2021), and difficulty in maintaining continuous motivation as the course progresses, which limits their professional development (Belay & Melesse, 2024; Zhang et al., 2021). Furthermore, online learners often engage in superficial shared practices with minimal in-depth discussion of teaching or pedagogical issues (Brown & Munger, 2010; Kelly & Antonio, 2016; Tsiotakis & Jimoyiannis, 2016). In-service teachers are more likely to focus on updating their knowledge and skills rather than engaging in reflective activities (Kwakman, 2003), which hinders deep knowledge building.
The concept of knowledge building was first introduced by Canadian education scholars Marleen Scardamalia and Carl Bereiter in the 1980 s (Bereiter & Scardamalia, 1989). Unlike traditional classroom learning, which focuses on knowledge acquisition, learning in a knowledge building community relies on social processes, emphasizing that knowledge can be deepened, understood, and innovated through collaboration and interaction among community members (Zhang et.al, 2011). They argue that knowledge building is more than just the transfer or reconstruction of knowledge; it is also about students becoming part of a culture of knowledge creation and advancing collective knowledge through collaborative inquiry (Bereiter & Scardamalia, 2014). This process emphasizes continuous improvement and deeper understanding. Effective teacher professional development should engage learners in a learning community to share different perspectives, encounter new ideas, and expand their understanding of teaching and learning through collaborative discussion of problems encountered in practice (Vescio et al., 2008). Knowledge building positively impacts learners’ depth of inquiry, collaboration, and knowledge-creation processes (Lee et al., 2006; van Aalst & Chan, 2007). The ability to construct new knowledge by interacting with peers is key to determining the quality of online collaborative learning (Lämsä et al., 2021). Nevertheless, learners still face many challenges in the process of knowledge building, such as inaccurate knowledge sharing due to the complexity of the learning content (Shin et al., 2018), and a lack of knowledge on how to improve and refine their knowledge (Chen et al., 2021), which often makes it difficult for them to achieve the desired learning outcomes. To address these problems, researchers have explored a variety of strategies to facilitate knowledge building. At the level of supporting tools, group awareness tools (Li et al., 2021), pedagogical scaffolding (Avcı, 2020), knowledge maps (Zheng et al., 2023), and collaborative scripts (Weinberger, et al., 2012) are widely used in instructional practice and can facilitate efficient collaborative knowledge building to a great extent. Learning analytics tools demonstrate the ability to help understand and optimize the learning process (Siemens & Baker, 2012), enhance feedback in collaborative learning (Banihashem et al., 2022), and effectively facilitate learners’ knowledge building (Feng et al., 2019). Furthermore, socially shared regulation can support collaborative knowledge building by helping learners co-regulate collaborative learning activities through setting goals, planning, monitoring, and adjusting the learning process (Zabolotna et al., 2023), as well as reflecting on and evaluating the learning process (Hadwin, et al., 2011). However, learners often struggle with these tasks, especially in the areas of reflection, assessment, and self-monitoring.
In recent years, AI in education has demonstrated significant potential, particularly in enhancing personalized learning and fostering high-level knowledge building among students, enabling self-reflection, collaboration, and personalized learning for in-service teachers (Blanchard et al., 2016; Sadler et al., 2020). Gen AI provides personalized feedback and step-by-step guidance in teaching and learning environments to aid in learners’ knowledge building and stimulate deep cognitive engagement. Studies leveraging ChatGPT have enhanced learners’ online collaborative learning experiences, providing efficient and flexible learning support (Zhu et al., 2023). GenAI, exemplified by ChatGPT, offers innovative approaches to facilitating advanced knowledge building among learners.
Application of GenAI in teaching and learning
Generative Artificial Intelligence (GenAI) refers to a subset of AI that focuses on computer models designed to generate content. These models are capable of generating multimodal outputs including text, images, 3D models, video, audio, and software code, in response to user input prompts (Chiu, 2024a, 2024b). These technologies leverage advanced natural language processing and deep learning techniques to generate human-like text based on input data, enabling them to engage in meaningful conversations and provide relevant, context-aware information (Goodfellow et al., 2020). GenAI technologies, particularly large language models like ChatGPT, have the potential to transform higher education by enhancing teaching, learning, and student engagement (Chiu, 2024a, 2024b).
GenAI’s pivotal role in education lies in its capacity to enrich the learning experience by generating highly original content in response to user prompts (Chan & Hu, 2023). Tailoring to individual needs and preferences revolutionizes how learners acquire and engage with information, thereby enhancing accessibility and engagement in education for all (Chang & Kidman, 2023; Chiu, 2024a). These GenAI benefits can bridge learning disparities, deepen comprehension, and support self-paced learning, particularly in online environments that facilitate flexible, ubiquitous education. Generative AI encompasses all facets of personalized learning for learners. For learners, Gen AI encompasses all facets of personalized learning. In terms of personalized resource recommendation and learning path planning, by analyzing students’ behavioral data and learning goals, Gen AI can generate interactive learning content for students (Kadaruddin, 2023), assist students in adjusting their learning plans according to their progress and needs (Bahroun et al., 2023), thereby enhancing the efficiency and depth of knowledge acquisition. In terms of intelligent tutoring, Gen AI-powered chatbots serve as integral components of intelligent tutoring systems, offering interactive learning services and personalized tutoring (Chang & Kidman, 2023; Ilieva et al., 2023), facilitating step-by-step guidance (Chiu, 2024a, 2024b). In terms of automated assessment and feedback, generative AI can automate the evaluation of student performance in assignments and exams, while also providing timely feedback (Wilson & Nishomoto, 2023). These assessments not only enhance the efficiency of the evaluation process but also help teachers reduce their workload. For teachers, GenAI empowers them generate course materials, including lesson plans and quizzes, and to tailor personalized courses and content to meet learners’ specific needs, thereby enhancing the overall quality of education (Chen et al., 2023). GenAI conserves educators’ time and effort by synthesizing and rewriting existing content, enabling them to concentrate on the more complex facets of curriculum development and instruction.
Few studies indicate that Gen AI, represented by ChatGPT, can generate new ideas and insights to assist learners during challenging learning activities (Baidoo-Anu & Owusu Ansah, 2023; Wu et al., 2024), provide strategies for learning improvement (Kasneci et al., 2023; Wu et al., 2024), and hold potential to facilitate self-regulated learning among learners (Chiu, 2024b; Molenaar et al., 2023; Wu et al., 2024; Xia et al., 2023). Many studies have confirmed the potential of GenAI to enhance teaching practices, particularly in fostering higher-order cognitive development. In the example of ChatGPT, it has been noted that it can act as a powerful cognitive agent to help learners understand complex concepts (Yilmaz & Yilmaz, 2023), develop critical thinking by providing various strategies for problem solving (Santos, 2023), and promote advanced collaboration and knowledge building among learners (Azaria et al., 2024). In addition, several studies have demonstrated that interactions with AI can lead to increased cognitive engagement (Asiri et al., 2021; Liang et al., 2023). Nazari et al. (2021) found in a randomized controlled trial that students, by continuously receiving AI-generated feedback, could reflect on their writing process, actively revise and enhance their work, and thus achieve a higher level of cognitive engagement in the knowledge building process. Engagement with AI systems has been shown to enhance students’ questioning abilities and creative problem-solving strategies (Wood et al., 2024).
Simultaneously, GenAI encounters challenges in teaching practice, particularly in integrating multidisciplinary knowledge and diverse media resources (e.g., text, images, videos), which often hinder its ability to generate constructive and practical teaching recommendations. An et al. (2024) emphasize that GenAI tools require effective integration with well-designed collaborative scaffolds to achieve optimal outcomes. Furthermore, concerns have been raised among educators regarding the negative impacts of GenAI, particularly in terms of academic integrity and over-reliance on technology (Tlili et al., 2023; Wise et al., 2024). Despite these risks and challenges, the significant potential of GenAI in education remains undeniable. In the current educational landscape, GenAI is increasingly being integrated into education, driving comprehensive transformations across the field. The impact and application of GenAI products, such as ChatGPT, Bing AI, and Bard, within education have garnered significant scholarly discourse and attention. Nevertheless, the empirical studies on the application of GenAI within actual classroom settings remain limited, and the pedagogical effects of GenAI, as well as its influence on the higher-order cognitive development of learners in knowledge construction, remain to be fully elucidated.
In recent years, increasing attention has been devoted to the analysis of teachers’ learning processes, with a particular emphasis on their collaborative and cognitive development (Walkoe& Luna, 2019). For instance, Zhang et al. (2017) explored the interaction network and social knowledge building behavioral patterns in online collaborative learning activities of elementary school teachers by combining social network analysis, content analysis and lag sequence analysis. Ouyang et al. (2021) examined the interaction processes of teachers in collaborative problem solving using multiple methods, including content analysis, lag sequence analysis, and frequent sequence mining. These studies have highlighted that cognitive development is a dynamic process necessitating a fine-grained approach for exploration and analysis (An et al., 2024). Therefore, exploring the fine-grained processes of in-service teachers’ cognitive development is crucial, as it plays a pivotal role in fostering their higher-order skills and enabling them to achieve advanced levels of knowledge building. Epistemic network analysis (ENA) has emerged as a powerful methodological tool for modeling and visualizing the complex structure of cognitive and collaborative learning processes. Unlike other tools, ENA not only facilitates the comparison of network structures across conditions but also provides interpretable visual representations of statistical differences. Swiecki et al. (2020) demonstrated that ENA could reveal individual performance differences in collaborative discourse while enabling statistical validation. Additionally, ENA can model the unique contributions of individuals to collaborative discourse while considering the group context, allowing for the simultaneous analysis of both individuals and groups within the same model (Tan et al., 2024). As a result, ENA has been widely applied in diverse educational contexts, including complex thinking and knowledge construction (Csanadi et al., 2018; Oshima et al., 2018), collaborative problem-solving (Bressler et al., 2019; Swiecki et al., 2020), socioemotional analysis (Prieto et al., 2021), and teacher professional development (Bauer et al., 2020; Fernandez-Nieto et al., 2021; Phillips et al., 2023; Zhang et al., 2022). While these studies offer valuable insights, the integration of GenAI into teacher learning contexts remains underexplored from a theoretical and analytical perspective. GenAI, particularly large language models (LLMs), introduces novel opportunities for personalized scaffolding, real-time feedback, and co-construction of knowledge in collaborative learning settings. From a sociocultural and constructivist perspective, the role of GenAI can be conceptualized not merely as a tool, but as a cognitive partner that mediates and transforms epistemic practices. However, the implications of this paradigm shift for in-service teacher learning require systematic investigation.
Therefore, this study leverages ENA to analyze the epistemic processes of in-service teachers in a GenAI-assisted collaborative learning environment. Specifically, we adopt a leading Chinese large language model as a pedagogical agent, integrating it into a GenAI-assisted instructional model designed to provide adaptive scaffolding. The study divides the whole teaching process into three stages, each aligned with varying levels of learning tasks, and collects the discussion data of in-service learners at each stage to thoroughly analyze and assess the role of GenAI in the online knowledge building process of these learners. The following three questions are mainly explored:
Q1: Can GenAI-based strategy facilitate high-level knowledge building for in-service learners compared to traditional learning?
Q2: What are the differences in learners’ knowledge building process under GenAI-based strategy compared to traditional learning?
Q3: What are the characteristics of learners’ knowledge building processes at different learning stages under GenAI-based strategy compared to traditional learning?
Research methodology
Participants
This study involved 143 in-service postgraduate students majoring in English education enrolled in a “Learning Sciences” course at University B in China. The participants were students from two parallel classes. To investigate the impact of GenAI on teacher professional learning, one class was assigned as the experimental group, while the other served as the control group. The experimental group (n = 74; 7 males, 67 females) engaged with GenAI-assisted learning throughout the course, while the control group (n = 69; 11 males, 58 females) followed a traditional online learning model. The two groups were generally comparable in terms of professional background. In the experimental group, 17 participants were employed in non-educational sectors, while the rest were educators, with an average of 2.6 years of teaching experience. In the control group, 10 participants came from non-educational fields, and the rest were educators, with an average teaching experience of 2.7 years.
