The Conditions for the Application of Artificial Intelligence in Ideological and Political Education

In recent years, the discipline of ideological and political education has paid close attention to the development of the artificial intelligence industry, hoping to apply artificial intelligence to combine the traditional advantages of ideological and political education with information technology. However, existing research often only relies on the technical characteristics of artificial intelligence to infer the ways it can empower ideological and political education, without fully considering the conditions for the application of artificial intelligence in ideological and political education. As a result, much research merely theoretically envisions the possibility of applying artificial intelligence, neglecting the practical constraints on the “landing” of artificial intelligence in the field of ideological and political education. Therefore, this article first looks ahead to the prospects of artificial intelligence empowering ideological and political education from the technical principles of artificial intelligence, then discusses how to create conditions for the application of artificial intelligence in ideological and political education starting from the core elements of artificial intelligence such as data, computing power, and algorithms, and finally improves the technical application norms that conform to the characteristics of both ideological and political education and artificial intelligence. Only in this way can discussions on the prospects of applying artificial intelligence in ideological and political education be both imaginative and realistic.

近年来，思想政治教育学科十分关注人工智能行业的发展，希望应用人工智能促成思想政治教育传统优势和信息技术的结合。然而，现有研究往往只依托人工智能的技术特性来推断其赋能思想政治教育的方式，尚未充分考虑到思想政治教育应用人工智能的实现条件，导致很多研究都是在理论上设想应用人工智能的可能性，忽视了人工智能在思想政治教育领域“落地”的现实制约因素。因此，本文首先从人工智能的技术原理出发展望它赋能思想政治教育的前景，进而讨论如何从数据、算力、算法等人工智能核心要素出发为思想政治教育应用人工智能创造条件，并完善符合思想政治教育和人工智能特性的技术应用规范，这样对思想政治教育应用人工智能前景的讨论才既具有想象力又具有现实性。

I. Multiple Scenarios for the Application of Artificial Intelligence in Ideological and Political Education

一、思想政治教育应用人工智能的多种场景

When discussing intelligence, people commonly think of products such as ChatGPT and Sora developed by OpenAI, “Wenxiaoyan (Wenxin Yiyan)” developed by Baidu, “Doubao” developed by ByteDance, and DeepSeek developed by Hangzhou DeepSeek Company. These products are called generative artificial intelligence. They generally rely on deep learning technology, autonomously learning the statistical laws of word-to-word connections from massive data, thereby acquiring the ability to generate natural language. Data, algorithms, and computing power are the basic elements supporting this type of artificial intelligence. According to the classification of artificial intelligence products, the current mainstream artificial intelligence belongs to the category of specialized, weak artificial intelligence: “specialized” means that this artificial intelligence is always trained for a specific purpose, mainly used for natural language processing; “weak” means that although this artificial intelligence can learn in an unsupervised state, it does not possess self-awareness, can only complete tasks assigned by humans, cannot autonomously propose tasks, and cannot refuse instructions issued by humans. When we currently say that ideological and political education applies artificial intelligence, we refer to the application of this specialized, weak artificial intelligence. Because it integrates various existing resources on the internet (including some research results in ideological and political education), it can empower ideological and political education and is expected to promote the improvement of theoretical research and practical work levels in ideological and political education. Only through learning and mastering the basic principles of artificial intelligence can we transcend science fiction literature about artificial intelligence and the marketing advertisements of intelligent product development enterprises, and make a generally accurate judgment about the scenarios and methods of applying artificial intelligence in ideological and political education.

智能时，人们普遍想到的是OpenAI公司开发的ChatGPT和Sora，百度公司开发的“文小言（文心一言）”，字节跳动公司开发的“豆包”以及杭州深度求索公司开发的DeepSeek等产品。这些产品被称为生成式人工智能，它们普遍依托深度学习技术，通过自主学习从海量数据中掌握词与词之间联结的统计学规律，从而具备生成自然语言的能力，数据、算法和算力是支撑这种人工智能的基本要素。按照人工智能产品的分类方式，当前主流的人工智能属于专用、弱人工智能：“专用”指的是这些人工智能总是为特定目的训练而成的，主要用于自然语言处理；“弱”指的是这种人工智能虽然可以在无监督状态下学习，但它不具备自我意识，只能完成人类给定的任务，不能自主提出任务，也不能拒绝人类发出的指令。当前我们说思想政治教育应用的人工智能，指的就是应用这种专用、弱人工智能，由于它整合了互联网上现有的各类资源（包括部分思想政治教育的研究成果），因而能够为思想政治教育赋能，有望促进思想政治教育理论研究和实际工作水平的提升。通过对人工智能基本原理的学习与掌握，我们才能够超越关于人工智能的科幻文学和智能产品开发企业的营销广告，对思想政治教育应用人工智能的场景和方式有一个大体准确的判断。

