A. Current Frameworks for Generative AI in Education
A.教育领域生成式人工智能的现有框架

In recent years, research and practice have increasingly focused on the integration of generative AI in educational settings [4], [7]. Various frameworks have been developed to guide educators and policymakers in harnessing the potential of generative AI tools.
近年来，研究和实践越来越关注在教育环境中整合生成式人工智能 [4] ， [7] 。人们制定了各种框架，以指导教育工作者和决策者利用生成式人工智能工具的潜力。

UNESCO has led the way in creating AI competency frameworks for both teachers and students, with the goal of fostering a thorough understanding of AI's importance in education. Both frameworks designed for teachers and for students emphasize adherence to human rights, the protection of human dignity and privacy, and the reinforcement of human agency [8]. However, the practical application of these frameworks in everyday educational settings remains challenging. The UNESCO guidelines do not provide detailed, actionable instructions that teachers can readily apply within varied classroom environments.
教科文组织率先为教师和学生制定了人工智能能力框架，目的是促进人们全面了解人工智能在教育中的重要性。为教师和学生设计的这两个框架都强调遵守人权、保护人的尊严和隐私，以及加强人的能动性 [8] 。 然而，在日常教育环境中实际应用这些框架仍然具有挑战性。教科文组织的指导方针没有提供详细的、可操作的指示，教师无法在不同的课堂环境中随时应用。

The Australian Framework for Generative Artificial Intelligence in Schools [13] outlines a set of guidelines tailored to the educational context, namely, teaching and learning; human and social well-being; transparency; fairness; accountability; and privacy, security, and safety. Regarding teaching and learning, the framework's holistic approach is commendable as it addresses several aspects:
《澳大利亚学校人工智能生成框架》 [13] 概述了一套针对教育背景的指导方针，即教学；人类和社会福祉；透明度；公平性；问责制；以及隐私、安保和安全。在教学方面，该框架的整体方法值得称赞，因为它涉及多个方面：

1. the reinforcement of educational practices through impactful integration;
   通过有影响力的整合加强教育实践；

2. instructional methods that cultivate a critical understanding of AI;
   培养对人工智能的批判性理解的教学方法；

3. teacher expertise to augment rather than replace human teaching;
   教师的专业知识是对人类教学的补充，而不是替代；

4. the development of critical thinking skills to foster intellectual growth;
   培养批判性思维能力，促进智力发展；

5. learning designs that prioritize students’ needs;
   优先考虑学生需求的学习设计；

6. the maintenance of academic integrity to guarantee the ethical application of AI in educational settings.
   维护学术诚信，保证人工智能在教育环境中的应用符合道德规范。

However, this framework may lack the specificity needed for practical application, specifically clear-cut guidance on how to implement AI tools in varied classroom situations. The effective integration of AI into curricula and teaching practices will require detailed strategies [14], which are not provided sufficiently by the current framework.
然而，这一框架可能缺乏实际应用所需的具体性，特别是缺乏关于如何在不同课堂情境中实施人工智能工具的明确指导。要将人工智能有效地融入课程和教学实践，就必须制定详细的策略 [14] ，而目前的框架并没有充分提供这些策略。

In addition to these official guidelines, several scholars have proposed frameworks. For example, Su and Yang [15] proposed a theoretical framework, “IDEE,” for generative AI in education. Their framework includes steps such as identifying desired outcomes, determining the level of automation, considering ethical impacts, and assessing effectiveness. Yet, this framework does not offer concrete steps for practical implementation in diverse educational settings. Chan [16] developed an AI education policy for higher education and indicated that students in higher education should take an active role in policy. Furthermore, Kong et al. [17] proposed a 6-P pedagogy, comprising a plan, prompt, preview, produce, peer-review, and portfolio-tracking framework, to guide university students’ academic writing.
除了这些官方指南，还有一些学者提出了一些框架。例如，Su 和 Yang [15] 为教育领域的生成式人工智能提出了一个理论框架 "IDEE"。他们的框架包括确定预期结果、确定自动化程度、考虑伦理影响和评估有效性等步骤。然而，该框架并未提供在不同教育环境中实际实施的具体步骤。Chan [16] 为高等教育制定了一项人工智能教育政策，并指出高等教育中的学生应在政策中发挥积极作用。此外，Kong 等人 [17] 提出了一种 6-P 教学法，包括计划、提示、预览、制作、同行评审和作品集跟踪框架，以指导大学生的学术写作。

These endeavors reveal a gap in research and an essential need for a K–12-specific framework that aligns generative AI's innovative capabilities with the developmental requirements of younger students. Such a framework should offer clear guidance for implementation across different educational environments, with a focus on centering human needs and perspectives.
这些努力揭示了研究方面的差距，也表明我们亟需一个针对 K-12 阶段的框架，将生成式人工智能的创新能力与低年级学生的发展要求结合起来。这种框架应为在不同的教育环境中实施提供明确的指导，重点是以人的需求和视角为中心。

