以人工智能为代表的新技术已呈现出变革教育系统的巨大潜力，引发各界对未来教育变革的深思。文献分析表明，个性化培养是未来教育教学变革的趋势和目标，也是应对我国教育均衡发展中挑战的关键策略之一。作者在博士后期间围绕个性化培养完成了一系列科研项目工作和学术研究，本文是对这些工作的总结。首先，报告了在《人工智能与未来教育发展研究》项目中作者作为核心骨干的研究团队对未来教育特征的讨论，指出个性化培养是未来教育变革的趋势和目标。然后，结合文献回顾了作为个性化培养基础的个性化学习的内涵和工作机理，以及人工智能促进个性化培养的潜力。接着，在个性化学习的框架下，以学习行为分析和教学互动反馈策略作为切入点，通过两个小型实证研究（研究1和研究2），结合数据分析结果讨论了研究对人工智能支撑的个性化培养的若干启示：（1）研究1基于在线学习活动的细粒度日志数据，通过综合运用多种学习分析技术（滞后序列分析、 时间日志数据分析等）来识别关键的学习行为特征，发现不同性别的学习者在学习过程中随着时间变化的行为模式特点和行为事件转换模式特点，识别了不同性别的学习者在特定类型的学习活动中呈现出来的优势和不足。据此，为个性化教学干预的设计提供依据，从而为有针对性的进行个性化培养提供支持；（2）研究2挖掘了教师在在线学习中教学反馈策略应用的数据，使用多案例研究法从一线教师的反馈实施实践中挖掘出11种教学反馈策略，还考察了教师在不同复杂度的学习任务中使用这些反馈的模式，以及分析了教师所用的反馈策略在有效性方面的特征。不同类型的教学反馈是实行个性化反馈的基本材料和内容，是被个性化处理的对象类型之一。针对不同特征类型的学习者，在特定的学习情境（如不同特征的学习任务和不同的学习表现）下，依照特定的个性化规则给学生提供个性化的教学反馈。按照这种个性化机制，借力AI技术，可以从多个维度助力实施个性化培养。

对于提出的个性化培养策略，即使可以通过实验室实验或者小规模教学实验的方法验证其性能，还需要大规模的将个性化培养应用到教育教学实践的一线，考察其最终效果，这就需要经过特定的实施路径，即教育社会实验研究。实施人工智能支撑的个性化培养教育社会实验，目的在于检验人工智能支撑的个性化培养措施在一线的、真实的、动态进展的教育教学社会进程中对学生学习、教师教学、教学环境等带来的综合性社会影响。本文报告了对教育社会实验持续进行的多阶段探索。首先提出了教育社会实验研究的概念框架、主要组成要素与过程模型（研究3）。然后以智能导学系统（ITS）的社会实验为例，通过系统化文献综述（研究4），总结了实际操作中的社会实验研究采用的实验设计方法，数据类型和分析技术。综述ITS教育社会实验结果显示，教育社会实验具有高复杂性、大规模、长周期的特点。社会实验往往考察实验干预手段对多类主体、多个维度的影响；此外，不同研究反应的分析结果还具有不一致性，有效结果的出现往往具有一定的前提条件等。教育社会实验与通常的实验室实验以及小规模的教学实验研究不同，必须要了解并考虑社会实验实施环境的特征与影响，考虑实验参与人员的多样特征。实验环境中多种因素往往交互作用并对实验结果产生复杂影响。因此在选择参与对象，设计实验，以及分析数据时必须考虑这些因素并采取措施处理这些复杂状况。研究还总结了社会实验中常见的挑战和影响社会实验成功实施的关键因素。这些发现为进行教育社会实验指明了工作重点，为后续相关研究提供了可能的方向。

人工智能支撑的个性化培养是推动未来教育发展的重要策略，也是实现教育公平发展的有效手段。同时，这也是一个高度复杂的系统性工程。在博士后既定的时间框下，本研究仅仅呈现了对这个宏大研究蓝图上较为细微的切入点的探索。因此，本研究在深度和广度上还存在局限性。后续可以沿着学习行为分析和个性化教学策略的研究方向，结合这些结果进行个性化学习的小规模实验研究。然后通过教育社会实验研究的方法，以来自一线的综合证据，检验所提议的个性化培养策略，检验其在教育教学中产生的复杂影响。然后在人工智能技术的支持下，更有信心的把个性化培养大规模推广到教育教学的真实环境中进行应用，助力教育教学变革。

The new technologies represented by artificial intelligence presents a huge potential of transforming educational systems, motivating people to imagine the futures of education. The analysis of literature about the characteristics of future education shows that personalized learning is the trend and goal of educational transformations, and it is also one of the key strategies to cope the challenges emerged in the balanced development of education in China. In this context, I have completed a series of research projects and studies related to personalized learning during my post-doctoral journey. This report is a summary of the major outcomes. First, I reported the findings and discussions based on my work related to the project “Artificial intelligence and the futures of education, which indicated that personalized learning is the goal and direction of educational transformation. Then, I described the concept of personalized learning and the working mechanism of using artificial intelligence to support personalized learning based on literature. Under the framework of personalized learning, using learning analytics and instructional feedback strategies as start points and examples, I conducted two empirical studies (study 1 and study 2). The implications for implementing personalized learning supported by artificial intelligence is discussed. For study 1, based on fine-grained log data, I used learning analytics techniques and identified key patterns and features in learning behaviors, such as the behavior patterns and behavioral transition patterns of learners with different genders in the learning process, and discovered the advantages and disadvantages of different learners in the learning process. This finding serves as a guide to enable personalized learning, which is the basis for the design of personalized learning interventions. Study 2 focuses on instructional feedback provided by teachers in online contexts. 11 types of instructional feedback, and the patterns of using feedbacks for different types of learning tasks were identified. The characteristics of the effectiveness of feedback were also considered and discussed. Different types of instructional feedback are necessary resources for personalized feedback and they are the objects to be personalized. One way to achieve personalized learning is to provide personalized instructional feedback to different types of learners in specific learning situations.

In addition, in order to apply personalized learning to the real context of education and teaching on a large scale, this article proposes a specific approach: Educational Social Experiment Study. To conduct education social experimental study on artificial intelligence-supported personalized learning, we need to investigate the overall effectiveness of artificial intelligence-supported personalized learning in the real and dynamic social procedure of education, teaching and learning. Based on the consideration of the complexity and diversity in education and teaching practice, a process model for conducting educational social experimental research was proposed. Based on a systematic literature review, I summarized the experimental design methods, data types and analysis techniques used in practical social experimental research focused on Intelligent tutoring system (ITS). The analysis of the results of the educational social experiment indicated that the researchers of educational social experiment need to consider the characteristics and influence of the social environment where the study is conducted, the diversity in the participants involved in the experiments, and the influence of various factors on the experimental results, which make the results complicated. I also summarized the common challenges emerged in social experiments and the mentioned key factors affecting the successful implementation. Based on these findings, I suggested the critical part of the work in conducting social experiments in education, and provided possible directions for subsequent research.

Considering the complexity of the theme personalized learning supported by artificial intelligence, this research only explores the several aspects of the big research blueprint. Thus, there are still limitations in the depth and width of this report. The follow-up work should combine the study of learning behavior analysis and instructional feedback strategies and conduct empirical research on personalized learning. Following the approach of educational social experimental research, large-scale application and studies of personalized learning with the support of artificial intelligence technology in education and teaching practice in the real educational context should be conducted.