To control for potential confounding variables and enhance the internal validity of the study, a stratified random grouping strategy was employed during the formation of learning teams. Specifically, participants were first stratified based on learning styles and leadership tendencies and then randomly assigned to groups of four to six members. Each group was composed to ensure heterogeneity in learning styles and leadership profiles, thereby achieving relative homogeneity across groups. This design aimed to minimize interference from individual differences and strengthen the methodological rigor of the experimental framework. Throughout the intervention, both groups were taught by the same experienced instructor using identical instructional content and materials, ensuring consistency in teaching practices apart from the variable under investigation. In addition, to account for prior knowledge differences, all participants completed a standardized pre-course knowledge assessment developed by the institution, covering foundational concepts relevant to the course. The results indicated that all learners met the passing threshold, suggesting a relatively uniform baseline of prior knowledge. At the beginning of the course, a validated learning motivation questionnaire (Hwang et al., 2013; see Appendix A) was administered to both groups to examine potential differences in motivational levels. The results of the questionnaire are shown in Table 1. Statistical analysis showed no significant difference between the two groups (control group mean = 4.59, experimental group mean = 4.46, p = 0.109 > 0.05). In addition, all participants provided informed consent prior to participation, confirming their voluntary involvement and agreement to the anonymous use of their data for scientific research purposes.
Table 1. Survey results on motivation of two groups
Dimension | Group | Mean | Sd | t | p |
|---|---|---|---|---|---|
Learning motivation | Control group | 4.59 | 0.42 | 1.61 | 0.109 |
Experimental group | 4.46 | 0.41 |
Instructional model
Grounded in constructivist learning theory, this study introduces a GenAI-assisted instructional model designed to foster deep learning and competency development among in-service teachers through technological integration (see Fig. 1). The model comprises three progressive learning stages: competency acquisition, meaning comprehension, and application and transfer. The learning tasks within each stage are progressively challenging, and GenAI offers tailored scaffolds and strategies to support students in achieving cognitive development at varying levels, aligned with their individual needs. In this study, the experimental group was instructed using the GenAI-assisted instructional model, whereas the control group adhered to a conventional online teaching approach.
[See PDF for image]
Fig. 1
GenAI-Assisted instructional model
During the Competency Acquisition stage, GenAI primarily delivers a knowledge framework to assist learners in constructing a robust knowledge structure. Through personalized learning support, GenAI dynamically generates learning content tailored to the learner’s ability level, informed by their background knowledge and learning progress. The objective of this phase is to enable learners to master foundational knowledge and core skills, ensuring their ability to develop deeper understanding in subsequent learning. GenAI facilitates the internalization of knowledge and the formation of a systematic knowledge network by delivering a clear knowledge framework and structured content.
During the Meaning Comprehension stage, GenAI guides learners in comprehending the deeper significance of knowledge through contextualized interactions. At this stage, GenAI emphasizes the comprehension and application of knowledge. It assists learners in integrating abstract concepts with concrete scenarios by simulating authentic contexts or offering innovative solutions, thereby fostering in-depth processing and internalization of knowledge.
During the Application and Transfer stage, GenAI supports learners in transferring acquired knowledge to novel contexts and solving complex problems through adaptive feedback mechanisms. This stage prioritizes knowledge transfer and innovation, with GenAI delivering timely, personalized feedback through design thinking and other methodologies. This enables learners to reflect on their learning processes and refine their strategies, ultimately achieving higher-order cognitive abilities.
Design of the learning process
This study is based on the online training course “Learning Science” conducted by University B in Beijing, based on the Learning Cell System (LCS, etc.edu.cn) for data collection and organization. The course, compulsory for in-service teachers, spans a 16 week semester, structured around theme-based learning modules and facilitated through online discussions. The whole instructional framework of the course can be divided into three stages (see Fig. 2).
[See PDF for image]
Fig. 2
The research design
Stage1: Competency Acquisition. This stage is dedicated to the foundational understanding of key concepts through introductory Q&A sessions, where learners engage in discussions on topics like “What is learning science?” or “What is academic ethics?” to explore conceptual frameworks and share personal insights. Figure 3 demonstrates the differential learning processes between the two groups using a representative learning task from Stage 1 as an example. During Stage 1, while the control group relied primarily on online learning resources and traditional community Q&A for problem-solving and knowledge acquisition, the experimental group adopted a distinctive approach: they addressed learning challenges through peer feedback while simultaneously leveraging GenAI’s powerful information retrieval capabilities to systematically construct their knowledge frameworks.
[See PDF for image]
Fig. 3
Learning case in Stage 1
Stage2: Meaning Comprehension. Learners are tasked with integrating and internalizing the concepts learned with their prior experiences. Discussions, such as “How do cognitive apprenticeship and metacognitive strategies inform your teaching practices?” encourage reflection and application of theoretical knowledge to practical scenarios. Learners are required to understand what they have learned from their existing experiences. Figure 4 presents the differences in learning processes between the two groups, using a representative learning task from Stage 2 as an example. During Stage 2, while the control group employed conventional peer interaction and feedback mechanisms to deepen knowledge understanding, the experimental group engaged in real-time interaction with GenAI to receive immediate feedback and personalized tutoring. Building upon this foundation, GenAI provided targeted solutions to help learners identify and address knowledge gaps, thereby facilitating deeper conceptual understanding.
[See PDF for image]
Fig. 4
Learning case in Stage 2
Stage3: Application and Transfer. This stage requires learners to apply their acquired knowledge by devising a research proposal or a plan for instructional enhancement. Discussions revolve around actionable topics like “Designing a Knowledge-Building English Classroom,” culminating in the completion of research design and implementation plans. Figure 5 demonstrates the learning process differences between the two groups using a representative Stage 3 learning task as an example. During this advanced learning stage, both groups engaged with higher-order tasks. The control group primarily relied on peer interactions to obtain feedback after proposing their instructional designs, subsequently refining their work based on this feedback. In contrast, the experimental group leveraged GenAI tools to acquire innovative pedagogical ideas and solutions. Building on this advantage, experimental group members could then share their GenAI-assisted design proposals in the platform’s discussion forum, where peer feedback enabled further optimization of their instructional designs.
[See PDF for image]
Fig. 5
Learning case in Stage 3
Following each topic within the three learning phases, the instructor assigns discussion tasks on the platform, which students are required to complete before a specified deadline. In the control group’s learning process, the primary resources are those predefined by the instructor, and learners primarily acquire knowledge through literature reviews and structured discussions. Learners in the control group experienced a longer feedback cycle, necessitating greater reliance on individual experience and group collaboration for knowledge integration and application.
In the experimental group, to ensure the output results were both reasonable and scientifically valid, the instructor explained and demonstrated how to use GenAI as a learning aid in the classroom, using a prominent large language model in China as an example, prior to the formal discussion. Over the initial two weeks of the course, the instructor demonstrated the use of GenAI to support the completion of learning tasks, as depicted in Fig. 6. Through task assignments and instructor-led feedback demonstrations, learners in the experimental group achieved proficiency in using prompts to generate high-quality outputs from GenAI. Specifically, the prompt design framework for GenAI-assisted learning is structured as follows: prompt design = role setting + task + generating subject + details of learning needs + learning form (Wang et al., 2024). In-service learners can integrate their professional experience with GenAI support to accomplish learning tasks. GenAI delivers dynamic and personalized learning resources, facilitates real-time interaction and feedback, and rapidly generates tools such as knowledge maps and case studies to enable learners to efficiently construct knowledge systems.
[See PDF for image]
Fig. 6
The instructor demonstrates how to use GenAI to assist in completing tasks
Data collection
To analyze the impact of GenAI on the knowledge building process of in-service learners across different learning phases, this study divided the whole teaching process into three phases and collected discussion data from both groups actively participating in topic-based learning on the platform. After eliminating some discussions that were not related to the course content, a total of 2053 valid data were obtained (including 1159 from the control group and 894 from the experimental group). As depicted in Table 2, the experimental group exhibits a higher volume of discussion data compared to the control group during stage 1 and stage 2 of the course, with a decrease in the later stage. This disparity may stem from GenAI’s ability to dynamically tailor content to learners’ needs, which reduces the confusion and knowledge blindness at the beginning stage of learning. The intensive interaction facilitated by GenAI early in the semester may enhance learning efficiency and potentially diminish the necessity for later remediation of knowledge gaps.
Table 2. Stage-based comparison of knowledge building frequencies across two groups
Stage | Control group | Experimental group | ||
|---|---|---|---|---|
Frequency (n) | Proportion (%) | Frequency (n) | Proportion (%) | |
Stage 1 | 227 | 19.59 | 188 | 21.03 |
Stage 2 | 554 | 47.8 | 520 | 58.16 |
Stage 3 | 378 | 32.61 | 186 | 20.81 |
Total | 1159 | 100 | 894 | 100 |
Data analysis
This study adopted the interaction analysis model proposed by Gunawardena et al. (1997) as the foundational coding framework, categorizing knowledge building into five distinct levels: information sharing, information analysis, information negotiation, information modification, and information application. Recent studies, including Liu et al. (2022), have expanded the analytic framework to include evaluation and reflection, processes that can involve either a cognitive process of evaluating cognitive objects or a metacognitive process of evaluating and reflecting on cognitive activities. However, existing frameworks often fail to distinguish between these processes. Consequently, this study integrated information evaluation into the coding framework, which is divided into six dimensions: information sharing, information analysis, information negotiation, information modification, information application and information evaluation. Within this framework, information evaluation is to evaluate and reflect on opinions, comments, programs, etc. Table 3 provides the dimensions, definitions, coding guidelines, and examples of the coding framework.
Table 3. A coding framework for in-service teachers’ knowledge building process
Categories | Description | Exemplar excerpts from student discussion data |
|---|---|---|
Information sharing (C1) | Sharing information, viewpoints, comments, etc., related to the topic or task. Includes supplementing personal evaluation | “Learning science is a highly practical course. It introduces us to a new perspective on education, exploring how humans learn.” |
Information analysis (C2) | Comparison, analysis, interpretation, questioning, refuting, arguing, reasoning, summarizing, or revising information | “I have a deeper understanding of scaffolding. Like a building, scaffolding not only supports the construction of the building, but additional scaffolding is added when the construction workers need to reach a higher position and can be removed when the building is completed. In an effective learning environment, scaffolding is gradually added, modified and removed as learners need it, and eventually the scaffolding will disappear completely…” |
Information negotiation (C3) | Expanding ideas by questioning and understanding others’ perspectives, adjusting personal viewpoints, and reaching consensus or unified solutions | “I wonder if my peers have participated in literary circle reading activities?” |
Information modification (C4) | Test and revise ideas | “The concepts are well summarized; in the process, I suggest incorporating the teacher as a guiding role; and the evaluation criteria can be enriched for reference.” |
Information application (C5) | Apply and integrate knowledge to form programs, plans, goals, etc | “After studying the course ‘Project-based Learning and Problem-based Learning’ shared by the teacher, I have gained a deeper understanding and knowledge of project-based learning and problem-based learning, which also guides me to further reflect on how to design project-based learning tasks in the classroom. I take the 2018 gaokao English Tianjin volume topic ‘‘world youth robotics skills competition’ as an example to illustrate how to design project-based learning tasks in the classroom ……” |
Information evaluation (C6) | Evaluate and reflect on ideas, comments, programs, etc | “It’s a good project practice, which is also an interdisciplinary practice activity and exercises students’ abilities in various aspects.” |
This study conducted a content analysis of platform discussion data utilizing the six-dimensional coding framework presented in Table 3, which is specifically designed to assess the knowledge building process of learners. To ensure the accuracy of the coding results, two researchers who were familiar with the course content and the coding framework independently coded part of the discussion data (approximately 30% of the total dataset). The Cohen’s Kappa coefficient of 0.834 was calculated after the two researchers completed the coding independently, signifying a high level of reliability in the coding outcomes. Following this, the two researchers discussed and negotiated to resolve coding inconsistencies, and the remaining data were divided equally before coding was completed by each of the two researchers.