B. Generative AI: Empowering Lifelong Learning
B.生成式人工智能：增强终身学习能力

Generative AI has catalyzed individualized learning [2], [3], [17]. By ensuring that learning experiences are tailored, generative AI provides a foundation for the continuous engagement and skill development that are essential for sustained educational growth throughout one's life [7], [9]. Chiu's [7] research delves into AI's diverse impacts on education, highlighting four pivotal roles of AI that support the development of lifelong learners. The first role involves using generative AI for personalization, such that tasks are customized to each learner's skill level. However, personalization may cause students to become too dependent on generative AI and take a passive approach to learning [18]. In K–12 education, therefore, it is vital to not only implement generative AI tools but also educate students about the foundational AI technologies, including the concepts of tokens, transformer algorithms, self-attention mechanisms, and embeddings. Students must comprehend these concepts to appreciate AI's capabilities and constraints and thus avoid an excessive dependence on technology.
生成式人工智能催化了个性化学习 [2] , [3] , [17] 。通过确保量身定制的学习体验，生成式人工智能为持续参与和技能发展奠定了基础，而持续参与和技能发展对于人一生的持续教育成长至关重要 [7] , [9] 。Chiu 的 [7] 研究深入探讨了人工智能对教育的各种影响，强调了人工智能在支持终身学习者发展方面的四个关键作用。第一个作用是利用生成式人工智能实现个性化，从而根据每个学习者的技能水平定制任务。然而，个性化可能会导致学生过于依赖人工智能生成器，从而采取被动的学习方式 [18] 。因此，在 K-12 教育中，至关重要的是不仅要实施生成式人工智能工具，还要教育学生了解人工智能的基础技术，包括标记、变换器算法、自我关注机制和嵌入等概念。学生必须了解这些概念，才能理解人工智能的能力和限制，从而避免过度依赖技术。

The second role of AI is its ability to enhance human–computer interactions. Michel-Villarreal et al. [19] found that using AI chatbots can improve students’ communication skills [20]. However, it is crucial to supplement human–computer interactions with teacher-led discussions to nurture empathy, emotional intelligence, and understanding, which AI currently cannot provide.
人工智能的第二个作用是增强人机互动的能力。Michel-Villarreal等人 [19] 发现，使用人工智能聊天机器人可以提高学生的交流技能 [20] 。 然而，关键是要通过教师主导的讨论来补充人机互动，以培养同理心、情商和理解力，而这正是人工智能目前无法提供的。

The third role of AI involves providing feedback on students’ work. In K–12 education, immediate feedback can help students to quickly identify areas for improvement and understand complex concepts [21]. Nevertheless, teachers should offer in-depth feedback that encompasses the more qualitative facets of students’ work, such as creativity and critical analysis. In addition, it is important to recognize that the content generated by AI may lack the nuance and depth of human insights [6], [22].
人工智能的第三个作用是为学生的作业提供反馈。在 K-12 教育中，即时反馈可以帮助学生快速确定需要改进的地方并理解复杂的概念 [21] 。不过，教师应提供深入的反馈，包括学生作业中更多的定性方面，如创造性和批判性分析。此外，必须认识到，人工智能生成的内容可能缺乏人类见解的细微差别和深度 [6] ， [22] 。

In its fourth role, AI can increase the engagement of students through prompt-based interactions with generative AI [23]. For K–12 students, continued engagement is vital for nurturing a lasting enthusiasm for learning. Thus, teachers should develop AI-assisted learning scenarios that encourage exploration, spark curiosity, and facilitate interactive learning [20], [24]. Teachers should address specific AI mechanisms, such as tokens, transformer algorithms, and embeddings, in their curricula to improve students’ understanding of generative AI.
人工智能的第四个作用是，通过与生成性人工智能 [23] 进行基于提示的互动，提高学生的参与度。对于 K-12 年级的学生来说，持续的参与对于培养持久的学习热情至关重要。因此，教师应开发人工智能辅助学习场景，鼓励探索，激发好奇心，促进互动学习 [20] ， [24] 。教师应在课程中涉及具体的人工智能机制，如代币、变换算法和嵌入，以提高学生对生成式人工智能的理解。

C. TPACK Framework and Generative AI in Education
C.TPACK 框架和教育领域的生成式人工智能

The TPACK framework is a vital model for understanding the integration of technology into educational practices [25], [26], [27]. Shulman [25] introduced the concept of PCK, highlighting the importance of teachers’ integrated understanding of both the content they teach and the pedagogical methods best suited for delivering this content. Mishra and Koehler [26] introduced technology knowledge (TK), and proposed that effective teaching in the digital age requires an understanding of how technology can be combined with pedagogical methods and CK. The TPACK framework thus consists of three primary forms of knowledge: PK, CK, and TK. It further includes four hybrid forms of knowledge that emerge at the intersections of the primary forms: PCK, technological pedagogical knowledge, TCK, and TPACK [26, p. 1025].
TPACK 框架是理解将技术融入教育实践的一个重要模型 [25] , [26] , [27] 。舒尔曼 [25] 提出了 PCK 的概念，强调了教师对所教内容和最适合传授这些内容的教学方法的综合理解的重要性。Mishra 和 Koehler [26] 介绍了技术知识（TK），并提出数字时代的有效教学需要了解如何将技术与教学方法和 CK 结合起来。因此，TPACK 框架包括三种主要的知识形式：PK、CK 和 TK。它还包括四种混合形式的知识，这些知识出现在主要形式的交叉点上：PCK、技术教学知识、TCK 和 TPACK [26，第 1025 页]。

The TPACK framework guides teachers in creating and implementing technology-enhanced learning experiences that are content-specific and pedagogically sound [27], [28]. TPACK reflects the dynamic integration of the technology, content, and instruction knowledge domains. It plays a pivotal role in guiding the contextual application of technology in instructional contexts [26] and is instrumental in cultivating teachers’ competency in integrating technology seamlessly into curriculum-specific teaching [27], [29].
TPACK 框架指导教师创建和实施技术强化学习体验，这些体验既针对具体内容，又符合教学规律 [27] , [28] 。TPACK 反映了技术、内容和教学知识领域的动态整合。它在指导技术在教学情境中的应用方面起着举足轻重的作用 [26] ，并有助于培养教师将技术无缝整合到特定课程教学中的能力 [27] , [29] 。

