随着人工智能在各社会领域的广泛普及与应用，教育人工智能为教育活动带来便利的同时也引发新的现实问题，亟需立法规范。具有独特生成性和类人属性的教育人工智能，在教育活动中不再仅仅作为工具发挥作用，而展现出某种特殊的主体性，使得在教育关系及其他社会关系中出现了新型“人-技术”关系，对现行立法提出了挑战，要求立法适应社会现实，对新的社会关系进行调整。尽管这一问题已经引发我国与世界各国关注，但理论研究仍不够深入。本研究以动态道义逻辑理论为基础，聚焦教育人工智能在社会活动和教育活动中展现出的特殊的主体性为立法带来的新挑战，综合运用文献研究法、文本研究与法比较研究等方法，探究教育人工智能法律规范内在逻辑，以及如何基于这种逻辑构建教育仍智能法律规范。

基于动态道义逻辑，本研究确定价值”“主体”“客体”“断定”为四个基本要素，据此构建出教育人工智能法律规范的逻辑模型。由此形成了本研究的论证思路，逐步回应了以下问题：第一，探究依据怎样的原则构建规范才能解决教育人工智能为立法带来的新问题，立法的价值依据是什么，价值依据的内在来源是什么，价值间的关系是怎样的，如何平衡价值间的关系并作出价值选择。第二，教育人工智能法律规范的法律主体是谁，教育人工智能的主体性是否使其成为法律主体，教育人工智能的主体性对人作为法律主体带来哪些影响；第三，教育人工智能法律规范的法律客体是谁，对客体进行规范的依据是什么，应当如何对客体进行规范；第四，在教育人工智能法律规范中，需要被规范的行为有哪些，应当如何对这些行为进行断定，也就是依据立法价值应当对法律行为作出什么样的判断，对行为的应为、可为、禁为的判断构成了什么样的权利义务关系。

从结论上看，本研究主要包括以下观点：

首先，在教育人工智能法律规范中，立法价值应解决教育人工智能为现行立法带来的独特问题，也就是回应教育人工智能特殊的主体性如何协调人的主体性、在法律规范中得到表达的问题。第一，法律作为人类社会生活的规范，必须以人作为主体的需求为先，教育人工智能的主体性必须融入人的主体性进行表达。第二，教育人工智能主体性向人之主体性的融入在立法价值上具体表现为教育人工智能的价值应当融入教育价值与法的价值进行表达。其中，教育的价值取向要求在主体性实现的基础上追求超主体性的实现，法律追求的价值则寻求合规律性与合目的性的统一。但是，在不同价值的相互交融中，价值间的冲突也显现出来。第三，面对价值间的冲突，静态的价值选择无法成为解决方案，必须在价值的动态关系中，将价值的冲突转向价值的一致，在价值的动态运作过程中寻找价值间的一致与平衡关系。无论是教育价值还是法的价值，都在主客观之间、主体性与超主体性之间寻求平衡与统一，这也为教育人工智能多元互动的立法价值取得了一种平衡思路，即在主体性与超主体性的辩证统一中调和互动的多元价值。

其次，教育人工智能的主体性是社会活动和教育活动中由于自身特性产生的主体性，并非法律层面上的主体性，因此并未影响到法律主体的类型，但对人作为法律主体的内涵和属性产生影响，换言之，教育人工智能的主体性并不使其成为法律主体，法律的主体仍然是人，但是对人作为法律主体的概念理解扩大了。第一，教育人工智能不是法律主体，教育人工智能立法的法律主体是教育人工智能的研发者、供应者、使用者、监管者，具体包括教育人工智能的工程师/设计师，教育人工智能的生产与销售企业，学校及其他教育机构、教师、学生及其监护人、其他公民及法人组织、政府机关等。第二，尽管对人的法律主体地位没有影响，教育人工智能的主体性却在融入人的主体性之时，冲击了人的固有边界，增添了人的技术属性，促使人的概念向后人类扩展。