This study employed two analytical approaches to uncover the effects of GenAI assistance on the knowledge building processes of in-service learners: comparative frequency analysis and epistemic network analysis. Based on the above coding framework, the study initially cleaned and coded the data, subsequently calculating the frequency of each knowledge-building dimension for both learner groups. Subsequently, the coded discussion data were analyzed using the ENA web tool (https://app.epistemicnetwork.org) to map the epistemic networks for both groups of learners, both holistically and across the three learning phases. This analysis aimed to analyze differences in knowledge building between the two groups. Finally, the ENA web tool was employed to map the epistemic networks of knowledge building for both groups, thereby exploring the roles Gen AI plays within the knowledge building process.
Results and discussion
To address question 1 and question 2, this study analyzes learners’ academic performance and the conversation data of two groups on the platform, aiming to explore whether GenAI can promote learners’ high-level knowledge building compared with the traditional model and to identify the characteristics presented in the process of knowledge building. To address question 3, this study collected conversation data across three learning stages for stage-wise comparative analysis, with the goal of examining the distinct roles GenAI assumes during various stages of knowledge building.
An analysis of the role of GenAI in knowledge building
Analysis of learners’ academic performance
At the conclusion of the instructional activities, learners were required to work collaboratively in groups to produce a comprehensive instructional strategy or research design report. Final evaluations were conducted through a combination of instructor evaluation and peer assessment, generating a composite score that served as an indicator of learners’ academic performance and learning outcomes. The evaluation criteria covered four key dimensions: the practical value and innovativeness of the selected topic, the coherence between the research design and research questions, the theoretical appropriateness of the instructional design, and the procedural rigor of the implementation process.
To examine performance differences between the experimental and control groups, a Mann–Whitney U test was conducted (see Table 4). The results revealed a statistically significant difference in academic performance (z = − 5.723, p = 0.00 < 0.05), with the experimental group obtaining significantly higher scores. These findings offer preliminary empirical evidence that the GenAI-assisted instructional model can more effectively enhance the professional development of in-service teachers compared to conventional instructional approaches. Notably, its benefits appear particularly salient in the domains of instructional planning and pedagogical innovation, suggesting considerable potential for technology-supported learning in professional teacher education.
Table 4. Results of academic performance assessment between two groups
Group | N | Mean | Sd | z | p |
|---|---|---|---|---|---|
Control group | 69 | 84.076 | 6.214 | − 5.723 | 0.000 |
Experimental group | 74 | 89.450 | 3.293 |
Frequency analysis of conversation behavior
In this study, the frequency and percentage of each dimension of knowledge building among the two groups of learners were quantified and are detailed in Table 5. The statistical analysis reveals two primary results. Result (i): Within the control group, the dimensions of “information sharing” (29.42%) and “information evaluation” (30.97%) were predominant, whereas the experimental group had a higher proportion of “information analysis” (42.95%) and “information evaluation” (19.57%). Result (ii): When compared with the control group, the experimental group demonstrated superior performance in the dimensions of “information analysis” (42.95 vs 9.74%) and “information modification” (7.71 vs 0.17%). Conversely, the control group outperformed in “information evaluation” (30.97 vs 19.57%), “information sharing” (29.42 vs 17.56%), and “information negotiation” (13.97 vs 2.23%).
Table 5. Frequency statistics of knowledge building dimensions across two groups
Code | Control group | Experimental group | ||
|---|---|---|---|---|
Frequency (n) | Proportion (%) | Frequency (n) | Proportion (%) | |
C1 | 341 | 29.42 | 157 | 17.56 |
C2 | 113 | 9.74 | 384 | 42.95 |
C3 | 162 | 13.97 | 20 | 2.23 |
C4 | 2 | 0.17 | 69 | 7.71 |
C5 | 148 | 12.76 | 86 | 9.61 |
C6 | 359 | 30.97 | 175 | 19.57 |
Irrelevant speech | 34 | 2.93 | 3 | 0.33 |
Total | 1159 | 100 | 894 | 100 |
It can be seen from the result (ii) that the GenAI-assisted learning model may have advantages in supporting high-level information processing and analysis but may fall short in facilitating collaborative interactions such as information sharing and negotiation. The superior performance of the experimental group in “information analysis” and “information modification” suggests that GenAI-assisted learning excels in aiding students in analyzing and applying information more deeply. This could be attributed to GenAI tools’ capacity to generate and synthesize vast amounts of information, thereby assisting students in filtering and deeply contemplating the rich information. With GenAI’s assistance, learners can reduce the cognitive load associated with information search and preliminary processing, freeing up cognitive resources to engage in higher-order cognitive activities such as information analysis. Consequently, the GenAI model appears to better support the development of higher-order cognitive skills in knowledge building, including critical analysis and synthesis. The relatively poor performance of the experimental group in “information sharing” and “information negotiation” may indicate the GenAI model’s limitations in promoting interaction, collaboration, and knowledge sharing among students. In a group or team learning environment, students might become overly reliant on AI-generated content, leading to a more individual-centered mode of analysis and potentially neglecting communication and collaboration with peers. The higher frequency of “information sharing” and “information negotiation” in the traditional model suggests that students are more inclined to construct knowledge through communication, which fosters a sense of collaboration and teamwork. Furthermore, it is noteworthy that, as shown in result (i), both the experimental and control groups performed well in “information evaluation”, with the experimental group showing slightly weaker performance. This may be related to the characteristics of the content generated by GenAI. While GenAI can rapidly produce a large volume of information to aid student learning, its automated content generation might lead to a lack of critical scrutiny regarding the authenticity or accuracy of the content, thereby diminishing the initiative of information evaluation. In contrast, in the traditional model, students rely more on autonomy to acquire and validate information, which may motivate them to be more proactive in information evaluation.
Epistemic network analysis
In order to compare and analyze the differences in knowledge building process between the two groups of learners, this study quantitatively coded the learners’ online discussion data and drew epistemic network diagrams based on the ENA web tool. Epistemic network analysis (ENA), a method for visualizing complex cognitive structures, was used to create multi-dimensional relational networks that intuitively represent the knowledge linkages and conceptual interactions of learners during online discussions.
In this study, we use the ENA web tool to calculate the co-occurrence of each element of the two groups of learners and draw an epistemic network diagram of knowledge building to realize the visual presentation of learners’ knowledge building (as shown in Figs. 7 and 8). Figure 7 illustrates the epistemic network plots of the two groups of learners, in which the dots represent the centroid of all students participating in online learning, the squares represent the average centroid of the epistemic network structure for each group, and the dashed boxes indicate the 95% confidence intervals. As can be seen in Fig. 7, the centroids of the epistemic networks for the two groups are in different positions, suggesting significant disparities in their cognitive network structures. To statistically validate these differences, this study conducted a Mann–Whitney U-test to compare the mean centroids of the epistemic networks between the groups. The results showed that the two groups of learners differed significantly in the X-dimension (p = 0.00 < 0.05).
[See PDF for image]
Fig. 7
Epistemic network centroid (blue for experimental group and red for control group)
[See PDF for image]
Fig. 8
Epistemic network of two groups (blue for experimental group and red for control group). Panel (a) shows the control group's epistemic network, (b) shows the experimental group's epistemic network, and (c) presents the superimposed subtraction diagrams of the networks for both groups
The analysis of epistemic networks, based on both single dimensions and the associations between dimensions, can reveal the differences in knowledge building between two groups of learners. Figure 8a and b present the epistemic network diagrams for the experimental and control groups, while Fig. 8c displays the superimposed subtraction diagrams of the networks for both groups. The superimposed subtraction diagrams of the epistemic networks visualize the differences between the same elemental connections of the learners in the GenAI-assisted and traditional modes, presenting the difference between the stronger connections minus the weaker ones as well as the distances between the centers of mass of the cognitive networks of the two groups.
The analysis of single dimensional features reveals a significant difference in the epistemic network focus between the experimental and control groups. From Fig. 8a and b, it can be concluded that most of the discussions in the control group focused on the dimensions of “information sharing” and “information evaluation”, whereas the experimental group concentrated more on “information analysis” and “information evaluation”. This result aligns with earlier findings (Result i), indicating that GenAI-assisted learning encourages deeper engagement with information processing and analysis, while learners in the traditional mode focus more on communication and sharing of information.
In terms of association dimension characteristics, the strength of connections between different elements in the epistemic networks exhibited greater variability between the groups. From the comprehensive analysis of the data in Fig. 8 and Table 6, it can draw the result (iii): learners of both groups show stronger strength in the connection of “information sharing-information analysis”, but weaker in “information negotiation-information modification” and “information modification-information application”. Further comparisons lead to result (iv): the control group has higher strength in the connections of the elements “information sharing-information negotiation”, “information sharing-information application”, “information sharing-information evaluation” and “information negotiation-information evaluation” than the experimental group. Conversely, the experimental group has higher strength in the connections of the elements “information sharing-information modification”, “information analysis-information modification”, “information analysis-information application”, “information analysis-information evaluation” and “information modification-information evaluation” than the control group.
Table 6. Co-occurrence coefficients of epistemic networks across two groups
connection | Control group | Experimental group |
|---|---|---|
C1-C2 | 0.25 | 0.27 |
C1-C3 | 0.26 | 0.02 |
C1-C4 | 0 | 0.11 |
C1-C5 | 0.3 | 0.06 |
C1-C6 | 0.47 | 0.18 |
C2-C3 | 0.12 | 0.04 |
C2-C4 | 0 | 0.18 |
C2-C5 | 0.13 | 0.32 |
C2-C6 | 0.19 | 0.46 |
C3-C4 | 0 | 0.02 |
C3-C5 | 0.11 | 0.03 |
C3-C6 | 0.2 | 0.07 |
C4-C5 | 0.01 | 0.06 |
C4-C6 | 0 | 0.19 |
From result (iv), it is evident that the association dimension characteristics highlight GenAI’s strength in facilitating the establishment of deep knowledge building relationships among learners. However, these results also reveal its relative weakness in supporting negotiation and application processes, a pattern consistent with the findings from the result (ii). In the control group, the connections between elements such as “information sharing-information negotiation,” “information sharing- information application,” “information sharing-information evaluation,” and “information negotiation-information evaluation” are stronger than those in the experimental group. This suggests that knowledge building in the traditional mode is more centered on information sharing, which fosters feedback, evaluation, and negotiation. These processes promote collaboration and interaction among students. In contrast, the experimental group shows significantly higher connection strengths in the association dimensions of “information sharing-information modification,” “information analysis-information modification,” “information analysis-information application,” “information analysis-information evaluation,” and “information modification-information evaluation.” This indicates that learners in the GenAI-assisted mode are more focused on information analysis. With the support of GenAI-generated content, learners can interpret and analyze information from multiple perspectives, making further modifications, applying knowledge, and evaluating their findings. This approach encourages critical reflection, helping students to develop deeper insights and enhancing their ability to reconstruct information. As a result, the GenAI-assisted model is more effective in supporting students’ engagement in higher-order cognitive processes, facilitating a deeper understanding of the material and promoting critical thinking. As demonstrated in result (iii), learners performed better in the primary stages of knowledge building—specifically, in “information sharing” and “information analysis,” which form the foundational stages of many collaborative learning tasks. However, these early stages may remain at a shallow level, failing to lead to deeper restructuring or reorganization of information. Negotiation, modification, and application are crucial stages in knowledge building, directly impacting learners’ ability to process information at a deeper level and construct new meanings. Weak performance in these areas can hinder the quality of knowledge building, potentially resulting in superficial learning. Since this study divides the teaching process into three stages, each with a different task, in order to further analyze the role played by GenAI in different tasks, a comparative analysis will be carried out by stage in Sect. “Intrinsic factors in GenAI’s role in knowledge building for In-Service teachers: A cross-temporal analysis”.