Several recent studies have explored teachers’ acceptance of generative AI in education using the technology acceptance model [30], [31], [32]. However, teacher development programmes in which TPACK is aligned with generative AI are scarce. Our study concentrates on four dimensions of CK: CK, TCK, PCK, and TPACK. CK underpins students’ learning and SRL objectives, TCK explores the affordances of generative AI for SRL, and PCK encompasses lesson planning strategies in which generative AI is integrated to support SRL. TPACK synthesizes all three dimensions into a cohesive strategy for designing student-centered courses that use generative AI to promote SRL in a unit teaching and learning.
最近有几项研究利用技术接受模型 [30] 、 [31] 、 [32] 探讨了教师对教育领域生成式人工智能的接受程度。 然而，将TPACK与生成式人工智能相结合的教师发展计划却很少见。我们的研究集中于四个维度的 CK：CK、TCK、PCK 和 TPACK。CK是学生学习和自学能力目标的基础，TCK探讨了生成式人工智能对自学能力的影响，PCK则包含了将生成式人工智能融入其中以支持自学能力的备课策略。TPACK 将所有三个维度综合成一个有凝聚力的策略，用于设计以学生为中心的课程，在单元教学中使用生成式人工智能促进 SRL。

To address the noted gap in research, the current study was conducted to support a teacher development programme tailored to integrate generative AI with teaching and learning. The attention, relevance, confidence, and satisfaction (ARCS) model of motivational design principles, developed by Keller [33], was adopted to guide the programme. The four components of ARCS are also crucial to the design of teaching and learning activities in using generative AI for engaging students for SRL development, drawing their attention, making the learning activities relevant to their experience, fostering confident development in using generative AI, and ultimately, enabling satisfaction after the learning process.
为了弥补研究方面的不足，本研究开展了一项教师发展计划，旨在支持将生成式人工智能与教学相结合的教师发展计划。该计划采用了凯勒 [33] 提出的动机设计原则--注意力、相关性、信心和满意度（ARCS）模型作为指导。ARCS 的四个要素对于设计使用生成式人工智能的教与学活动也至关重要，它们能吸引学生参与，促进他们的自学能力发展，吸引他们的注意力，使学习活动与他们的经验相关，培养他们使用生成式人工智能的自信心，并最终使他们在学习过程后获得满足感。

D. SRL via Generative AI
D.通过生成式人工智能实现 SRL

Zimmerman's three phases of SRL, namely, forethought, performance, and self-reflection, are essential for guiding students through domain-specific learning processes [10], [34]. The integration of generative AI into a curriculum can enhance each phase of SRL by providing individualized learning materials and immediate feedback that facilitates reflection on learning strategies and outcomes. However, current K–12 educational frameworks lack a systematic approach to integrate TPACK with generative AI and thus foster SRL.
齐默尔曼（Zimmerman）提出的自学学习（SRL）的三个阶段，即前思、表现和自省，对于指导学生完成特定领域的学习过程 [10] ， [34] 至关重要。将生成式人工智能整合到课程中，可以通过提供个性化的学习材料和即时反馈，促进对学习策略和结果的反思，从而增强自学能力的每个阶段。然而，目前的 K-12 教育框架缺乏将传统知识包（TPACK）与生成式人工智能相结合，从而促进 SRL 的系统方法。

In the forethought phase of SRL, generative AI can be used to analyze students’ existing knowledge and learning preferences, enabling the provision of tailored learning goals and resources [35], [36]. This individualized learning aids in establishing intrinsic motivation and strategic planning [19].
在 SRL 的前瞻阶段，生成式人工智能可用于分析学生的现有知识和学习偏好，从而提供量身定制的学习目标和资源 [35] ， [36] 。这种个性化学习有助于建立内在动力和战略规划 [19] 。

During the performance phase, generative AI can provide immediate feedback to empower students to monitor their understanding and quickly adjust their learning strategies [14]. For example, writing prompts can facilitate dialogue with generative AI. However, young learners are still developing their metacognitive and cognitive abilities [37], [38]; accordingly, they may experience frustration or confusion when working with AI. Thus, teachers play a vital role in providing emotional support and motivation to help students to overcome challenges [39].
在表现阶段，生成式人工智能可以提供即时反馈，使学生能够监控自己的理解并迅速调整学习策略 [14] 。例如，写作提示可以促进与生成式人工智能的对话。然而，青少年学生的元认知和认知能力仍处于发展阶段 [37] ， [38] ；因此，他们在使用人工智能时可能会遇到挫折或困惑。因此，教师在提供情感支持和激励以帮助学生克服困难方面起着至关重要的作用 [39] 。

In the self-reflection phase, students engage in self-evaluation to assess their own performance and reflect on their interactions with generative AI, enabling them to better plan for subsequent learning tasks [40]. This reflection is a key step toward consolidating learning and preparing for the future.
在自我反思阶段，学生进行自我评价，评估自己的表现，并反思自己与生成式人工智能的互动，从而更好地规划后续学习任务 [40] 。这种反思是巩固学习成果和为未来做好准备的关键一步。