Artificial intelligence will empower academic research in ideological and political education. Artificial intelligence “concentrates” and “condenses” various types of knowledge accumulated in the past. When researchers use artificial intelligence to conduct research, they can stand on the “shoulders of giants” of various disciplines to engage in new thinking. For an individual researcher, applying artificial intelligence can enable them to acquire content that exceeds their personal knowledge accumulation and experience, thereby helping researchers expand their thinking. For example, in the study of social trends of thought, artificial intelligence can quickly integrate previous research results on similar issues, helping to complete tedious tasks such as searching, surveying, statistics, and sorting. Some parts of ideological and political education research particularly require the collection and analysis of empirical data. Artificial intelligence has obvious advantages in this regard; it can grasp and analyze the attention, search heat, public sentiment, and public opinion trends of certain topics, thus providing new methods for ideological and political education research. As technologies such as text-to-image and text-to-video gradually mature, the presentation methods of ideological and political education research results will also become richer. By continuously adjusting the presentation of results through human-machine dialogue, unexpected results may “emerge” from the answers given by artificial intelligence, which will further stimulate theoretical research to deepen. When artificial intelligence makes tasks such as information collection, historical sorting, and literature review easier, researchers in ideological and political education may be able to devote more energy to promoting theoretical innovation itself.

人工智能将赋能思想政治教育学术研究。人工智能中“浓缩”“凝结”了以往积累下来的各类知识，研究者们运用人工智能开展研究时就能够站在各个学科“巨人的肩膀”上展开新的思考。对单个研究者来说，应用人工智能能够习得一些超出个人知识积累和经历的内容，进而帮助研究者拓展思路。比如在社会思潮研究中，人工智能能够快速整合以往同类问题的研究成果，帮助完成搜索、调查、统计、梳理等繁琐工作。思想政治教育研究中有一部分特别需要搜集和分析实证数据，人工智能在这方面的优势很明显，它能够掌握与分析某些话题的关注度、搜索热度、公众情绪和舆论走势等，这就提供了思想政治教育研究的新方法。随着文生图、文生视频等技术的逐步成熟，思想政治教育研究成果的呈现方式也将变得丰富。通过人机对话不断调整成果的呈现方式，人工智能给出的答案中有可能“涌现”出意想不到的结果，这又会进一步刺激理论研究走向深入。当人工智能让信息收集、历史梳理、文献综述等变得容易时，思想政治教育研究者就有可能将更多的精力用于推进理论本身的创新。

Applying artificial intelligence helps to enhance the targetedness and effectiveness of ideological and political education. Artificial intelligence can be a very good teaching assistant. Currently, artificial intelligence has integrated a considerable portion of online teaching resources, demonstrating powerful capabilities in areas such as studying classic Marxist works, interpreting the latest Party and state documents and meeting spirits, and answering various questions that perplex students. This will, to a certain extent, help achieve the sharing of high-quality resources, personalized learning with real-time companionship, and lifelong learning in ideological and political education. When we use past data from ideological and political education to train artificial intelligence products, such products will provide assistance in curriculum design, organizing teaching materials, expanding work ideas, and assisting in teaching evaluation. For example, the widespread application of artificial intelligence can provide targeted analysis of learning situations, helping teachers conduct targeted teaching for different majors (distinguishing between science/engineering and humanities/social sciences), different educational stages (distinguishing between universities, middle schools, and primary schools), and different types of schools (distinguishing between undergraduate universities and vocational colleges). Another example is that theoretical teaching in ideological and political education is always inevitably abstract; at this point, text-to-video type artificial intelligence can play an auxiliary teaching role. This type of artificial intelligence can not only restore value content to specific life scenarios, such as restoring situations like the “trolley problem” or “moral blackmail” through text-to-video, but can even generate some scenarios that are currently difficult to experience firsthand but conform to theoretical conceptions, such as envisioning the situation of immense material abundance and distribution according to need in communist society through text-to-video and virtual reality. Values education particularly requires situational experience. Using artificial intelligence technology to generate educational scenarios, allowing students to experience the practical basis of values in specific scenarios, is conducive to improving the effectiveness of ideological and political education.