Discussion
Building on the previous analysis, the following responses address the findings related to Question 1 and Question 2. Regarding Question 1, the results indicate that the GenAI-assisted instructional model not only enhances the academic performance of in-service learners but also presents clear advantages over traditional learning models, particularly in supporting learners’ engagement with high-level information processing and critical thinking. Learners in the GenAI-assisted mode outperformed those in the traditional model in dimensions such as “information analysis” and “information modification”. The latter, “information modification”, involves critical evaluation and restructuring of information, including the analysis of peers’ opinions, which is considered an advanced level of knowledge building (Gunawardena et al., 1997; Roseli & Umar, 2015). This suggests that GenAI facilitates cognitive processes by reducing the load associated with information search and preliminary processing, allowing learners to allocate more cognitive resources to higher-level cognitive activities such as analysis and modification, and to a certain extent, it can effectively improve learners’ information analysis, and critical thinking. This finding is consistent with existing research, which has pointed out that GenAI can provide learners with personalized support and feedback to stimulate critical thinking by challenging them with customized sets of questions that are tailored to the student’s proficiency level (Azaria et al., 2024). Moreover, GenAI’s ability to understand complex queries and deliver real-time, relevant responses enhances learners’ comprehension of the subject matter (Elkhodr et al., 2023; Farrokhnia et al., 2024). This capability enables students to analyze material more effectively, thereby improving their research skills and critical reflection (Kasneci et al., 2023). Scholars have also noted GenAI’s potential to help students identify complex concepts and enhance problem-solving abilities within their discipline (Lamb et al., 2021; Ledesma et al., 2017). The transformative impact of GenAI on teaching and learning practices has been widely recognized. Research indicates that GenAI not only improves academic test scores compared to traditional learning environments (Alneyadi & Wardat, 2023), but also fosters higher engagement and better overall learning experiences (Ferrarelli & Iocchi, 2021; Ledesma & García, 2017).
Regarding Question 2, the results indicate that, compared to traditional learning models, GenAI-assisted learning fosters deeper individual cognition while, to some extent, reducing social interaction and collaboration among learners. In the knowledge-building process, learners in the traditional mode perform better in dimensions such as “information evaluation”, “information sharing”, and “information negotiation”, while learners using GenAI excel in “information analysis” and “information modification”. This suggests that the traditional model relies more on group negotiation and information sharing, where learners are more inclined to reach consensus through collaborative discussion. In contrast, the GenAI-assisted model emphasizes individual analysis and independent thinking, fostering a more autonomous learning process. “Information negotiation” is essential as it facilitates the transition from basic information sharing to higher-level knowledge building, where learners engage with differing perspectives and negotiate solutions (Yang et al., 2015). This process also helps learners develop generalized skills such as collaborative skills, which are critical for working in teams (Engelmann et al., 2009). In the experimental group, the increased use of GenAI resulted in a more individualized learning process, leading to fewer opportunities for social interaction and collaboration, which may limit the development of teamwork skills. Research has shown that learners involved in online collaborative learning, unlike face-to-face interaction, often lack the group cohesion necessary for effective knowledge building, making it harder to create new insights and meaning (Pifarré et al., 2014). Furthermore, UNESCO’s, 2021 report Reimagining Our Future Together highlights the need for future education systems to not only develop cognitive abilities but also equip students with a full range of social and emotional skills, including teamwork and collaboration (UNESCO, 2021). These skills will be essential in preparing students for future societal needs. Therefore, educators must strike a balance between fostering individual cognitive skills and promoting social collaboration. In the future, teachers should focus on strengthening learners’ collaborative abilities while using generative AI tools thoughtfully and responsibly. By integrating these tools into teaching practices, educators can help develop the skills necessary for success in a collaborative, future-oriented society.
Intrinsic factors in GenAI’s role in knowledge building for in-service teachers: a cross-temporal analysis
Frequency analysis of conversation behavior
In this study, the frequency and percentage of each dimension of knowledge building across different learning stages were counted for both the control and experimental groups, with the statistical results presented in Table 7. The following key results can be derived from the data.
Table 7. Frequency statistics of knowledge building dimensions across different learning stages in two groups
Stage | Categories | Control group | Experimental group | ||
|---|---|---|---|---|---|
Frequency (n) | Proportion (%) | Frequency (n) | Proportion (%) | ||
Stage 1 | C1 | 74 | 34.10 | 99 | 53.51 |
C2 | 18 | 8.29 | 85 | 45.95 | |
C3 | 60 | 27.65 | 0 | 0 | |
C4 | 0 | 0 | 0 | 0 | |
C5 | 7 | 3.23 | 0 | 0 | |
C6 | 58 | 26.73 | 1 | 0.54 | |
Stage 2 | C1 | 119 | 21.83 | 48 | 9.27 |
C2 | 65 | 11.93 | 154 | 29.73 | |
C3 | 74 | 13.58 | 20 | 3.86 | |
C4 | 0 | 0 | 69 | 13.32 | |
C5 | 74 | 13.58 | 67 | 12.93 | |
C6 | 213 | 39.08 | 160 | 30.89 | |
Stage 3 | C1 | 148 | 40.77 | 10 | 5.41 |
C2 | 30 | 8.26 | 145 | 78.38 | |
C3 | 28 | 7.71 | 0 | 0 | |
C4 | 2 | 0.55 | 0 | 0 | |
C5 | 67 | 18.46 | 19 | 10.27 | |
C6 | 88 | 24.24 | 11 | 5.95 | |
Result (iv): In the first stage, “information sharing” was the dominant dimension for both the control and experimental groups. However, the experimental group had a significantly higher percentage (53.51 vs. 34.10%). At this stage, there were significant differences between the two groups of learners. Compared to the control group, the experimental group exhibited more “information sharing” (53.51 vs. 34.10%) and “information analysis,” (45.95 vs. 8.30%). In contrast, the control group showed higher percentages in “information negotiation” (27.65 vs. 0.00%) and “information evaluation” (26.73 vs. 0.54%).
Result (v): In the second stage, “information evaluation” had the highest percentage for both groups, but the control group had a higher proportion (39.08 vs. 30.89%). During this phase, the experimental group had a greater emphasis on “information analysis” (29.73 vs. 11.93%) and “information modification” (13.32 vs. 0.00%), while the control group showed higher percentages in “Information sharing” (21.83 vs. 9.27%) and “Information negotiation” (13.58 vs. 3.86%).
Result (vi): Compared with the first two phases, the two groups of learners showed significant changes in their conversational behavior in the third phase. In the third stage, “Information sharing” was the most prevalent dimension in the control group (40.77%), while “information analysis” dominated in the experimental group (78.38%). At this stage, the experimental group outperformed the control group in “information analysis” (78.38 vs. 8.26%). The control group, on the other hand, performed better in “information sharing” (40.77 vs. 5.41%), “information application” (18.46% vs. 10.27%), and “information evaluation” (24.24 vs. 5.95%).
Overall, there were significant differences between the two groups of learners. The experimental group demonstrated a more focused approach to “information analysis,” while the control group’s conversational behavior appeared more scattered. This suggests that GenAI-assisted learning facilitates deeper information analysis to a considerable extent, with this effect progressively strengthening across the three learning stages. As can be seen from the result (iv), in the first stage, the experimental group exhibited significantly higher percentages of “information sharing” and “information analysis” compared to the control group. This indicates that GenAI’s intervention helped learners filter and analyze relevant information more efficiently during the stages of information reception and initial comprehension. In contrast, the control group showed higher percentages in “information negotiation” and “information evaluation,” suggesting a greater reliance on peer discussion and external feedback to validate information. However, this approach appears less effective at this stage, as it does not support as much in-depth individual analysis. As shown in Result (v), in the second stage, the experimental group showed a significantly higher percentage in “information analysis” and “information modification” compared to the control group. Notably, the experimental group made more substantial progress in “information modification,” which highlights GenAI’s role in promoting critical reflection and self-correction. In contrast, the control group exhibited higher percentages in “information sharing” and “information negotiation,” indicating that learners in this group preferred a more collaborative approach to knowledge building through discussion and negotiation. While this approach aids in deepening understanding, it is somewhat limited in fostering individualized, in-depth analysis and independent thinking.
As shown in Result (vi), in the third stage, the experimental group showed a dominant focus on “information analysis”, with 78.38% of their interactions falling under this dimension. This reflects the GenAI-assisted approach’s ability to foster deeper and more independent analysis. In summary, GenAI-assisted learning significantly enhanced learners’ ability to analyze and reflect during the knowledge building process, promoting higher autonomy and deeper cognitive engagement. In contrast, the traditional teaching model places more emphasis on negotiation and information sharing to facilitate group-based knowledge building.
Epistemic network analysis
To further analyze the differences in knowledge building across different learning stages, this study employed epistemic network analysis to compare and examine the knowledge building processes of the two groups of learners at each stage. Epistemic network analysis provides a concrete representation of the connections between the discourse behaviors of the two groups, enabling a clearer examination of the differences in the development of learners’ knowledge building processes across the two learning modes. Figure 9 presents the epistemic network diagrams for both groups at each of the three learning stages. The analysis reveals significant differences between the two groups along the X-axis at all three stages of learning (p = 0.00 < 0.05), while no significant differences were observed along the Y-axis.
[See PDF for image]
Fig. 9
Epistemic network and its subtraction for two groups at three stages (blue for experimental group and red for control group)
In Stage 1 (1–2 weeks), learners primarily discussed conceptual explanations or shared personal prior experiences. Figure 9a, b, and c illustrate notable differences in the epistemic network structures of the two groups during the early stage of instruction. Further analysis leads to the following two key results. Result (vii): The control group concentrated on the dimensions of “information sharing”, “information negotiation”, and “information evaluation”. In the association dimension, the control group demonstrated stronger connections in the following pairs: “information sharing-information negotiation”, “information sharing-information evaluation”, and “information negotiation-information evaluation”. Result (viii): In contrast, the experimental group focused on “information sharing” and “information analysis”. The experimental group showed a stronger correlation between “information sharing-information analysis” compared to the control group. Results (vii) revealed that, in the traditional learning model, learners were more inclined to share their understanding, discuss and negotiate viewpoints, and engage in preliminary evaluation of information during the initial phase of the task. This indicates a more interaction- and collaboration-oriented approach at the start of the task. From result (viii), it can be seen that in the GenAI-assisted mode, learners are more concerned with acquiring and analyzing information and less concerned with negotiation and evaluation when approaching a new topic, which may be attributed to the fact that the GenAI tool provides a large amount of information to help learners move directly to the stage of information analysis, which leads to faster deeper understanding and processing of knowledge. As shown in the elemental connections, learners in the GenAI-assisted mode demonstrated a stronger correlation between information sharing and information analysis, which supports the idea that the GenAI tool encourages independent, deep cognitive processing. In this model, students rely less on dialogue and interaction to clarify and expand ideas, instead focusing more on individual cognitive processing.
In Stage 2 (3–9 weeks), learners are tasked with understanding and integrating what they have learned from their prior experiences. Figure 9d, e, and f illustrate significant differences in the epistemic network structures between the two groups of learners during the middle phase of the teaching period. Further analysis leads to the following two key results. Result (ix): In the control group, learners’ conversational behaviors at this stage focused on five dimensions: information sharing, information analysis, information negotiation, information application, and information evaluation. In terms of the association dimension, the strongest connection was found between “information sharing-information evaluation”, with other associations appearing more balanced. This suggests that learners in the control group demonstrated more complex interaction patterns through a diverse range of discourse behaviors. Result (x): In contrast, the experimental group focused on four dimensions: information analysis, information evaluation, information modification, and information application. In terms of association, the experimental group showed stronger connections between “information analysis-information evaluation”. From Result (ix), it is evident that in the traditional model, learners placed greater emphasis on information sharing, information negotiation, and information evaluation during the process of knowledge internalization. These behaviors helped learners deepen their understanding of the material through mutual discussions and evaluations. The high strength of the connection “information sharing-information evaluation” indicates that learners in the control group tend to calibrate their understanding through sharing, peer feedback, and collective evaluation. This approach reflects a characteristic of traditional learning, where knowledge internalization is facilitated through interactive exchanges and discussions. It supports learners in integrating others’ perspectives with their own, enhancing the application and relevance of the knowledge. In contrast, as shown in Result (x), learners in the GenAI-assisted mode placed greater emphasis on information analysis, information evaluation, and information modification in the middle phase of the learning process. This shift enabled them to more proactively reconstruct their knowledge through in-depth interactions with GenAI-generated content. The stronger connection between Information analysis and Information evaluation suggests that learners in the experimental group were more focused on critically analyzing and reflecting on the information. This process enhances learners’ ability to engage in self-feedback and critically examine their understanding of the material. As learners revised their existing knowledge, they engaged in higher-level knowledge building, which is characteristic of the GenAI-assisted model. Compared to the control group, the GenAI-assisted mode places greater emphasis on information reprocessing and reflective application, allowing learners to achieve a higher degree of autonomy and personalized understanding in their knowledge internalization.