However, the absence of an integrated TPACK and generative AI framework for SRL in K–12 education has left teachers without a clear strategy to harness AI's full potential in promoting SRL. A dedicated framework could guide teachers in effectively combining their pedagogical expertise and CK with generative AI to foster and support SRL. Such a framework would ensure a human-centered pedagogical approach and equip students with the necessary skills and mindset for lifelong learning.
然而，在 K-12 教育中，由于缺乏针对 SRL 的综合 TPACK 和生成式人工智能框架，教师们没有明确的策略来利用人工智能的全部潜力来促进 SRL。一个专门的框架可以指导教师有效地将他们的教学专业知识和 CK 与生成性人工智能结合起来，以促进和支持 SRL。这样一个框架将确保以人为本的教学方法，并使学生具备终身学习所需的技能和心态。

Against this background, a human-centered learning and teaching framework (HCLTF) was proposed in this study. This framework uses generative AI for SRL development through domain knowledge learning in K–12 settings. In addition, a 10-week teacher development programme was introduced to equip in-service primary teachers with conceptual knowledge about generative AI and the skills necessary to use generative AI tools for creating differentiated instructional materials and fostering students’ individualized learning.
在此背景下，本研究提出了一个以人为本的学习与教学框架（HCLTF）。该框架利用生成式人工智能，在 K-12 环境中通过领域知识学习促进自学能力的发展。此外，本研究还引入了一个为期10周的教师发展计划，以帮助在职小学教师掌握生成式人工智能的概念知识，以及使用生成式人工智能工具创建差异化教学材料和促进学生个性化学习所需的技能。

The programme includes an introduction to generative AI (e.g., concepts of generative AI, tokens, self-attention mechanisms, transformer, supervised and unsupervised learning, and reinforcement learning) and the application of generative AI to course design. The TPACK and ARCS models were integrated to guide the design of the programme. The proposed HCLTF was used to guide the pedagogical design of diverse case studies on subjects, such as Chinese language, English language, mathematics, general studies, and programming. The impact of the teacher professional development programme was examined by addressing the following three questions.
该课程包括对生成式人工智能的介绍（如生成式人工智能的概念、代币、自我注意机制、转换器、监督和非监督学习以及强化学习）以及生成式人工智能在课程设计中的应用。TPACK 模型和 ARCS 模型相结合，为课程设计提供指导。拟议的 HCLTF 用于指导不同学科（如中文、英文、数学、通识教育和编程）案例研究的教学设计。教师专业发展计划的影响通过以下三个问题进行了研究。

1. To what extent did the teacher development programme improve teachers’ comprehension of generative AI concepts?
   教师发展计划在多大程度上提高了教师对生成式人工智能概念的理解？

2. To what extent did the teacher development programme enhance teachers’ perceptions of the domains of CK, TCK, PCK, and TPACK?
   教师发展计划在多大程度上提高了教师对知识、能力和社区知识、个人成长知识和知识包领域的认识？

3. To what extent did the teacher development programme enhance teachers’ perceived ability to design courses that would increase students’ attention, relevance, confidence, and satisfaction?
   教师发展计划在多大程度上提高了教师设计课程的能力，从而提高学生的注意力、相关性、自信心和满意度？

A. Introduction to the Proposed Framework
A.拟议框架简介

Fig. 1 depicts the proposed HCLTF. The framework is visually articulated as a Venn diagram composed of three overlapping circles representing the domains of learning, teaching, and generative AI. The following section explains the three domains and the three areas where they intersect.
图 1 描述了拟议的 HCLTF。该框架被形象地表述为一个维恩图，由三个重叠的圆圈组成，分别代表学习、教学和生成式人工智能领域。下文将解释这三个领域以及它们相互交叉的三个方面。

1. Learning: The learning domain, guided by human-centric principles, positions students at the core of the educational experience. It accentuates the facilitation of the three stages of SRL. In this domain, the goal is to empower students to take ownership of their learning process through setting goals, monitoring learning processes, and reflecting on their learning outcomes.
   学习：学习领域遵循以人为本的原则，将学生定位为教育体验的核心。它强调促进自学学习的三个阶段。在这一领域，目标是通过设定目标、监控学习过程和反思学习成果，让学生掌握自己的学习过程。

2. Teaching: The teaching domain is dedicated to pedagogical approaches that seamlessly integrate generative AI into educational settings. This domain involves guiding teachers to integrate generative AI into their teaching strategies, ensuring that the technology is aligned with pedagogical goals and enhances content delivery. Within this framework, the ARCS model is used to assist teachers in contemplating strategies to enhance students’ motivation during the learning process.
   教学：教学领域致力于将生成式人工智能无缝融入教育环境的教学方法。这一领域涉及指导教师将人工智能生成技术融入教学策略，确保技术与教学目标相一致，并增强内容的传授。在这一框架内，ARCS 模型被用来帮助教师思考在学习过程中提高学生学习积极性的策略。

3. Generative AI: The generative AI domain is focused on providing technological affordances and immediate feedback. In this context, generative AI is a tool that can be used to offer individualized feedback to students.
   生成式人工智能：生成式人工智能领域侧重于提供技术能力和即时反馈。在这种情况下，生成式人工智能是一种可用于向学生提供个性化反馈的工具。

4. Learning–Teaching: At the nexus of the learning and teaching domains, the focus shifts to teachers’ guidance and facilitation roles. Teachers are instrumental in steering students through the SRL process by providing affective and social support to foster a supportive learning environment [41].
   学与教：在学与教领域的结合点上，重点转向教师的指导和促进作用。教师通过提供情感和社会支持，营造有利的学习环境，在引导学生完成自学学习过程中发挥着重要作用 [41] 。