应用人工智能有助于提升思想政治教育的针对性和实效性。人工智能可以是很好的助教。目前，人工智能已经整合了相当一部分网络上的教学资源，在马克思主义经典著作研读、党和国家最新文件和会议精神阐释、学生困惑的各类问题的解答等方面显示了强大的功能，将在一定程度上有利于实现思想政治教育中优质资源的共享、实时陪伴的个性化学习和终身学习。当我们运用以往的思想政治教育数据训练人工智能产品时，这种产品将在课程设计、整理教学素材、开拓工作思路、辅助教学评价方面提供帮助。比如，人工智能的广泛应用能够提供有针对性的学情分析，帮助教师针对不同专业（理工科与人文社会科学专业的区分）、不同学段（大中小学的区分）、不同类型的学校（本科高校与高职高专院校的区分）进行有针对性的教学。再如，思想政治教育中理论教学总是难免抽象，此时文生视频类人工智能将起到辅助教学的作用。这种人工智能不仅能够将价值观的内容还原到具体的生活场景中，如通过文生视频的方式还原“电车难题”“道德绑架”等情境，甚至可以生成一些当前很难现场感受但符合理论构想的场景，如通过文生视频和虚拟现实的方式设想共产主义社会的物质财富极大涌流和按需分配情境。价值观教育特别需要情境体验，借助人工智能技术生成教育场景，让学生在具体场景中体会价值观的现实基础，有利于提升思想政治教育的实效。

Artificial intelligence will participate in some routine work in ideological and political education, can provide data support for carrying out refined and segmented ideological and political education, and enhance the targetedness of ideological and political education. For example, in some news reporting and policy propaganda work, generative artificial intelligence can quickly generate diverse propaganda materials and carry out customized production and precise dissemination based on the behavioral habits of internet users. The Third Plenary Session of the 20th Central Committee of the CPC proposed to “form a mechanism for segmented and precise implementation of online ideological and moral education.” [1]1 The Third Plenary Session of the 20th Central Committee of the CPC: a major meeting of the Communist Party of China’s top leadership, held in July 2024, which focused on deepening reform and advancing Chinese modernization. Whether based on the requirements of the Party and the state, or on the development plans of the artificial intelligence industry under the new situation of rapid information technology development, artificial intelligence has great potential in empowering segmented and precise ideological and moral education. In China’s development plan for artificial intelligence, one of the tasks to achieve key breakthroughs in critical common technologies is natural language processing technology, “researching short text computation and analysis technology, cross-language text mining technology and semantic understanding technology oriented towards machine cognitive intelligence, and human-machine dialogue systems for multimedia information understanding.” [2]2 This is a reference to China’s national development plan for artificial intelligence, likely the “New Generation Artificial Intelligence Development Plan” issued by the State Council in 2017.17 In terms of evaluating the effectiveness of ideological and political education, artificial intelligence technology helps to continuously track changes in the ideological concepts of educational objects. By tracking the changes in the thoughts and behaviors of the educated and analyzing relevant data, the effectiveness of ideological and political education can be tested.

人工智能将参与思想政治教育中的部分常规工作，能够为开展精细化、分众化的思想政治教育提供数据支撑，提升思想政治教育的针对性。比如在一些新闻报道、政策宣传工作中，生成式人工智能能够快速生成多样化的宣传材料，并针对网络用户行为习惯进行定制化生产和精准化传播。党的二十届三中全会提出，要“形成网上思想道德教育分众化、精准化实施机制”。[1]无论是根据党和国家的要求，还是基于信息技术迅猛发展新形势下人工智能行业的发展规划，人工智能在赋能思想道德教育分众化、精准化方面都大有可为。在我国关于人工智能的发展规划中，在关键共性技术上要实现重点突破的任务之一就是自然语言处理技术，“研究短文本的计算与分析技术，跨语言文本挖掘技术和面向机器认知智能的语义理解技术，多媒体信息理解的人机对话系统”。[2]17在思想政治教育效果的评价方面，人工智能技术有助于持续追踪教育对象的思想观念变动，通过追踪受教育者的思想、行为变化以及对相关数据的分析研究，检验思想政治教育的实效性。

II. Creating the Technical Conditions for the Application of Artificial Intelligence in Ideological and Political Education

二、创造思想政治教育应用人工智能的技术条件

The reason why artificial intelligence can demonstrate astonishing capabilities in information retrieval, human-machine dialogue, and content production is inseparable from sufficient guarantees in data, algorithms, and computing power. The application of artificial intelligence in ideological and political education also requires resource guarantees in these three aspects, so that the currently formed basic large models can be specially trained to empower ideological and political education activities.