In Stage 3 (10–16 weeks), the primary learning task involves proposing a research design or an instructional improvement program by synthesizing the knowledge acquired during the previous stages. Figure 9g, h, and i reveal the following two key results. Result (xi): In the control group, learners’ conversational behaviors at this stage primarily centered around “information sharing”, “information application”, and “information evaluation”. In terms of associations, the control group exhibited strong connections in “information sharing-information application” and “information sharing-information evaluation”. Result (xii): In the experimental group, learners’ conversations were focused on “information analysis” and “information application”. The association dimension revealed stronger connections in “information analysis-information application” and “information analysis-information evaluation”. As can be seen from the result (xi), it is apparent that in the traditional mode, as learners approached the comprehensive task, they focused on sharing knowledge and experience, as well as evaluating feedback to refine and validate the program’s rationale. This approach reflects a reliance on group wisdom for validating and improving the design, with less emphasis on individual, in-depth analysis or reflection. The “information sharing- information evaluation” and “information sharing-information application” connections highlight the group-based nature of knowledge building in the control group, where collaborative discussion and feedback help shape the design. In contrast, result (xii) illustrates that learners in the GenAI-assisted mode placed greater emphasis on information analysis and evaluation. These analyses, coupled with reflective evaluation, enabled learners to apply knowledge more innovatively and transfer it to new contexts. The higher connection strength in “information analysis-information application”, “information analysis-information sharing” and “information analysis-information evaluation” indicates that learners in the experimental group engaged in multiple levels of cognitive processing during the design phase. This included deep analysis, feedback evaluation, reflective thinking, and the application of knowledge. The GenAI-assisted model encouraged learners to engage in a more systematic and reflective approach to problem-solving, which facilitated a higher level of autonomy and cognitive depth in their design process.
Overall, the experimental group (GenAI-assisted mode) and the control group (traditional mode) demonstrated significant cognitive differences across the three learning stages. As the teaching tasks evolved from basic knowledge comprehension to more advanced application and innovation design, the results from (vii), (ix), and (xi) suggest that the traditional model facilitated rapid knowledge building in the early and middle stages, particularly in tasks related to basic knowledge comprehension and internalization. This was largely due to the control group’s emphasis on knowledge sharing and collaborative learning within the group. In these stages, the traditional model promoted effective information exchange, enabling students to grasp foundational knowledge quickly and facilitating group-based consensus and evaluation. However, as tasks progressed to higher-order cognitive activities (particularly in advanced stages requiring innovation and complex problem-solving), the traditional model demonstrated significant limitations. The control group’s designs often relied on group wisdom, with relatively less emphasis on individual reflection or personalized innovation, as indicated by the findings in Results (vii), (ix), and (xi). This reliance on group collaboration hindered the deeper, individual cognitive processing required for innovation and the development of unique, personalized solutions. In contrast, the experimental group (GenAI-assisted mode), as shown in Results (viii), (x), and (xii), exhibited stronger tendencies toward individual analysis and autonomous cognitive processing from the outset. The integration of GenAI tools enabled these learners to achieve a deeper, more personalized understanding of the material early on. In the middle and later stages, the GenAI-assisted mode proved particularly effective in supporting critical reflection, individualized modification, and innovative application. These cognitive shifts allowed students in the experimental group to develop more unique insights and innovative solutions, aligning with the demands of higher-order cognition and creative problem-solving.
Discussion
In response to question 3, the comparison of results across the three learning phases reveals that GenAI provides varying levels of support depending on the complexity of the learning tasks. Compared to the traditional model, GenAI-assisted learners exhibited less significant benefits in low-order tasks, while demonstrating more pronounced advantages in higher-order, complex tasks. The relatively limited role of GenAI in lower-order tasks stems from the fact that learners can achieve these tasks independently, as they require less cognitive effort and minimal support. In contrast, GenAI enhances learners’ critical reflection and innovative thinking in higher-order tasks by offering extensive information resources and advanced analysis tools, enabling deeper, more personalized reflection and creative problem-solving. Existing research highlights the potential of GenAI to support learners in higher-order tasks, such as imaginative and high-quality creative writing prompts in writing tasks (Gupta et al., 2023), which enhance critical writing skills (Liebrenz et al., 2023). Therefore, in classroom teaching and professional training, educators should strategically incorporate GenAI tools according to the complexity and goals of learning tasks. For low-order tasks, the focus should be on enhancing foundational skills and summarizing key concepts. In contrast, higher-order tasks should employ more constructive and creative teaching approaches, fostering critical thinking and innovation, and facilitating the effective management and application of knowledge to support personalized growth and development. This dynamic interplay between GenAI’s support features and task-based practice enhances each other, providing new perspectives and innovative approaches for improving teaching strategies.
In summary, GenAI demonstrates distinct advantages in facilitating higher-order knowledge building tasks. Its role in supporting critical thinking, innovative application, and complex data analysis provides both a theoretical foundation and a practical basis for the deeper integration of GenAI technology in future educational practices. This finding suggests that future educational research should focus on the differentiated application of GenAI tools across various cognitive tasks. This would help provide more targeted technical support to meet the personalized learning needs in diverse educational contexts.
Conclusions, implications, and future directions
This study provides an empirical foundation for the practical application of generative AI in teaching and learning by examining the impact of AI tools on learners’ knowledge-building processes. Based on the findings, it proposes the following prospects for teacher professional development. These recommendations aim to guide and optimize learning in the era of intelligent technologies.
Teacher training should focus on collaborative skill development
Based on the results, this study draws several conclusions regarding the cognitive and collaborative implications of GenAI-assisted learning for in-service teachers. On the one hand, it can be seen that GenAI can significantly enhance learners’ academic performance and promote the development of cognitive skills such as information analysis and critical thinking. This result is consistent with existing research findings, and many scholars who have studied the application of GenAI have pointed out that GenAI enhances learners’ collaborative learning achievement to some extent (An et al., 2024; Liu et al., 2024). On the other hand, the results suggest that GenAI may reduce opportunities for social interaction and collaboration. This result diverges from existing research findings. It has been shown that GenAI can positively facilitate group discussion and accelerate the formation of collective intelligence in complex tasks (Yilmaz & Yilmaz, 2023). This discrepancy may be related to the participants’ discussion patterns and the overall atmosphere. In the experimental group, after sharing personal insights and engaging in peer interactions, participants frequently utilized GenAI tools for individual task comprehension, cognitive development, and planning, potentially accounting for the limited collaborative interactions observed. Notably, the GenAI instructional model proposed in this study was designed to facilitate higher-order cognitive activities among in-service teachers. Although the course objectives encouraged participants to actively engage in high-quality interactions with others, in practice, teachers often relied on their existing experiences, preferred solving problems independently rather than collaborating in groups, and struggled to construct knowledge in group settings (Barratt-Pugh et al., 2019; Hong et al., 2011; Walton & Rusznyak, 2019). Most in-service teachers tended to create posts to share their detailed insights. Particularly with the involvement of GenAI, in-service teachers could obtain more immediate and comprehensive responses through interactions with the tool, which led to a preference for working independently rather than collaboratively. Furthermore, analysis of the interaction data revealed that exchanges among in-service learners and their peers were often confined to emotional expressions, such as agreement or gratitude. For example, phrases like “I agree with this.”, “Great job”, or “Thanks for sharing” were common, but these types of exchanges were insufficient to foster meaningful and constructive interactions. Productive interactions are multifaceted, reciprocal, and iterative processes that depend on active engagement with both peers and content (Zhang et al., 2017). Each iteration is essentially cyclical, where one interaction receives input from another and provides output for subsequent interactions (Hou, 2015). Collaboration, however, is a core competency in 21st-century education, fostering essential skills such as teamwork, communication, and the ability to integrate multiple perspectives—skills that are vital for addressing complex problems and creative tasks. These findings carry important implications for the design of AI-supported professional development.
Based on these findings, this study proposes the following instructional strategies to enhance the collaborative capabilities of in-service teachers in an AI-supported environment. First, it is essential to provide an efficient and accessible online learning environment equipped with various synchronous and asynchronous discussion tools to support teacher collaboration. Simultaneously, in-service teachers should be guided on how to effectively utilize GenAI tools to foster deeper collaboration. The effective support of both the environment and tools can significantly enhance the efficiency of collaboration. Research by An et al. (2024) suggests that insufficient personal learning time for teachers may be a major factor affecting their interaction levels and social knowledge construction. A well-designed online discussion environment and tools may help address this issue. Second, instructors need to implement appropriate instructional strategies or scaffolding to facilitate high-quality collaborative knowledge co-construction among in-service learners. In teacher training, the instructor plays a pivotal role and must possess relevant skills, including posing questions, maintaining focus in discussions, providing active feedback, establishing norms, regular monitoring, and offering technical assistance (Wang, 2008). Instructional strategies such as providing metacognitive strategies, fostering collective responsibility, and stimulating autonomous agency can improve teachers’ collaborative construction levels (Scardamalia, 2002). For instance, some studies have introduced knowledge frameworks or metacognitive note-taking to guide teachers during their learning process. By employing metacognitive scaffolding strategies, these studies demonstrated effective improvements in collaborative learning outcomes (Ouyang et al., 2021). Additionally, fostering a sense of community is a critical factor. Research indicates that a sense of collective responsibility can significantly influence the knowledge building process (Ouyang et al., 2021). During the learning process, it is important to actively engage learners’ collective agency, encouraging them to take on the role of knowledge constructors. For example, instructors might introduce role rotation exercises, allowing in-service learners to take on various roles in collaborative tasks. This approach can enhance learners’ collaborative communication skills and improve their ability to engage in productive teamwork. Research has shown that assigning roles in collaborative settings can effectively boost participants’ engagement, foster task-related dialogues, and increase learners’ awareness of the collaborative process (Schellens et al., 2017). Studies suggest that when collaborative skill development is combined with AI technology, learners can not only deepen their cognitive understanding but also enhance key soft skills necessary for future success (Ruiz-Rojas et al., 2024). These findings underscore the importance of preparing learners not only in terms of their academic knowledge but also in their interpersonal and teamwork skills. In summary, teacher training within the GenAI-assisted learning environment should prioritize the development of collaborative skills, and exploring the integration of AI technologies with collaborative learning strategies. This approach can foster a synergistic enhancement of both cognitive and collaborative competencies, equipping learners with the skills needed to thrive in future learning and work environments.