5. Learning–Generative AI: At the intersection of the learning and generative AI domains, neuroscience-informed educational implications are used to enhance attention, engagement, error-feedback, and reflection and thus support students’ SRL [42]. Accordingly, students are encouraged to actively find and correct errors in the content generated by AI.
   学习-生成式人工智能：在学习和生成式人工智能领域的交叉点上，神经科学的教育意义被用来提高注意力、参与度、错误反馈和反思，从而支持学生的自学能力 [42] 。因此，我们鼓励学生在人工智能生成的内容中主动发现并纠正错误。

6. Central Zone—Learning, Teaching, and Generative AI: The central zone, where all three domains converge, symbolizes the unified essence of teaching and learning. This should be a synergistic space, where the framework domains are combined to create a cohesive and dynamic learning environment that supports SRL and prepares students for the demands of the 21st century.
   中心区--学习、教学和生成性人工智能：中心区是所有三个领域的交汇处，象征着教与学的统一本质。这应该是一个协同增效的空间，在这里，各框架领域相互结合，创造出一个具有凝聚力和活力的学习环境，以支持自学能力的培养，并使学生为满足 21 世纪的需求做好准备。

Fig. 1.  图 1.

HCLTF that uses generative AI in K–12 settings.
在 K-12 环境中使用生成式人工智能的 HCLTF。

Show All

B. Case Study: Generative AI Tools as Partners for Enhancing Primary Students’ Chinese Writing
B.案例研究：生成式人工智能工具作为提高小学生中文写作水平的合作伙伴

This case study demonstrates the application of the HCLTF for designing learning and teaching activities in using generative AI (e.g., ChatGPT) for supporting Grade 5 students’ Chinese writing in Hong Kong, in which students are guided by their teachers to partner with ChatGPT across a five-stage lesson plan comprising topic preparation, spoken to written language revision, structure refinement, word refinement, and reflection and review (see Fig. 2).
本案例研究展示了应用 HCLTF 设计学与教活动，使用生成式人工智能（如 ChatGPT）来支持香港五年级学生的中文写作，其中学生在教师的指导下与 ChatGPT 合作，完成了包括主题准备、口语到书面语修改、结构完善、词语完善和反思与回顾五个阶段的课程计划（见图 2 ）。

1. Stage 1—Topic preparation: In this stage, the teacher assigns essay topics and clearly explains the expectations. Students engage in goal-setting and planning their essays. In this stage, generative AI has not yet been introduced.
   第 1 阶段--题目准备：在这一阶段，教师布置作文题目，并明确说明对学生的要求。学生参与目标设定和作文规划。在这一阶段，尚未引入生成式人工智能。

2. Stage 2—Spoken to written language revision: The teacher facilitates discussions and guides students in the process of brainstorming ideas orally [43]. Students create drafts of their essays by verbally articulating their thoughts and converting them to text using tools, such as Google Docs, which support speech-to-text functionality. The students then refine their spoken language drafts to align with written language standards by removing colloquialisms and replacing them with more formal language and adding correct punctuation. Next, the students use ChatGPT to revise their spoken language drafts into a more formal written language (Version 1). They learn to interact with the AI interface through prompts, compare their work with the AI's output, and resolve the differences. A comparison table is created to contrast the student version with the generative AI version 1 (v1) and thus provide immediate feedback on the outcomes. Throughout this stage, teachers’ scaffolding provides crucial support.
   第二阶段--从口头到书面的语言修正： [43] 教师主持讨论，引导学生集思广益。学生通过口头表达自己的想法，并使用支持语音到文本功能的工具（如 Google Docs）将其转换为文本，从而创作出作文草稿。然后，学生通过删除口语化用语，代之以更正式的语言，并添加正确的标点符号，完善口语草稿，使其符合书面语言标准。接下来，学生使用 ChatGPT 将口语草稿修改为更正式的书面语言（版本 1）。他们学会通过提示与人工智能界面进行交互，将自己的作品与人工智能的输出进行比较，并解决其中的差异。创建一个对照表，将学生版本与生成式人工智能版本 1（v1）进行对比，从而对结果提供即时反馈。在整个阶段中，教师的支架提供了重要的支持。

3. Stage 3—Structure refinement: At the beginning of this stage, the teacher introduces structural knowledge and organizes further discussions to deepen the students’ understanding of essay structure. The students apply this knowledge to enhance the organization of their essays, considering elements, such as introductions, transitions, and conclusions. After an initial student-led revision, generative AI tools are used to further refine the structure (v2). The students are encouraged to critically and comparatively evaluate the structure proposed by ChatGPT against their own essay, and teacher-led discussions are used to consolidate the learning points from prior revisions.
   第三阶段--结构完善：在这一阶段的开始，教师介绍结构知识并组织进一步讨论，以加深学生对作文结构的理解。学生运用这些知识来加强文章的组织，考虑导语、过渡句和结论等要素。在学生主导的初步修改后，使用生成式人工智能工具进一步完善结构（v2）。我们鼓励学生根据自己的文章对 ChatGPT 提出的结构进行批判性的比较评估，并通过教师引导的讨论来巩固之前修改中的学习要点。

4. Stage 4—Word refinement: This stage focuses on linguistic precision. The teacher continues to provide guidance and to facilitate the refinement of language and word choice. The students focus on enhancing the phrases and terminology used in their essays. Subsequently, they can ask ChatGPT to integrate advanced rhetorical devices, such as metaphors or personification, into their writing. Finally, they compare their revisions with ChatGPT's suggestions (v3) to determine the most effective expression.
   第 4 阶段--词语精炼：这一阶段的重点是语言的准确性。教师继续提供指导，促进语言和用词的精炼。学生将重点放在加强作文中使用的短语和术语上。随后，他们可以要求 ChatGPT 将比喻或拟人等高级修辞手法融入他们的写作中。最后，他们将自己的修改意见与 ChatGPT 的建议（v3）进行比较，以确定最有效的表达方式。