人工智能之所以能够在信息检索、人机对话、内容生产方面展现出惊人的能力，离不开数据、算法和算力等方面的充分保障。思想政治教育应用人工智能同样需要这三个方面的资源保障，使当前已经形成的基础大模型经过专门训练后得以赋能思想政治教育活动。

The application of artificial intelligence in ideological and political education requires certain macro-level guarantee conditions, which are difficult to achieve solely through the efforts of ideological and political educators. However, with the country’s significant progress in the field of artificial intelligence and its emphasis on the integration of ideological and political education with information technology, these guarantees will become increasingly adequate. From the perspective of the national artificial intelligence development plan, the application of AI in ideological and political education is an inevitable trend. Xi Jinping pointed out: “China attaches great importance to the profound impact of artificial intelligence on education, actively promotes the deep integration of AI and education, drives educational reform and innovation, fully leverages the advantages of AI, and accelerates the development of education that accompanies everyone throughout life, is equally accessible to everyone, suits everyone, and is more open and flexible.” [3]3 Xi Jinping, “A Letter to the International Conference on Artificial Intelligence and Education,” *People’s Daily*, May 17, 2019. In 2017, the State Council issued the *New Generation Artificial Intelligence Development Plan* (hereinafter referred to as the *Plan*), which sets forth the goals, tasks, and technical roadmap for China’s AI development in 2020, 2025, and 2030. The *Plan* states that it is necessary to establish a new generation of basic theoretical systems for AI, a new generation of key common technology systems for AI, and to coordinate the layout of AI innovation platforms. “Carry out the construction of smart campuses, and promote the application of AI throughout the entire process of teaching, management, and resource development. Develop comprehensive three-dimensional teaching fields and online learning platforms based on big data intelligence. Develop intelligent education assistants, and establish intelligent, fast, and comprehensive education analysis systems. Establish a learner-centered educational environment, provide precisely targeted educational services, and achieve the customization of daily education and lifelong education.” [2]2 This is a reference to China’s national development plan for artificial intelligence, likely the “New Generation Artificial Intelligence Development Plan” issued by the State Council in 2017.27 The training and maintenance of AI models is an extremely resource-intensive activity, and in recent years, the cost issue of applying AI has received widespread attention. Large models require a large number of high-performance chips during pre-training and massive amounts of training corpora during prolonged training, which necessitates energy support. Collecting various types of data generated by educators and educational objects imposes new requirements on data storage and processing capabilities, as well as the operational and maintenance costs of AI. Against the backdrop of building a resource-conserving and environment-friendly society, the application of AI in ideological and political education needs to consider energy constraints. The *Interim Measures for the Management of Generative Artificial Intelligence Services* released in July 2023 states: “Promote the construction of generative AI infrastructure and public training data resource platforms. Facilitate the coordinated sharing of computing power resources and enhance the utilization efficiency of computing power resources.” [4]4 The *Interim Measures for the Management of Generative Artificial Intelligence Services* is a Chinese regulatory document issued in July 2023 to govern the development and use of generative AI technologies. More relevant to ideological and political education activities is the *Outline for Building a Strong Education Country (2024–2035)*, which proposes a series of measures, including implementing the national education digitalization strategy, building a national education big data center, establishing an education-specific network and a computing power sharing network, and creating an AI education large model, among others. [5]5 The *Outline for Building a Strong Education Country (2024–2035)* is a Chinese policy document outlining long-term goals for educational development, including digitalization and AI integration. The state has already recognized the importance of building energy-efficient and low-carbon AI. Through technological innovation and resource sharing, the cost issues currently constraining AI development are expected to be resolved in the future.