The long-term sustainability of integrating GenAI into teacher professional development
The findings of this study clearly indicate that GenAI excels in supporting higher-order, complex tasks but exhibits limited effectiveness in addressing routine, lower-order tasks. The results reveal the underlying patterns of knowledge building among in-service teachers within the GenAI-assisted instructional model and offer important insights into the transformation of learning approaches in teacher professional development. Specifically, GenAI provides flexible cognitive support and contextual resources for in-service teachers, playing a vital role in facilitating self-directed, constructivist learning and helping them reconcile the tension between work and study. Existing research also suggests that GenAI plays a more significant role in complex tasks compared to simple ones (Yilmaz & Yilmaz, 2023). In educational contexts, the integration of GenAI should be reconsidered—not as a one-size-fits-all solution, but rather as a context-sensitive support system whose effectiveness depends on task complexity and the learner’s epistemic needs. In lower-order tasks, GenAI may serve a supplementary role in reducing cognitive load and facilitating information retrieval. In contrast, for higher-order tasks, its generative and interactive capabilities can stimulate learners’ critical thinking, creativity, and collaborative inquiry. This approach aligns with the core principle of Vygotsky’s Zone of Proximal Development (ZPD) theory: GenAI can function as a cognitive scaffold, extending learners’ developmental space beyond what they could achieve independently. Therefore, the instructional application of GenAI should be aligned with the cognitive complexity of the task and the technological proficiency of the teacher. Notably, beyond China, GenAI tools such as ChatGPT and Bard have been widely adopted in Europe and the United States across various educational contexts, including academic writing (Maghamil & Sieras, 2024), programming (Johnson et al., 2024), and other subject-specific applications. These applications reveal the potential of GenAI to personalize instruction, scaffold learner autonomy, and support disciplinary reasoning. Teachers must equip their students with the skills to understand and utilize the technologies essential for their future careers. To ensure the sustainable and equitable integration of GenAI into education, several practical and policy-level implications must be considered. Therefore, ensuring the long-term sustainability of GenAI in teacher training is imperative. First, teacher education curricula should explicitly incorporate GenAI literacy training that goes beyond tool operation to include critical evaluation, ethical use, and pedagogical design. Teachers must be empowered to make informed decisions about when, why, and how to integrate GenAI to foster deep learning and facilitate meaningful interactions. Additionally, policymakers should advocate for the integration of GenAI into curriculum standards, define its application scenarios and objectives in teaching and learning, and ensure effective implementation through teacher training and technical support. Finally, policymakers should encourage cross-disciplinary collaboration and promote the joint development of GenAI integration frameworks for diverse educational environments, involving educational technologists, subject teachers, and other stakeholders. Schools and educational institutions should implement long-term support mechanisms (Loble, 2023), including regular teacher workshops, technical support, and resource-sharing platforms, to enable teachers to continuously update their technological skills and adapt to the rapid evolution of educational technology.
Limitation and future research
This study provides an empirical case of GenAI-assisted learning, offering both theoretical grounding and practical guidance for establishing technology-enabled mechanisms for continuous teacher development. It holds significant educational impact and practical value for broader implementation. Furthermore, the study identifies three key limitations that should be addressed in future research. First, this study was conducted using a large language model developed in China. This type of GenAI platform generates responses based on cue words and logical relationships, which presents technical limitations. Specifically, it lacks true semantic understanding and real-world comprehension, leading to potential inaccuracies in output, as well as privacy and security concerns. The participants in this study were all in-service teachers with a certain level of subject matter expertise and the ability to critically assess the GenAI-generated answers. When applying this model to learners with less prior knowledge, especially those starting from a “zero base”, it is essential for teachers to provide appropriate guidance and interventions. Secondly, regarding data collection, this study relied on discussion data from the learning platform and learners’ performance scores, which may limit the comprehensiveness and accuracy of the results. To increase the validity and reliability of future research, it is recommended to collect data from multiple sources, such as semi-structured interviews, activity trace analysis, and other observational measures. This will provide a more holistic view of the impact of GenAI on learners’ knowledge building and ensure a more accurate assessment of its effects. Furthermore, regarding research methodology, while this study employed epistemic network analysis, the approach presents limitations in effectively differentiating between the quality and quantity of learner discussions. The methodology primarily relies on undirected weighted network models to represent and quantify connections within coded data, which can only reflect the closeness of inter-node relationships (i.e., connection weights) but fails to distinguish qualitative attributes such as the depth and innovativeness of contributions. Additionally, this undirected network structure cannot capture the temporal characteristics of behavioral events. As learner-GenAI interactions constitute a dynamic process, the current methodology inadequately analyzes the sequence and developmental trajectories of these interactions, thereby limiting the investigation of dynamic mechanisms in knowledge construction. To enhance analytical depth and methodological sophistication, future studies could implement improvements in two dimensions. First, adopting mixed-methods designs that integrate epistemic network analysis with qualitative coding. This would involve manual annotation to classify discussion quality (e.g., basic questioning, in-depth exploration, innovative application) and subsequently incorporating these quality dimensions into network models, enabling dual characterization of both quantitative and qualitative aspects. Second, employing temporally sensitive methods such as sequence analysis or ordered network analysis (ONA) to construct directed weighted networks. These approaches would facilitate analysis of behavioral event sequences, duration patterns, and evolutionary pathways in learner-GenAI interactions (Tan et al., 2024), thereby providing more comprehensive insights into knowledge construction mechanisms within GenAI-assisted learning environments. Finally, regarding research design, this study did not consider the influence of other potential factors on learners’ knowledge construction within a GenAI-assisted learning environment. Future research should explore the roles of self-regulation, learner motivation, leadership, technology acceptance, self-efficacy, and other individual differences in shaping learning outcomes in GenAI-assisted contexts. For instance, a pre-test and post-test design could be used to compare groups of learners with varying levels of self-regulation or leadership, further investigating the most effective instructional models in GenAI-supported environments. Investigating these factors could yield a deeper understanding of how learners interact with AI tools and the broader cognitive and motivational processes involved.
Acknowledgements
We would like to extend our gratitude to all the in-service teachers who participated in this course, as well as the teaching assistants Zhang Yixin and Guo Yunjing, for their active cooperation and invaluable support during the implementation of this study.
Author contributions
HZ: Conceptualization, writing-review & editing, methodology, investigation, data curation and visualization. QW: Conceptualization, writing-review & editing and supervision. All authors read and approved the final manuscript.
Funding
This study was funded by: Research on the Mechanism for the Automatic Generation of Personalized Learning Resources and its Use in Adaptive Learning (Grant Number: 62307004), National Natural Science Foundation of China; Fundamental Research Funds for the Central Universities (Grant Number: 2024JJ003;2024JX066).
Data availabilty
Data will be made available on request.
Declarations
Competing interests
The authors declare that there is no conflict of interest regarding the publication of this paper. No financial or personal relationships that could inappropriately influence or bias the content of this work.
Abbreviations
Generative artificial intelligence
Teacher professional development
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
Alneyadi, S; Wardat, Y. ChatGPT: Revolutionizing student achievement in the electronic magnetism unit for eleventh-grade students in Emirates schools. Contemporary Educational Technology; 2023; 15,
An, S; Zhang, S; Guo, T; Lu, S; Zhang, W; Cai, Z. Impacts of generative AI on student teachers’ task performance and collaborative knowledge construction process in mind mapping-based collaborative environment. Computers & Education; 2024; [DOI: https://dx.doi.org/10.1016/j.compedu.2024.105227]
Asiri, YA; Millard, DE; Weal, MJ. Assessing the impact of engagement and real-time feedback in a mobile behavior change intervention for supporting critical thinking in engineering research projects. IEEE Transactions on Learning Technologies; 2021; 14,
Avalos, B. Teacher professional development in teaching and teacher education over ten years. Teaching and Teacher Education; 2011; 27,
Avcı, Ü. Examining the role of sentence openers, role assignment scaffolds and self-determination in collaborative knowledge building. Educational Technology Research and Development; 2020; 68,
Azaria, A; Azoulay, R; Reches, S. ChatGPT is a remarkable tool—for experts. Data Intelligence; 2024; 6,
Bahroun, Z; Anane, C; Ahmed, V; Zacca, A. Transforming education: A comprehensive review of generative artificial intelligence in educational settings through bibliometric and content analysis. Sustainability; 2023; 15,
Baidoo-Anu, D; Ansah, LO. Education in the era of generative artificial intelligence (AI): Understanding the potential benefits of ChatGPT in promoting teaching and learning. Journal of AI; 2023; 7,
Banihashem, SK; Noroozi, O; van Ginkel, S; Macfadyen, LP; Biemans, HJ. A systematic review of the role of learning analytics in enhancing feedback practices in higher education. Educational Research Review; 2022; 37, [DOI: https://dx.doi.org/10.1016/j.edurev.2022.100489] 100489.
Barratt-Pugh, L; Zhao, F; Zhang, Z; Wang, S. Exploring current Chinese higher education pedagogic tensions through an activity theory lens. Higher Education; 2019; 77, pp. 831-852. [DOI: https://dx.doi.org/10.1007/s10734-018-0304-8]
Bauer, E; Sailer, M; Kiesewetter, J; Shaffer, DW; Schulz, C; Pfeiffer, J; Gurevych, I; Fischer, MR; Fischer, F. Pre-service teachers’ diagnostic argumentation: What is the role of conceptual knowledge and epistemic activities?. Proceedings of the fifteenth international conference of the learning sciences; 2020; International Society of the Learning Sciences (ISLS): pp. 2399-2400.
Belay, S; Melesse, T. Exploring the link between teachers’ motivation for continuous professional development and professional learning communities: A structural equation modeling approach. SAGE Open; 2024; [DOI: https://dx.doi.org/10.1177/21582440241281855]
Bereiter, C; Scardamalia, M. Resnick, LB. Intentional learning as a goal of instruction. Knowing, learning, and instruction: Essays in honor of Robert Glaser; 1989; Lawrence Erlbaum Associates: pp. 361-392.
Bereiter, C; Scardamalia, M. Tan, SC; So, HJ; Yeo, J. Knowledge building and knowledge creation: One concept, two hills to climb. Knowledge creation in education; 2014; Springer Singapore: pp. 35-52. [DOI: https://dx.doi.org/10.1007/978-981-287-047-6_3]
Blanchard, MR; LePrevost, CE; Tolin, AD; Gutierrez, KS. Investigating technology-enhanced teacher professional development in rural, high-poverty middle schools. Educational Researcher; 2016; 45,
Boling, CJ; Martin, SH. Supporting teacher change through online professional development. The Journal of Educators Online; 2005; 2,
Bozkurt, A; Karadeniz, K; Baneres, D; Guerrero-Roldán, AE; Rodríguez, ME. Artificial intelligence and reflections from educational landscape: A review of AI studies in half a century. Sustainability; 2021; 13,
Bressler, DM; Bodzin, AM; Eagan, B; Tabatabai, S. Using epistemic network analysis to examine discourse and scientific practice during a collaborative game. Journal of Science Education and Technology; 2019; 28, pp. 553-566. [DOI: https://dx.doi.org/10.1007/s10956-019-09786-8]
Brown, R; Munger, K. Learning together in cyberspace: Collaborative dialogue in a virtual network of educators. Journal of Technology and Teacher Education; 2010; 18,
Chan, CKY; Hu, W. Students’ voices on generative AI: Perceptions, benefits, and challenges in higher education. International Journal of Educational Technology in Higher Education; 2023; 20,
Chang, CH; Kidman, G. The rise of generative artificial intelligence (AI) language models—challenges and opportunities for geographical and environmental education. International Research in Geographical and Environmental Education; 2023; 32,
Chen, W; Tan, JS; Pi, Z. The spiral model of collaborative knowledge improvement: An exploratory study of a networked collaborative classroom. International Journal of Computer-Supported Collaborative Learning; 2021; 16,
Chen, Y; Jensen, S; Albert, LJ; Gupta, S; Lee, T. Artificial intelligence (AI) student assistants in the classroom: Designing chatbots to support student success. Information Systems Frontiers; 2023; 25, pp. 161-182. [DOI: https://dx.doi.org/10.1007/s10796-022-10291-4]
Chiu, TK. A classification tool to foster self-regulated learning with generative artificial intelligence by applying self-determination theory: A case of ChatGPT. Educational Technology Research and Development; 2024; [DOI: https://dx.doi.org/10.1007/s11423-024-10366-w]
Chiu, TK. Future research recommendations for transforming higher education with generative AI. Computers and Education: Artificial Intelligence; 2024; 6, [DOI: https://dx.doi.org/10.1016/j.caeai.2023.100197] 100197.