5. Stage 5—Reflection and review: In this stage, the teacher facilitates a discussion to enable reflection, comments on the students’ essays (v4), highlights areas of strength, and suggests improvements. The students summarize their learning experiences and the knowledge and skills gained throughout the cycles of interaction with ChatGPT.
   第五阶段--反思与回顾：在这一阶段，教师主持讨论以促进反思，对学生的论文（v4）进行评论，强调优势领域并提出改进建议。学生总结他们的学习经验以及在与 ChatGPT 互动的整个周期中所获得的知识和技能。

Fig. 2.  图 2.

Five stages of primary grade five students’ Chinese writing underpinned by the proposed framework.
以拟议框架为基础的小学五年级学生中文写作的五个阶段。

Show All

In the proposed framework, generative AI is used as a tool to support and enhance the writing process, from initial topic preparation to the final reflection. In each stage, learners are encouraged to actively engage with domain knowledge. As a result, the learners continuously refine their understanding. This iterative process of learning and application is fundamental for the development of SRL competencies.
在拟议的框架中，生成式人工智能被用作支持和加强写作过程的工具，从最初的题目准备到最后的反思。在每个阶段，我们都鼓励学习者积极参与领域知识的学习。因此，学习者会不断完善自己的理解。这种学习和应用的迭代过程是培养自学能力的基础。

A. Participants A.与会者

The study recruited a cohort of 31 in-service primary school teachers. A purposeful sampling technique was used. During the recruitment process, the researchers contacted the principals of several local primary schools and extended invitations to teachers who expressed interest in participating in the teacher development programme.
研究招募了 31 名在职小学教师。研究采用了有目的的抽样技术。在招募过程中，研究人员与当地几所小学的校长取得了联系，并向表示有兴趣参加教师发展计划的教师发出了邀请。

The gender distribution of the participants was nearly equal, with 16 female and 15 male teachers. Regarding educational attainment, most of the participants held a bachelor's degree (n = 21); the others possessed a master's degree (n = 10).
参与者的性别分布基本均衡，其中女教师 16 人，男教师 15 人。在学历方面，大多数参与者拥有学士学位（21 人），其他参与者拥有硕士学位（10 人）。

The participating teachers were responsible for teaching two or more subjects. Fig. 3 presents the subject distribution among the participants: computer science was most commonly taught, followed by general studies, mathematics, and languages.
参赛教师负责教授两个或两个以上的学科。图 3 显示了参与者的学科分布情况：教授计算机科学的最多，其次是通识教育、数学和语文。

Fig. 3.  图 3.

Distribution of subjects taught by teachers.
教师教授科目的分布情况。

Show All

B. Procedure B.程序

The teachers were invited to participate in a six-week programme that included: 1) a 30-h course intended to a foundational understanding of AI (e.g., what AI is, the five steps of machine learning, supervised learning, and unsupervised learning) and deep learning (e.g., data cleaning, data augmentation, neural networks, computer vision, convolution neural networks, and recurrent neural networks), and generative AI concepts and 2) a 30-h course on integrating AI into course design in areas, such as Chinese language, English language, mathematics, general studies, computer science, and music education.
教师们应邀参加了为期六周的课程，其中包括1) 30 小时的课程，旨在对人工智能（例如，什么是人工智能、机器学习的五个步骤、监督学习和无监督学习）和深度学习（例如，数据清理、数据增强、神经网络、计算机视觉、卷积神经网络和递归神经网络）以及生成式人工智能概念进行基础性了解；以及 2) 30 小时的课程，旨在将人工智能融入中文、英文、数学、通识教育、计算机科学和音乐教育等领域的课程设计中。

Throughout the programme, the teachers completed a series of assessments to gauge their progress and the efficacy of training. These assessments included pr-etest and post-test to measure their understanding of AI concepts and evaluate their knowledge of TPACK before and after the course and their self-perceived competency in designing courses, specifically in terms of increasing students’ attention, relevance, confidence, and satisfaction.
在整个课程期间，教师们完成了一系列评估，以衡量他们的进步和培训效果。这些评估包括前测和后测，以衡量他们对人工智能概念的理解，评估他们在课程前后对 TPACK 的了解，以及他们在设计课程方面的自我认知能力，特别是在提高学生的注意力、相关性、自信心和满意度方面。

In the final week of the programme, the teachers were required to write a self-reflective piece on their experiences and the insights gained from learning about the use of generative AI tools in education.
在课程的最后一周，教师们需要撰写一篇自我反思文章，介绍他们从学习在教育中使用生成式人工智能工具中获得的经验和启示。

C. Data Collection and Analysis
C.数据收集与分析

A. Effect of the Teacher Development Programme on Concept Test Scores
A.教师发展计划对概念测试成绩的影响

The mean differences between the preconcept and postconcept tests. The assumption of normality was not violated (Shapiro–Wilk test, p > 0.05). The pretest score (M = 3.581, SD = 1.205) was lower than the posttest (M = 4.871, SD = 1.648; paired-samples t-test), and this difference was significant, t(30) = 3.616, p < 0.001, 95% confidence interval (CI) [0.562, 2.019]. The results show that the teacher development programme had a positive effect on the teachers’ understanding of the AI concepts.
概念前和概念后测试的平均差异。未违反正态性假设（Shapiro-Wilk 检验，P > 0.05）。前测得分（M = 3.581，SD = 1.205）低于后测得分（M = 4.871，SD = 1.648；配对样本 t 检验），且差异显著，t(30) = 3.616，p < 0.001，95% 置信区间 (CI) [0.562, 2.019]。结果表明，教师发展计划对教师理解人工智能概念产生了积极影响。