思想政治教育应用人工智能需要一些宏观上的保障条件，凭借思想政治教育者的努力很难实现，但随着国家在人工智能领域的长足进步和对思想政治教育与信息技术结合的重视，这些保障将越来越充分。从国家的人工智能发展规划来看，思想政治教育应用人工智能是大势所趋。习近平指出：“中国高度重视人工智能对教育的深刻影响，积极推动人工智能和教育深度融合，促进教育变革创新，充分发挥人工智能优势，加快发展伴随每个人一生的教育、平等面向每个人的教育、适合每个人的教育、更加开放灵活的教育。”[3]2017年国务院发布了《新一代人工智能发展规划》（以下简称《规划》），提出了我国在2020、2025和2030年人工智能发展的目标、任务和技术路线。《规划》指出，要建立新一代人工智能基础理论体系、建立新一代人工智能关键共性技术体系、统筹布局人工智能创新平台等。“开展智能校园建设，推动人工智能在教学、管理、资源建设等全流程应用。开发立体综合教学场、基于大数据智能的在线学习教育平台。开发智能教育助理，建立智能、快速、全面的教育分析系统。建立以学习者为中心的教育环境，提供精准推送的教育服务，实现日常教育和终身教育定制化。”[2]27人工智能模型的训练和维护是一项极其耗费资源的活动，近年来应用人工智能的成本问题得到了普遍关注。大模型在预训练中需要大量高效能芯片，在长时间的训练中需要海量的训练语料，这就需要能源上的支持。采集教育者和教育对象产生的各类数据，对数据的存储和处理能力、人工智能的运行和维护成本等都提出了新的要求。在建设资源集约型、环境友好型社会的大背景下，思想政治教育应用人工智能需要考虑能源上的限制。2023年7月发布的《生成式人工智能服务管理暂行办法》中提出：“推动生成式人工智能基础设施和公共训练数据资源平台建设。促进算力资源协同共享，提升算力资源利用效能。”[4]同思想政治教育活动更为相关的政策是，《教育强国建设规划纲要（2024—2035年）》中提出了一系列举措，包括实施国家教育数字化战略，建好国家教育大数据中心，搭建教育专网和算力共享网络，打造人工智能教育大模型，等等。[5]国家已经注意到建设节能低碳的人工智能的重要意义，通过技术创新、资源共享等方式，当前制约人工智能发展的成本问题有望在未来解决。

Integrate the data resources accumulated in previous activities and absorb the data continuously generated when applying AI. Current AI research and development relies on deep learning technology to construct artificial neural networks, “automatically extracting key linguistic traces from human-scale corpora and using them for automatic language generation.” [6]6 This is a quotation from a scholarly source describing how AI language models extract linguistic patterns from large datasets. It uses massive amounts of data for debugging, modifies programs in a targeted manner based on feedback results, and finally achieves a correspondence between information input and output. According to the technical principles of deep learning, for an AI product to assist in ideological and political education work, it is necessary to input previous ideological and political education content, cases, processes, methods, and other materials into the AI system. Through repeated training, a relationship between specific inputs and specific outputs is established. Data quality has always been a key factor constraining the intelligence level of AI. The larger the data volume and the better the data quality, the more powerful the AI function. Conversely, in areas lacking data, AI may “speak nonsense with a straight face.” To train AI that adapts to the characteristics and reality of ideological and political education, it is necessary to extensively use the data generated in ideological and political education.

整合以往活动中已经积累的数据资源，吸收应用人工智能时不断产生的数据。目前人工智能研发依托深度学习技术，构建人工神经元网络，“从人类规模语料中自动提取关键性语言痕迹，并用于语言的自动生成”。[6]运用海量数据进行调试，根据反馈结果来针对性地修改程序，最后实现信息输入和输出之间的对应关系。按照深度学习技术原理，一款人工智能产品想要辅助思想政治教育工作，需要将以往的思想政治教育内容、案例、过程、方法等资料输入人工智能系统之中，通过反复训练，建立起特定输入和特定输出之间的关系。数据质量一直是制约人工智能的智能程度的重要因素，数据量越大、质量越好，人工智能的功能越强大，反之，在缺乏数据的领域，人工智能则可能出现“一本正经地胡说”的情况。要想训练出适应思想政治教育特性和实际的人工智能，就需要广泛使用思想政治教育中形成的数据。