Chu, HC; Hwang, GH; Tu, YF; Yang, KH. Roles and research trends of artificial intelligence in higher education: A systematic review of the top 50 most-cited articles. Australasian Journal of Educational Technology; 2022; 38,
Compen, B; De Witte, K; Schelfhout, W. The role of teacher professional development in financial literacy education: A systematic literature review. Educational Research Review; 2019; 26, pp. 16-31. [DOI: https://dx.doi.org/10.1016/j.edurev.2018.12.001]
Cooper, G. Examining science education in ChatGPT: An exploratory study of generative artificial intelligence. Journal of Science Education and Technology; 2023; 32,
Cress, U; Kimmerle, J. Suthers, DD; Lund, K; Rosé, CP; Teplovs, C; Law, N. Successful knowledge building needs group awareness: Interaction analysis of a 9th grade CSCL biology lesson. Productive multivocality in the analysis of group interactions; 2013; Springer, US: pp. 495-509. [DOI: https://dx.doi.org/10.1007/978-1-4614-8960-3_27] 1279.97010
Csanadi, A; Eagan, B; Kollar, I; Shaffer, DW; Fischer, F. When coding-and-counting is not enough: Using epistemic network analysis (ENA) to analyze verbal data in CSCL research. International Journal of Computer-Supported Collaborative Learning; 2018; 13, pp. 419-438. [DOI: https://dx.doi.org/10.1007/s11412-018-9292-z]
Dai, W; Lin, J; Jin, H; Li, T; Tsai, YS; Gašević, D; Chen, G. Can large language models provide feedback to students? A case study on ChatGPT. 2023 IEEE international conference on advanced learning technologies (ICALT); 2023; IEEE: pp. 323-325. [DOI: https://dx.doi.org/10.1109/icalt58122.2023.00100]
Dede, C; Jass Ketelhut, D; Whitehouse, P; Breit, L; McCloskey, EM. A research agenda for online teacher professional development. Journal of Teacher Education; 2009; 60,
Desimone, LM; Porter, AC; Garet, MS; Yoon, KS; Birman, BF. Effects of professional development on teachers’ instruction: Results from a three-year longitudinal study. Educational Evaluation and Policy Analysis; 2002; 24,
Elkhodr, M; Gide, E; Wu, R; Darwish, O. ICT students’ perceptions towards ChatGPT: An experimental reflective lab analysis. STEM Education; 2023; 3,
Engelmann, T; Dehler, J; Bodemer, D; Buder, J. Knowledge awareness in CSCL: A psychological perspective. Computers in Human Behavior; 2009; 25, pp. 949-960. [DOI: https://dx.doi.org/10.1016/j.chb.2009.04.004]
Farrokhnia, M; Banihashem, SK; Noroozi, O; Wals, A. A SWOT analysis of ChatGPT: Implications for educational practice and research. Innovations in Education and Teaching International; 2024; 61,
Feng, X., Van Aalst, J., Chan, C. K. K., & Yang, Y. (2019). Fostering collective knowledge advancement through idea-friend maps in a large knowledge-building community. Computer Supported Collaborative Learning. https://repository.isls.org/bitstream/1/1580/1/288-295.pdf
Fernandez-Nieto, G. M., Martinez-Maldonado, R., Kitto, K., & Buckingham Shum, S. (2021). Modelling spatial behaviours in clinical team simulations using epistemic network analysis: methodology and teacher evaluation. In LAK21: 11th international learning analytics and knowledge conference. Pp. 386–396.
Ferrarelli, P; Iocchi, L. Learning Newtonian physics through programming robot experiments. Technology, Knowledge and Learning; 2021; 26,
Fishman, B; Konstantopoulos, S; Kubitskey, BW; Vath, R; Park, G; Johnson, H; Edelson, DC. Comparing the impact of online and face-to-face professional development in the context of curriculum implementation. Journal of Teacher Education; 2013; 64,
Goodfellow, I; Pouget-Abadie, J; Mirza, M; Xu, B; Warde-Farley, D; Ozair, S; Bengio, Y. Generative adversarial networks. Communications of the ACM; 2020; 63,
Gunawardena, CN; Lowe, CA; Anderson, T. Analysis of a global online debate and the development of an interaction analysis model for examining social construction of knowledge in computer conferencing. Journal of Educational Computing Research; 1997; 17, pp. 397-431. [DOI: https://dx.doi.org/10.2190/7mqv-x9uj-c7q3-nrag]
Gupta, BB; Gaurav, A; Panigrahi, PK; Arya, V. Analysis of artificial intelligence-based technologies and approaches on sustainable entrepreneurship. Technological Forecasting and Social Change; 2023; 186, [DOI: https://dx.doi.org/10.1016/J.TECHFORE.2022.122152] 122152.
Hadwin, AF; Järvelä, S; Miller, M. Zimmerman, BJ; Schunk, DH. Self-regulated, co-regulated, and socially shared regulation of learning. Handbook of self-regulation of learning and performance; 2011; Routledge: pp. 65-84. [DOI: https://dx.doi.org/10.4324/9780203839010.ch5]
Hong, HY; Chen, FC; Chai, CS; Chan, WC. Teacher-education students’ views about knowledge building theory and practice. Instructional Science; 2011; 39, pp. 467-482. [DOI: https://dx.doi.org/10.1007/s11251-010-9143-4]
Hou, H. What makes an online community of practice work? A situated study of Chinese student teachers’ perceptions of online professional learning. Teaching and Teacher Education; 2015; 46, pp. 6-16. [DOI: https://dx.doi.org/10.1016/j.tate.2014.10.005]
Huang, L; Liang, M; Xiong, Y; Wu, X; Lim, CP. A systematic review of technology-enabled teacher professional development during COVID-19 pandemic. Computers & Education; 2024; [DOI: https://dx.doi.org/10.1016/j.compedu.2024.105168]
Hwang, GJ; Yang, LH; Wang, SY. A concept map-embedded educational computer game for improving students’ learning performance in natural science courses. Computers & Education; 2013; 69, pp. 121-130. [DOI: https://dx.doi.org/10.1016/j.compedu.2013.07.008]
Ilieva, G; Yankova, T; Klisarova-Belcheva, S; Dimitrov, A; Bratkov, M; Angelov, D. Effects of generative Chatbots in higher education. Information; 2023; 14,
Jensen, LX; Buhl, A; Sharma, A; Bearman, M. Generative AI and higher education: A review of claims from the first months of ChatGPT. Higher Education; 2024; [DOI: https://dx.doi.org/10.1007/s10734-024-01265-3]
Johnson, DM; Doss, W; Estepp, CM. Using ChatGPT with Novice Arduino programmers: Effects on performance, interest, self-efficacy, and programming ability. Journal of Research in Technical Careers; 2024; [DOI: https://dx.doi.org/10.9741/2578-2118.1152]
Kadaruddin, K. Empowering education through Generative AI: Innovative instructional strategies for tomorrow’s learners. International Journal of Business, Law, and Education; 2023; 4,
Kalinowski, E; Egert, F; Gronostaj, A; Vock, M. Professional development on fostering students’ academic language proficiency across the curriculum—A meta-analysis of its impact on teachers’ cognition and teaching practices. Teaching and Teacher Education; 2020; 88, [DOI: https://dx.doi.org/10.1016/j.tate.2019.102971] 102971.
Kasneci, E; Seßler, K; Küchemann, S; Bannert, M; Dementieva, D; Fischer, F; Kasneci, G. ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences; 2023; 103, [DOI: https://dx.doi.org/10.1016/j.lindif.2023.102274] 102274.
Kelly, N; Antonio, A. Teacher peer support in social network sites. Teaching and Teacher Education; 2016; 56, pp. 138-149. [DOI: https://dx.doi.org/10.1016/j.tate.2016.02.007]
Kennedy, A. Models of continuing professional development: A framework for analysis. Journal of in-Service Education; 2005; 31,
King, F. Evaluating the impact of teacher professional development: An evidence-based framework. Professional Development in Education; 2014; 40,
Kohnke, L; Moorhouse, BL; Zou, D. ChatGPT for language teaching and learning. Relc Journal; 2023; 54,
Kwakman, K. Factors affecting teachers’ participation in professional learning activities. Teaching and Teacher Education; 2003; 19,
Lamb, R; Hand, B; Kavner, A. Computational modeling of the effects of the science writing heuristic on student critical thinking in science using machine learning. Journal of Science Education and Technology; 2021; 30,
Lämsä, J; Hämäläinen, R; Koskinen, P; Viiri, J; Lampi, E. What do we do when we analyse the temporal aspects of computer-supported collaborative learning? A systematic literature review. Educational Research Review; 2021; 33, [DOI: https://dx.doi.org/10.1016/j.edurev.2021.100387] 100387.
Ledesma, EFR; García, JJG. Selection of mathematical problems in accordance with student’s learning style. International Journal of Advanced Computer Science and Applications; 2017; [DOI: https://dx.doi.org/10.14569/IJACSA.2017.080316]
Lee, EYC; Chan, CKK; van Aalst, J. Students assessing their own collaborative knowledge building. International Journal of Computer-Supported Collaborative Learning; 2006; 1,
Li, Y; Li, X; Zhang, Y; Li, X. The effects of a group awareness tool on knowledge construction in computer-supported collaborative learning. British Journal of Educational Technology; 2021; 52,
Liang, J; Wang, L; Luo, J; Yan, Y; Fan, C. The relationship between student interaction with generative artificial intelligence and learning achievement: Serial mediating roles of self-efficacy and cognitive engagement. Frontiers in Psychology; 2023; 14, 1285392. [DOI: https://dx.doi.org/10.3389/fpsyg.2023.1285392]
Liebrenz, M; Schleifer, R; Buadze, A; Bhugra, D; Smith, A. Generating scholarly content with ChatGPT: Ethical challenges for medical publishing. The Lancet Digital Health; 2023; 5,
Liu, M; Zhang, LJ; Biebricher, C. Investigating students’ cognitive processes in generative AI-assisted digital multimodal composing and traditional writing. Computers & Education; 2024; 211, [DOI: https://dx.doi.org/10.1016/j.compedu.2023.104977] 104977.
Liu, Q; Chen, F; Zhang, S; Chen, J; Yang, X. Zaixian xiezuo xuexi duihua fenxi moxing goujian ji yingyong——ji yu yuan renzhi shijiao de yanjiu [Construction and application of dialogue analysis model in online collaborative learning—Based on metacognitive perspective]. Dianhua Jiaoyu Yanjiu; 2022; 11, pp. 53-59. [DOI: https://dx.doi.org/10.13811/j.cnki.eer.2022.11.007]
Loble, L. (2023). The rise of ChatGPT shows why we need a clearer approach to technology in schools. The Conversation. Retrieved March 14, 2025 from http://theconversation.com/the-rise-of-chatgpt-shows-why-we-need-a-clearer-approach-to-technology-in-schools-199596
Luan, H; Geczy, P; Lai, H; Gobert, J; Yang, SJH; Ogata, H; Baltes, J; Guerra, R; Li, P; Tsai, CC. Challenges and future directions of big data and artificial intelligence in education. Frontiers in Psychology; 2020; 11, [DOI: https://dx.doi.org/10.3389/fpsyg.2020.580820] 580820.
Maghamil, M; Sieras, S. Impact of ChatGPT on the academic writing quality of senior high school students. Journal of English Language Teaching and Applied Linguistics; 2024; 6,
Molenaar, I; de Mooij, S; Azevedo, R; Bannert, M; Järvelä, S; Gašević, D. Measuring selfregulated learning and the role of AI: Five years of research using multimodal multichannel data. Computers in Human Behavior; 2023; 139, [DOI: https://dx.doi.org/10.1016/j.chb.2022.107540] 107540.