B. Effect of the Teacher Development Programme on TPACK Evaluation Scores
B.教师发展计划对 TPACK 评估得分的影响

Table I shows the descriptive statistics of teachers' perceptions of TPACK. A Wilcoxon signed-rank test showed that the teacher development programme led to statistically significant increases in the teachers’ CK (Z = –4.562, p < 0.001), PCK (Z = –4.632, p < 0.001), TCK (Z = –4.396, p < 0.001), and TPACK (Z = –4.561, p < 0.001).
表 I 显示了教师对专题知识包看法的描述性统计。Wilcoxon 符号秩检验表明，教师发展计划使教师的 CK（Z = -4.562，p < 0.001）、PCK（Z = -4.632，p < 0.001）、TCK（Z = -4.396，p < 0.001）和 TPACK（Z = -4.561，p < 0.001）都有了显著提高。

These results reveal that the teacher development programme was highly effective in enhancing teachers’ knowledge and skills across the CK, PCK, TCK, and TPACK domains. The statistically significant improvements in the survey scores suggest that the programme increased not only the teachers’ theoretical understanding of the integration of technology with pedagogy and content but also their ability to apply this knowledge in practice.
这些结果表明，教师发展计划在提高教师的知识和技能（CK、PCK、TCK 和 TPACK）方面非常有效。调查得分在统计上的显著提高表明，该计划不仅提高了教师对技术与教学法和内容整合的理论理解，还提高了他们在实践中应用这些知识的能力。

To triangulate the quantitative data, the teachers’ self-reflective writings were analyzed, and some examples were selected.
为了对定量数据进行三角测量，我们对教师的自我反思文章进行了分析，并选取了一些例子。

1. CK

2. The course provides information on AI, which can be passed on to students, and emphasises the importance of understanding AI in education. (T1)
   该课程提供有关人工智能的信息，可将这些信息传递给学生，并强调在教育中了解人工智能的重要性。(T1)

3. This course has really opened my eyes to the role generative AI can play in our classrooms. It is not just another buzzword; it is a tool we can use to really make a difference in how we teach and how our students learn. I will pass on what I have learnt about AI to my students and help them understand its growing importance in our world. (T17)
   这门课程让我真正看到了生成式人工智能在课堂上可以发挥的作用。它不仅仅是另一个流行词，而是我们可以用来真正改变我们的教学方式和学生的学习方式的工具。我将把我所学到的人工智能知识传授给我的学生，帮助他们了解人工智能在我们这个世界上日益增长的重要性。(T17)

2. PCK

5. There is a need to deepen teachers’ understanding of AI to facilitate the birth and transformation of pedagogy. I learnt a lot from this course. (T23)
   有必要加深教师对人工智能的理解，以促进教学法的诞生和变革。我从这门课程中学到了很多。(T23)

6. After taking this programme, I see the power AI has to reshape [how] we teach. I have gained so much insight into how AI can help tailor our teaching methods to each unique student. The framework we have been introduced to is very practical. It underscores the importance of keeping our students at the heart of their learning journey. By harnessing the power of AI, we can create a learning environment that not only recognises but also celebrates each student's individual needs and potential. (T16)
   参加这个课程后，我看到了人工智能重塑我们教学方式的力量。我对人工智能如何帮助我们针对每个独特的学生量身定制教学方法有了更深入的了解。向我们介绍的框架非常实用。它强调了让学生在学习过程中处于核心地位的重要性。通过利用人工智能的力量，我们可以创造一个不仅承认而且赞美每个学生的个人需求和潜力的学习环境。(T16)

3. TCK

1. Words like “tokens” and “transformers” were foreign to me, and I was worried that machines might take our jobs. But this course showed me that is not the case. (T14)
   代币 "和 "变压器 "这样的词对我来说很陌生，我担心机器会抢走我们的工作。但这门课程告诉我，事实并非如此。(T14)

4. TPACK

2. Integrating AI into our teaching is not just about understanding the technology or the content. I have learnt that using AI tools effectively requires a careful balance. This course has given me the confidence to strategically use AI in my lesson plans. (T29)
   将人工智能融入我们的教学，不仅仅是要了解技术或内容。我了解到，有效使用人工智能工具需要谨慎平衡。这门课程让我有信心在教学计划中战略性地使用人工智能。(T29)

3. AI can give me insights into where a student might be struggling, allowing me to intervene with the right kind of support at the right time. The TPACK framework reminds me that technology is a tool to facilitate this journey, not the journey itself. My role is to use that tool to build a bridge between students’ current abilities and their potential, guiding them towards becoming lifelong, self-motivated learners. (T15)
   人工智能可以让我深入了解学生可能在哪些方面遇到困难，让我能够在适当的时候提供适当的支持。TPACK 框架提醒我，技术是促进这一旅程的工具，而不是旅程本身。我的职责是利用这一工具，在学生的现有能力和潜能之间架起一座桥梁，引导他们成为终身的、自我激励的学习者。(T15)