At present, it appears that at least three types of data should be absorbed into model training: First, the research results already accumulated in ideological and political education and related disciplines, whether in the form of books and periodicals or digital results aggregated in databases such as “CNKI,” are high-quality data. However, due to issues such as copyright and cost, they have not yet been fully absorbed into AI model training. In the future, through digital conversion and knowledge sharing, their entry into model training will greatly enhance the ability of AI to assist academic research in ideological and political education. Second, the Communist Party of China has always attached great importance to ideological and political education, and has accumulated a great deal of experience in the history of Party ideological and political education. This experience exists widely in the form of excellent courseware, typical cases, and teaching resources. After conversion through specific technologies to enter the training process of large models, it will make the application of AI in ideological and political education activities more convenient. Third, with the continuous advancement of the integrated construction of ideological and political education across universities, middle schools, and primary schools, the collection of teacher and student behavioral data is expected to achieve continuous tracking vertically and continuous expansion horizontally. As the data collected by AI becomes more extensive, the intelligent services it can provide will become more personalized and targeted. In summary, if the above three types of data can be aggregated into AI training and application, it will greatly enhance the level of ideological and political education.

目前看来至少有三类数据应该吸收到模型训练中：第一，思想政治教育以及相关学科中已经积累的研究成果，无论是以书籍报刊形式存在的还是类似“中国知网”等数据库中汇聚的数字化成果，都是优质数据，但由于版权、成本等问题，目前还没有被充分吸收到人工智能模型训练中。未来通过数字化转换和知识共享进入模型训练中，将大大提升人工智能助力思想政治教育学术研究的能力。第二，中国共产党历来重视思想政治教育，在党的思想政治教育史上积累了非常多的经验，这些经验以优秀课件、典型案例、教学资源的形态广泛存在于各处，经过特定技术的转化使其进入大模型的训练过程，会使思想政治教育活动应用人工智能更为便利。第三，随着大中小学思想政治教育一体化建设的持续推进，师生行为数据的采集有望实现纵向上的持续追踪、横向上的不断拓展。随着人工智能采集到的数据越来越广泛，它所能提供的智能服务将更加个性化、更具针对性。总之，以上三类数据若能汇聚到人工智能训练和应用中，将大大提升思想政治教育的水平。

Ideological and political educators should participate in the development and maintenance of AI products for ideological and political education. Currently, AI products on the market, such as “Doubao” and “Wenxiaoyan,” use some publicly available information on the internet during their initial training. They can answer any question, but many of their answers are not precise. Some scholars refer to this current type of large model as a “basic large model,” which “when applied to specific industries or fields, still needs to be fine-tuned with specialized data and knowledge to achieve knowledge transfer and skill acquisition, optimizing it into an ‘industry large model’ or ‘specialized large model’ that meets the needs of the specific industry or field.” [7]7 This is a quotation from a scholarly source distinguishing between general-purpose “basic large models” and domain-specific “industry large models.” The personnel involved in training basic large models mainly include model designers, software engineers, annotators, and model trainers, who primarily engage in algorithm and architecture design, programming, and model training and adjustment. Ideological and political educators may not be proficient in computer technology, but they can and should play an important role in developing AI specifically for ideological and political education activities. In the design activities of large models, they can use their professional knowledge in ideological and political education to provide suggestions for adjusting various parameters in the model, ensuring that technological development aligns with the core socialist values; in data annotation and model training, they can judge the quality of data input into the model through professional research, which will ensure the intelligence level of the model from the source; in the application of AI, they can propose adjustments and modifications from the perspective of educators to help with the iterative upgrading of AI products.

思想政治教育者应参与到思想政治教育类人工智能产品开发和维护中来。目前市场上的人工智能产品，比如“豆包”“文小言”等，在前期训练时，它们使用的是当前网上的一些公开信息，什么问题都可以回答，但其中许多问题都回答得不精。有学者称当前的这种大模型为“基础大模型”，“在应用到具体行业或领域时还需要结合专门的数据和知识进行微调，使它实现知识迁移和技能习得，优化为符合具体行业或领域需要的‘行业大模型’或‘专用大模型’”。[7]参与基础大模型训练的主要有模型设计师、软件工程师、标注师和模型训练师等，他们主要进行算法及架构设计、编程以及模型的训练和调整等工作。思想政治教育者可能不擅长计算机技术，但是在开发专门应用于思想政治教育活动的人工智能时可以且应该发挥重要作用。在大模型设计活动中，运用思想政治教育专业知识为调整模型中各项参数提供建议，确保技术发展与社会主义核心价值观相契合；在数据标注和模型训练中，通过专业的研究判断输入模型的数据质量，这将从源头处确保模型的智能水平；在人工智能的应用中，从教育者的视角出发提出调整和修改建议，以帮助人工智能产品的迭代升级。