再次，教育人工智能不是法律主体，不意味着它不需要法律规范，相反，作为法律规范的客体，教育人工智能应当符合特定的标准，满足人作为主体的价值需求。构建具体标准应当以立法价值为指导，也就是基于教育人工智能的主体性应受限于人的主体性的原则确定教育人工智能的标准，围绕着人作为主体、在教育活动中尤其表现为围绕着受教育者为主体展开。第一，教育人工智能应符合教育性标准，指向教育目的且符合教育规律，这要求教育人工智能产品应有设计其教育目标及其实现方案的技术脚本与使用计划。第二，教育人工智能应符合鲁棒性与准确性标准，以保障技术能够稳定、准确地实现教育目标。这要求教育人工智能产品应配备专门的研发与评估团队，负责保障技术能力等级的持续提升。第三，教育人工智能应符合道德性与安全性标准，以保障技术在实现教育目标时不对人的其他价值需求造成负面影响，这要求教育人工智能产品应包含可迭代的风险管理系统，对漏洞和风险进行技术自查从而及早干预。第四，教育人工智能应符合可干预性标准，以保障当技术不能实现教育目标时有挽救的渠道，这要求教育人工智能产品中应留存可供工程师监督的后台路径，增加人类监督。第五，教育人工智能应符合透明性标准，以保障当技术造成了侵权后果后时有追责的途径，这要求教育人工智能产品应留存运行记录以便需要时进行追溯、核查。

最后，在立法价值的指导下，提供、部署、监管环节的权利义务关系和权力责任关系是教育人工智能法律规范的重点。规范相关主体的行为并不单纯意味着限制人工智能发展，而是在以保障受教育权为核心的前提下，通过规约教育人工智能和促进教育人工智能为教育服务两种手段进行立法规范。依据立法价值，教育人工智能法律规范必须将教育人工智能的主体性依托于人的主体性、将教育人工智能的价值融入教育价值和法的价值，即，对行为的规范应围绕人，尤其围绕教育中受教育者的受教育权展开。第一，私法领域应重点规范提供者、部署者的权利义务关系，要求教育人工智能产品的提供者提供适合受教育者使用的产品，要求教育人工智能产品的运营商提供能够保障受教育权不受侵犯的运营服务，要求出于职业身份使用、部署教育人工智能产品的公民与法人组织以教育目的为指导开展对教育人工智能的使用与部署；第二，公法领域应重点规范监管者的权力责任关系，要求权力主体依法开展对各相关主体履行义务情况的监督，在行使权力的同时通过规约自身和提供保障的方式承担责任，规范自身对权力的运用限制于职权范围内，对教育人工智能产品的各类应用行为提供创新政策支持，并通过构建保障制度的方式履行责任。

With the widespread popularization and application of artificial intelligence in various social fields, educational artificial intelligence not only brings convenience to educational activities, but also causes new practical problems, which need to be regulated by legislation. As an educational artificial intelligence with unique generativity and human-like attributes, it no longer plays a role as a tool in educational activities, but shows some special subjectivity, which makes a new "human-technology" relationship appear in educational relations and other social relations, posing challenges to the existing legislation, requiring legislation to adapt to the social reality and adjust the new social relations. Although this problem has aroused the attention of our country and other countries, the theoretical research is still not deep enough. Based on the theory of dynamic deontic logic, this study focus on the new challenges brought by the special subjectivity of educational artificial intelligence in social and educational activities for legislation. It comprehensively uses methods such as literature research, text research, and comparative research to explore the internal logic of legal norms for educational artificial intelligence, and how to construct intelligent legal norms for education based on this logic.

Based on dynamic deontic logic, this study identifies “value”, “subject”, “object”, and “assertion” as four basic elements, and constructs a logical model for legal norms of educational artificial intelligence based on them. This formed the argumentation approach of this study, gradually responding to the following questions: First, to explore how to build norms according to the principles to solve the new problems brought by educational artificial intelligence legislation, what is the value basis of legislation, what is the internal source of the value basis, what is the relationship between values, how to balance the relationship between values and make value choices; Second, who is the subject of regulation, whether the subjectivity of educational artificial intelligence makes it a legal subject, and what impact the subjectivity of educational artificial intelligence has on human being as a legal subject; Third, who is the object of the norm, what is the basis of the norm object, and how to regulate the object; Fourth, what are the behaviors that need to be regulated, how should we judge these behaviors, that is, what kind of judgment should be made on the legal behavior according to the legislative value, and what kind of rights and obligations are constituted by the judgment on the should, can and forbidden of the behavior.