C. Effect of the Teacher Development Programme on Perceived Ability to Teach With Generative AI Under the ARCS
C.教师发展计划对在 ARCS 下使用生成式人工智能教学的认知能力的影响

Statistically significant mean differences in attention, relevance, confidence, and satisfaction between the presurvey and postsurvey were assessed (paired samples t-test). The assumption of normality was not violated (Shapiro–Wilk's test, p > 0.05). Table II tabulates the descriptive data on teachers’ perceived ability to use generative AI to design courses that can increase students’ attention, relevance, confidence, and satisfaction.
对调查前和调查后在注意力、相关性、信心和满意度方面的平均差异进行了统计评估（配对样本 t 检验）。没有违反正态性假设（Shapiro-Wilk 检验，p > 0.05）。表 II 列出了教师对使用生成式人工智能设计课程以提高学生注意力、相关性、自信心和满意度的感知能力的描述性数据。

TABLE I Descriptive Statistics of Teachers’ Perceptions of TPACK
表 I 教师对专业知识包看法的描述性统计

TABLE II Descriptive Statistics of Teachers’ Perceived Ability to Design Courses Under ARCS
表 II 教师对 ARCS 系统下课程设计能力的描述性统计表

For attention, a statistically significant increase was observed from the presurvey (M = 3.151, SD = 0.950) to the postsurvey (M = 4.366, SD = 0.598), t(df) = 7.762, p < 0.001. The effect size was large (Cohen's d = 0.87). For relevance, the participants reported a significant improvement from the presurvey (M = 3.000, SD = 0.878) to post-survey (M = 4.226, SD = 0.669), t(df) = 8.161, p < 0.001. The effect size was large (Cohen's d = 0.84). For confidence, there was a significant increase from the presurvey (M = 3.108, SD = 0.900) to the post-survey (M = 4.333, SD = 0.644), t(df) = 7.652, p < 0.001. The effect size was large (Cohen's d = 0.89). For satisfaction, there was a significant increase from the presurvey (M = 3.290, SD = 0.811) to the post-survey (M = 4.366, SD = 0.663), t(df) = 6.966, p < 0.001. The effect size was large (Cohen's d = 0.86).
在注意力方面，从调查前（M = 3.151，SD = 0.950）到调查后（M = 4.366，SD = 0.598），观察到了统计学上的显著增长，t(df) = 7.762，p < 0.001。效应大小很大（Cohen's d = 0.87）。在相关性方面，参与者报告说，从调查前（M = 3.000，SD = 0.878）到调查后（M = 4.226，SD = 0.669），他们的相关性有了显著提高，t（df）= 8.161，p < 0.001。效应大小较大（Cohen's d = 0.84）。在自信心方面，从调查前（M = 3.108，SD = 0.900）到调查后（M = 4.333，SD = 0.644），自信心明显增加，t（df）= 7.652，p < 0.001。效应大小较大（Cohen's d = 0.89）。在满意度方面，从调查前（M = 3.290，SD = 0.811）到调查后（M = 4.366，SD = 0.663），满意度有了显著提高，t（df）= 6.966，p < 0.001。效应大小较大（Cohen's d = 0.86）。

To add depth to the quantitative data, the teachers’ self-reflective narratives were carefully reviewed.
为了增加定量数据的深度，我们仔细审查了教师的自我反思叙述。

1. Attention 请注意

2. The training transformed my lesson kick-offs. I now use targeted questions that tap into students’ curiosity, leading to a noticeable boost in their participation from the start. (T16)
   培训改变了我的开课方式。现在，我使用有针对性的问题来激发学生的好奇心，使他们从一开始就明显提高了参与度。(T16)

3. After completing the programme, I have noticed a marked change in the way I approach the start of my lessons. This shift became evident when I observed my students more eagerly participating in discussions and activities right from the beginning of class. Generative AI can give me many amazing ideas. (T24)
   课程结束后，我发现自己上课的方式发生了明显的变化。当我观察到我的学生从课堂一开始就更热衷于参与讨论和活动时，这种转变就显而易见了。生成式人工智能能给我带来很多奇妙的想法。(T24)

2. Relevance 相关性

1. I am always striving to make my lessons relevant, but the programme gave me a new perspective on how to integrate real-world problems into my curriculum. By using AI to brainstorm real-time scenarios, I am able to bring context to theoretical concepts. (T29)
   我一直在努力使我的课程具有相关性，但该计划给了我一个新的视角，让我知道如何将现实世界的问题融入我的课程。通过使用人工智能对实时场景进行头脑风暴，我能够为理论概念提供背景。(T29)

3. Confidence 信心

1. After the first day of this programme, I encouraged students to converse with generative AI and critically evaluate its responses. They were developing critical thinking skills by identifying areas where AI's responses can be improved. (T8)
   第一天课程结束后，我鼓励学生与生成式人工智能对话，并对其反应进行批判性评估。他们通过找出人工智能反应可以改进的地方，培养了批判性思维能力。(T8)

4. Satisfaction 满意度

1. Integrating generative AI into my daily teaching practices has brought a new level of joy as I watch students take ownership of their learning. The changes I have implemented in the course, inspired by what I've learnt from this teacher development programme, have been met with enthusiastic feedback from students, affirming the value of these innovative tools in enhancing their educational experience. (T23)
   将生成式人工智能融入我的日常教学实践，给我带来了新的快乐，因为我看到学生们掌握了学习的主动权。受教师发展项目的启发，我在课程中实施了一些变革，并得到了学生们的热烈反馈，他们肯定了这些创新工具在提升教育体验方面的价值。(T23)