III. Establishing Norms for the Application of Artificial Intelligence in Ideological and Political Education

三、建立思想政治教育应用人工智能的规范

Ideological and political education involves issues such as national ideological security and the physical and mental health of educational objects. Before applying AI, management departments should establish scientific and clear norms, and continuously improve them in practical application, thereby preventing and resolving various risks, so that the application of AI truly assists ideological and political education.

思想政治教育涉及国家意识形态安全、教育对象的身心健康等问题，在应用人工智能之前，管理部门应该建立起科学、明确的规范，并在实际应用中不断完善，从而防范化解各类风险，让人工智能的应用真正为思想政治教育助力。

Establish the “intelligent” standards for intelligent products in ideological and political education. What constitutes intelligence, how many types of artificial intelligence exist, and in what direction AI should be developed are all contested issues within academia. Various internet companies are developing their own products and universally affix the “intelligent” label in their marketing. In practical application, different schools may cooperate with different internet enterprises and use different types of AI products. Tests have shown that posing the same question to different AI products yields inconsistent feedback, indicating that differences in algorithm design and data sources among different AIs lead them to produce different results. Some scholars have pointed out that the AI currently applied in teaching activities mainly includes four categories: behavior detection, learning models, prediction models, and intelligent assessment, each with its technical advantages and application limitations.[8]8 The reference [8]10 and [8]10-11 cite a Chinese academic source (likely a journal article or book) discussing the classification and limitations of AI in education, and the proposal for a third-party evaluation body. The specific source is not provided in the text.10 Teaching AI is still a considerable distance from comprehensive and mature application. Therefore, “the government could organize experts in AI and education to form an authoritative third-party evaluation body, in order to distinguish the genuine from the fake among the myriad teaching intelligent products on the market.”[8]8 The reference [8]10 and [8]10-11 cite a Chinese academic source (likely a journal article or book) discussing the classification and limitations of AI in education, and the proposal for a third-party evaluation body. The specific source is not provided in the text.10-11 For academic questions, AI responses should comply with norms regarding intellectual property rights, and adhere to maintaining the diversity of academic viewpoints and the rigor of academic argumentation. In traditional knowledge production activities, the accuracy and stability of knowledge production are guaranteed by the academic community; peer review, editorial review, and reader oversight are important means of ensuring the professionalism of knowledge production. However, the various types of information generated by AI have neither standardized norms nor the gatekeeping of the academic community, making it a challenge to ensure the accuracy of the knowledge they produce and disseminate. For various life-related questions, AI responses need to comply with laws, regulations, and public order and good customs, avoiding any discriminatory answers, and so forth. The AI industry is a technology-intensive industry. What products can be called “intelligent” products, and how many types and levels of “intelligent” products exist, are matters that ordinary users cannot accurately judge. This requires relevant departments to take the lead, with experts from related fields such as ideological and political education and AI jointly participating in formulating standards. As AI technology is updated, these standards must also be adjusted accordingly, so as to provide a universal reference for the application of AI in ideological and political education.

制定思想政治教育中智能产品的“智能”标准。关于什么是智能，有多少种人工智能，要朝什么方向研发人工智能，在学界都是有争议的问题。各个互联网公司都在开发自己的产品，并且在营销中普遍贴上“智能”的标签。在实际应用时，不同的学校可能会与不同的互联网企业合作，使用不同类型的人工智能产品。经过测试，同样的问题在不同的人工智能产品中提问，得到的反馈不尽相同，说明不同人工智能在算法设计、数据来源上的差异会导致它们给出不同的结果。有学者指出，目前在教学活动中应用的人工智能主要包括行为探测、学习模型、预测模型和智能评测四类，每个类别都有它的技术优势和应用限度。[8]10教学人工智能距离全面、成熟的应用还有不小的距离，因此，“政府或可组织人工智能、教育领域的专家成立权威的第三方评估机构，以对市场上形形色色的教学智能产品去伪存真”。[8]10-11对于学术类问题，人工智能的回答应该遵守知识产权方面的规范，坚持维护学术观点的多样性和学术论证的严谨性。传统的知识生产活动中，知识生产的准确性和稳定性是由学术共同体来保证的，同行评议、编辑审校、读者监督等是保证知识生产专业性的重要手段，而人工智能生成的各类信息既没有规范标准，也没有学术共同体的把关，如何保证它们生产和传播的知识是准确的，就成了一个难题。对于各类生活问题，人工智能的回答需要遵守法律法规、公序良俗，避免任何歧视性的答案，等等。人工智能产业是技术密集型产业，什么样的产品可以叫“智能”产品，“智能”产品有多少种类、多少层级，一般用户无法准确判断。这需要有关部门牵头，思想政治教育和人工智能等相关行业的专家共同参与制定标准。随着人工智能技术的更新，这些标准也要随之调整，这样才能够给思想政治教育应用人工智能提供一个普遍的参照。