Nazari, N; Shabbir, MS; Setiawan, R. Application of artificial intelligence powered digital writing assistant in higher education: Randomized controlled trial. Heliyon; 2021; 7,
Opfer, VD; Pedder, D. Conceptualizing teacher professional learning. Review of Educational Research; 2011; 81,
Oshima, J; Oshima, R; Fujita, W. A mixed-methods approach to analyze shared epistemic agency in jigsaw instruction at multiple scales of temporality. Journal of Learning Analytics; 2018; 5,
Ouyang, F; Hu, Y; Zhang, Y; Guo, Y; Yang, Y. In-service teachers’ knowledge building during face-to-face collaborative learning. Teaching and Teacher Education; 2021; 107, [DOI: https://dx.doi.org/10.1016/j.tate.2021.103479] 103479.
Parsons, SA; Hutchison, AC; Hall, LA; Parsons, AW; Ives, ST; Leggett, AB. US teachers’ perceptions of online professional development. Teaching and Teacher Education; 2019; 82,
Phillips, M; Trevisan, O; Carli, M; Mannix, T; Gargiso, R; Gabelli, L; Lippiello, S. Uncovering patterns and (dis) similarities of pre-service teachers through Epistemic Network Analysis. Society for information technology & teacher education international conference; 2023; Association for the Advancement of Computing in Education (AACE): pp. 1021-1030.
Pifarré, M; Cobot, R; Argelagós, E. Incidence of group awareness information on students’ collaborative learning processes. Journal of Computer Assisted Learning; 2014; 30, pp. 300-317. [DOI: https://dx.doi.org/10.1111/jcal.12043]
Powell, CG; Bodur, Y. Teachers’ perceptions of an online professional development experience: Implications for a design and implementation framework. Teaching and Teacher Education; 2019; 77, pp. 19-30. [DOI: https://dx.doi.org/10.1016/j.tate.2018.09.004]
Prieto, LP; Rodríguez-Triana, MJ; Ley, T; Eagan, B. The value of epistemic network analysis in single-case learning analytics: A case study in lifelong learning. Advances in quantitative ethnography: Second international conference, ICQE 2020, Malibu, CA ,USA ,February 1–3, 2021, Proceedings 2,; 2021; Springer International Publishing: pp. 202-217. [DOI: https://dx.doi.org/10.1007/978-3-030-67788-6_14]
Ray, PP. ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope. Internet of Things and Cyber-Physical Systems; 2023; 3, pp. 121-154. [DOI: https://dx.doi.org/10.1016/j.iotcps.2023.04.003]
Reeves, TD; Pedulla, JJ. Predictors of teacher satisfaction with online professional development: Evidence from the USA’s e-Learning for Educators initiative. Professional Development in Education; 2011; 37,
Roseli, MHM; Umar, IN. Students’ levels of knowledge construction and cognitive skills in an online forum learning environment. Social and Behavioral Sciences; 2015; 197, pp. 1983-1989. [DOI: https://dx.doi.org/10.1016/j.sbspro.2015.07.574]
Ruis, AR; Lee, SB. Advances in quantitative ethnography: Second international conference, ICQE 2020, Malibu, CA, USA, February 1–3, 2021, Proceedings; 2021; Springer Nature: [DOI: https://dx.doi.org/10.1007/978-3-030-67788-6]
Ruiz-Rojas, LI; Salvador-Ullauri, L; Acosta-Vargas, P. Collaborative working and critical thinking: Adoption of generative artificial intelligence tools in higher education. Sustainability; 2024; 16,
Sadler, TD; Friedrichsen, P; Zangori, L; Ke, L. Technology-supported professional development for collaborative design of COVID-19 instructional materials. Journal of Technology and Teacher Education; 2020; 28,
Sancar, R; Atal, D; Deryakulu, D. A new framework for teachers’ professional development. Teaching and Teacher Education; 2021; 101, [DOI: https://dx.doi.org/10.1016/j.tate.2021.103305] 103305.
Santos, RPD. Enhancing chemistry learning with ChatGPT and Bing Chat as agents to think with: A comparative case study. SSRN Electronic Journal; 2023; [DOI: https://dx.doi.org/10.2139/ssrn.4447416]
Scardamalia, M. Smith, B. Collective cognitive responsibility for the advancement of knowledge. Liberal education in a knowledge society; 2002; Open Court: pp. 67-98.
Schellens, T; Van Keer, H; Valcke, M; De Wever, B. The impact of role assignment as scripting tool on knowledge construction in asynchronous discussion groups. Computer supported collaborative learning 2005; 2017; Routledge: pp. 557-566. [DOI: https://dx.doi.org/10.3115/1149293.1149367]
Shin, Y; Kim, D; Jung, J. The effects of representation tool (visible-annotation) types to support knowledge building in computer-supported collaborative learning. Journal of Educational Technology & Society; 2018; 21,
Siemens, G; Baker, RSD. Learning analytics and educational data mining: Towards communication and collaboration. Proceedings of the 2nd international conference on learning analytics and knowledge; 2012; Association for Computing Machinery: pp. 252-254. [DOI: https://dx.doi.org/10.1145/2330601.2330661]
Swiecki, Z; Ruis, AR; Farrell, C; Shaffer, DW. Assessing individual contributions to collaborative problem solving: A network analysis approach. Computers in Human Behavior; 2020; 104, [DOI: https://dx.doi.org/10.1016/j.chb.2019.01.009] 105876.
Tan, Y; Swiecki, Z; Ruis, AR; Shaffer, D. Saqr, M; López-Pernas, S. Epistemic network analysis and ordered network analysis in learning analytics. Learning analytics methods and tutorials; 2024; Springer: [DOI: https://dx.doi.org/10.1007/978-3-031-54464-4_18]
Tlili, A; Shehata, B; Adarkwah, MA; Bozkurt, A; Hickey, DT; Huang, R; Agyemang, B. What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education. Smart Learning Environments; 2023; 10,
Tømte, CE; Gjerustad, C. Online or offline–Does it matter? A study of in-service teachers’ perceptions of learning outcomes in Norway. Nordic Journal of Digital Literacy; 2020; 15,
Tsiotakis, P; Jimoyiannis, A. Critical factors towards analysing teachers’ presence in on-line learning communities. The Internet and Higher Education; 2016; 28, pp. 45-58. [DOI: https://dx.doi.org/10.1016/j.iheduc.2015.09.002]
Uiterwijk-Luijk, L; Krüger, M; Zijlstra, B; Volman, M. Teachers’ role in stimulating students’ inquiry habit of mind in primary schools. Teaching and Teacher Education; 2019; 86, [DOI: https://dx.doi.org/10.1016/j.tate.2019.102894] 102894.
UNESCO. (2021). Reimagining our futures together: A new social contract for education [OL]. Retrieved November 23, 2024, from https://unesdoc.unesco.org/ark:/48223/pf0000379381.
van Aalst, J; Chan, CKK. Student-directed assessment of knowledge building using electronics portfolios. Journal of the Learning Sciences; 2007; 16,
Vescio, V; Ross, D; Adams, A. A review of research on the impact of professional learning communities on teaching practice and student learning. Teaching and Teacher Education; 2008; 24,
Walkoe, JDK; Luna, MJ. What we are missing in studies of teacher learning: A call for microgenetic, interactional analyses to examine teacher learning processes. Journal of the Learning Sciences; 2019; 29,
Walton, E; Rusznyak, L. Cumulative knowledge-building for inclusive education in initial teacher education. European Journal of Teacher Education; 2019; 43,
Wang, Q. A generic model for guiding the integration of ICT into teaching and learning. Innovations in Education and Teaching International; 2008; 45,
Wang, Q; Guo, F; Zhang, Y; Ma, S. Jiyu damoxing de shendu ketang biange yanjiu [Research on revolution path of deep-level thinking class]. Open Education Research; 2024; 30,
Weinberger, A; Biemans, HJ; Mulder, M; Chizari, M. Argumentation-based computer supported collaborative learning (ABCSCL): A synthesis of 15 years of research. Educational Research Review; 2012; 7, pp. 79-106. [DOI: https://dx.doi.org/10.1016/j.edurev.2011.11.006]
Wilson, S; Nishomoto, M. Assessing learning of computer programming skills in the age of generative artificial intelligence. Journal of Biomechanical Engineering. DOI; 2023; 10,
Wise, B; Emerson, L; Van Luyn, A; Dyson, B; Bjork, C; Thomas, SE. A scholarly dialogue: Writing scholarship, authorship, academic integrity and the challenges of AI. Higher Education Research & Development; 2024; 43,
Wood, D; Moss, SH. Evaluating the impact of students’ generative AI use in educational contexts. Journal of Research in Innovative Teaching & Learning; 2024; 17,
Wu, TT; Lee, HY; Li, PH; Huang, CN; Huang, YM. Promoting self-regulation progress and knowledge construction in blended learning via ChatGPT-based learning aid. Journal of Educational Computing Research; 2024; 61,
Xia, Q; Chiu, TKF; Chai, CS; Xie, K. The mediating effects of needs satisfaction on the relationships between prior knowledge and self-regulated learning through artificial intelligence chatbot. British Journal of Educational Technology; 2023; 54,
Xu, Y; Zhu, J; Wang, M; Qian, F; Yang, Y; Zhang, J. The impact of a digital game-based AI Chatbot on students’ academic performance, higher-order thinking, and behavioral patterns in an information technology curriculum. Applied Sciences; 2024; 14,
Yang, XM; Li, JH; Guo, XS; Li, XJ. Group interactive network and behavioral patterns in online English-to-Chinese cooperative translation activity. Internet and Higher Education; 2015; 25, pp. 28-36. [DOI: https://dx.doi.org/10.1016/j.iheduc.2014.12.003]
Yilmaz, R; Yilmaz, FGK. Augmented intelligence in programming learning: Examining student views on the use of ChatGPT for programming learning. Computers in Human Behavior: Artificial Humans; 2023; 1,
Yoon, SA; Miller, K; Richman, T. Comparative study of high-quality professional development for high school biology in a face-to-face versus online delivery mode. Educational Technology & Society; 2020; 23,
Yücel, ÜA; Usluel, YK. Knowledge building and the quantity, content and quality of the interaction and participation of students in an online collaborative learning environment. Computers & Education; 2016; 97, pp. 31-48. [DOI: https://dx.doi.org/10.1016/j.compedu.2016.02.015]
Zabolotna, K; Malmberg, J; Järvenoja, H. Examining the interplay of knowledge construction and group-level regulation in a computer-supported collaborative learning physics task. Computers in Human Behavior; 2023; 138, [DOI: https://dx.doi.org/10.1016/j.chb.2022.107494] 107494.
Zhang, X; Admiraal, W; Saab, N. Teachers’ motivation to participate in continuous professional development: Relationship with factors at the personal and school level. Journal of Education for Teaching; 2021; 47,
Zhang, J; Hong, HY; Scardamalia, M; Teo, CL; Morley, EA. Sustaining knowledge building as a principle-based innovation at an elementary school. Journal of the Learning Sciences; 2011; 20,
Zhang, S; Gao, Q; Sun, M; Cai, Z; Li, H; Tang, Y; Liu, Q. Understanding student teachers’ collaborative problem solving: Insights from an epistemic network analysis (ENA). Computers & Education; 2022; 183, [DOI: https://dx.doi.org/10.1016/j.compedu.2022.104485] 104485.
Zhang, S; Liu, Q; Chen, W; Wang, Q; Huang, Z. Interactive networks and social knowledge construction behavioral patterns in primary school teachers’ online collaborative learning activities. Computers & Education; 2017; 104, pp. 1-17. [DOI: https://dx.doi.org/10.1016/j.compedu.2016.10.011]
Zheng, L; Niu, J; Long, M; Fan, Y. An automatic knowledge graph construction approach to promoting collaborative knowledge building, group performance, social interaction and socially shared regulation in CSCL. British Journal of Educational Technology; 2023; 54,
Zhu, G; Fan, X; Hou, C; Zhong, T; Seow, P; Shen-Hsing, AC; Poh, TL. Embrace opportunities and face challenges: Using ChatGPT in undergraduate Students' collaborative interdisciplinary learning. arXiv; 2023; [DOI: https://dx.doi.org/10.48550/arXiv.2305.18616]
© The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.