This study mainly includes the following viewpoints:

First of all, in the law of educational artificial intelligence, the legislative value should solve the unique problems brought by educational artificial intelligence to the current legislation, that is, respond to the special subjectivity of educational artificial intelligence how to coordinate human subjectivity and be expressed in the legal norms. First, the law, as the norm of human social life, must take the needs of people as the subject as the first priority, and the subjectivity of educational artificial intelligence must be integrated into the subjectivity of people to express. Secondly, the integration of the subjectivity of educational artificial intelligence into the subjectivity of human beings in the legislative value is embodied in that the value of educational artificial intelligence should be expressed in the value of education and law. Among them, the value orientation of education requires the realization of super-subjectivity on the basis of the realization of subjectivity, while the value pursued by law seeks the unity of regularity and purpose. However, in the intermingling of different values, the conflict between values also appears. Third, in the face of the conflict between values, static value choice cannot be a solution, we must turn the conflict of values to the consistency of values in the dynamic relationship of values, and find the consistency and balance between values in the process of dynamic operation of values. Both the value of education and the value of law seek balance and unity between subjectivity and objectivity, between subjectivity and super-subjectivity, which also makes a balanced idea for the legislative value of multiple interaction of educational artificial intelligence, that is, to harmonize the multiple value of interaction in the dialectical unity of subjectivity and super-subjectivity.

Secondly, the subjectivity of educational artificial intelligence does not affect the type of legal subject, but affects the connotation of legal subject. In other words, the subjectivity of educational artificial intelligence does not make it a legal subject, and the subject of law is still human, but the concept of human as a legal subject is expanded. First, educational AI is not a legal subject, and the legal subject of educational AI legislation is to educate AI developers, suppliers, users and regulators, including educational AI engineers/designers, educational AI production and sales enterprises, schools and other educational institutions, teachers, students and their guardians, other citizens and legal entities, government agencies, etc. Second, although it has no impact on the legal subject status of human beings, the subjectivity of educational artificial intelligence shocks the inherent boundary of human beings, adds the technical attributes of human beings, and promotes the extension of the concept of human beings after human beings.

Thirdly, educational artificial intelligence is not the subject of law, does not mean that it does not need regulation, on the contrary, as the object of educational artificial intelligence law, educational artificial intelligence should meet specific standards and meet the value needs of people as the subject. The construction of specific standards should be guided by the legislative value, that is, the standards of educational artificial intelligence should be determined based on the principle that the subjectivity of educational artificial intelligence should be limited to human subjectivity. Therefore, the development of educational artificial intelligence standards should be centered around the human as the subject, especially in educational activities around the educatees as the subject. First, educational artificial intelligence should conform to educational standards, point to educational purposes and conform to educational laws, which requires educational artificial intelligence products to design their educational goals and implementation of the technical script and use plan. Second, educational AI should meet the standards of robustness and accuracy to ensure that the technology can achieve educational goals stably and accurately. This requires that educational AI products should be equipped with a dedicated R & D and evaluation team responsible for ensuring the continuous improvement of the technical capability level. Third, educational AI should meet ethical and safety standards to ensure that technology does not negatively affect other value needs of people when achieving educational goals, which requires educational AI products to contain an iterative risk management system, technical self-examination of vulnerabilities and risks, and early intervention. Fourth, educational artificial intelligence should meet the intervention standard to ensure that when the technology fails to achieve the educational goal, there is a channel to save, which requires that educational artificial intelligence products should retain a background path for engineers to supervise, and increase human supervision. Fifth, educational AI should meet transparency standards to ensure that when the technology causes infringement consequences, there is a way to be held accountable, which requires that educational AI products should keep operational records for traceability and verification when necessary.

Last but not least, under the guidance of legislative values, the rights, obligations, and responsibilities of providing, deploying, and regulating links are the focus of legal norms for artificial intelligence education. Standardizing the behavior of relevant entities does not simply mean restricting the development of artificial intelligence, but rather legislating and regulating educational artificial intelligence through two means: regulating educational artificial intelligence and promoting educational artificial intelligence to serve education, with the core of safeguarding the right to education. According to the legislative value, the law of educational artificial intelligence must rely on the subjectivity of educational artificial intelligence and integrate the value of educational artificial intelligence into the value of education and law, that is, the regulation of behavior should revolve around people, especially around the right to education of the educated in education. First, the field of private law should focus on regulating the rights and obligations of providers and deployers, requiring providers of educational artificial intelligence products to provide products suitable for use by educatees, and requiring operators of educational artificial intelligence products to provide operational services that can protect the right to education from infringement. Requiring citizens and legal organizations that use and deploy educational AI products for professional purposes to guide the use and deployment of educational AI;Second，the field of public law should focus on regulating the relationship between the power and responsibility of regulators, requiring the subjects exercising power to assume responsibility by regulating themselves and providing protection, regulating the use of power within the scope of their authority, providing innovative policy support for various application behaviors of educational artificial intelligence products, and fulfilling their responsibilities by building a guarantee system.