Clarify the rules for data collection and analysis in ideological and political education activities. The application of AI in ideological and political education requires a continuous influx of new data to ensure that AI is constantly upgraded through ongoing training. Based on different types of data, the rules for data collection and use should be clarified. First, much of the data accumulated in ideological and political education activities is stored in archives, libraries, and electronic databases in the form of files, books, and papers. Some of this data is free resources and can directly enter the model training process through digital conversion; some has clear intellectual property rights and needs to undergo data transactions in accordance with relevant national regulations before entering the training process. Second, when AI is applied in ideological and political education, activity data of teachers and students will be collected. Because internet companies have obvious advantages in computing power, algorithms, and data, the AI used in ideological and political education is mostly provided by internet enterprises with technology and equipment. Developing and maintaining AI requires substantial research investment, and market value and return on capital are matters that internet enterprises must consider. In recent years, internet enterprises have obtained the right to use user data by providing free network services and signing agreements. “Through such agreements, internet companies can permanently and freely use the vast amount of information produced by users, turning it into their valuable assets. At the same time, all behaviors of users when using network services are meticulously recorded and analyzed, placed into each person’s ‘digital archive.’”[9]9 The reference [9] cites a Chinese academic source discussing how internet companies obtain user data through free service agreements and create “digital archives.” The specific source is not provided in the text. “Free service—user agreement—data value-added” is the mainstream binding method between internet companies and users. Users express informed consent when signing the agreement. However, because internet enterprises hold the initiative in formulating and modifying user agreements, users’ informed consent is often only formal in nature. If data is collected in this manner, then the application of AI in ideological and political education may pose risks to the personal information rights of teachers and students. Therefore, when schools introduce intelligent technical services from internet enterprises, they should take the initiative, grasp the initiative in data collection and use, and not allow enterprises to arbitrarily dispose of ideological and political education activity data. Third, the teaching management experience accumulated by various schools in ideological and political education activities, as well as the large number of scientific research and teaching results formed by teachers in teaching, will be gathered through methods such as teaching demonstrations and lesson plan collections, in compliance with relevant laws and regulations such as the Copyright Law of the People’s Republic of China. The aggregation of such high-quality data is a powerful guarantee for AI-empowered ideological and political education activities.

明确思想政治教育活动中数据采集和分析的规则。思想政治教育应用人工智能需要源源不断的新数据，从而保证人工智能在持续训练中不断升级。根据数据的不同类型，应该明确数据采集和使用的规则。第一，思想政治教育活动中积累的许多数据以档案、书籍、论文等形式存储于档案馆、图书馆和电子数据库中。这些数据中有一部分是免费资源，通过数字化转换能够直接进入模型训练过程；一部分具有明确的知识产权，需要按照国家相关规定进行数据交易后再进入训练过程。第二，思想政治教育应用人工智能时会采集师生的活动数据。由于互联网公司在算力、算法、数据等方面有明显优势，思想政治教育中应用的人工智能大多由互联网企业提供技术和设备。开发和运维人工智能需要大量的科研投入，市场价值、资本回报率是互联网企业必须考虑的事情。近年来互联网企业通过提供免费的网络服务，以签订协议的方式来获得用户数据的使用权。“通过这类协议，互联网公司可以无偿永久性使用用户生产的大量信息，成为其宝贵的资产。同时，用户使用网络服务的一切行为也会被精细地记录和分析，放入每个人的‘数字档案’。”[9]“免费服务—用户协议—数据增值”是互联网公司与用户的主流绑定方式，用户在签订协议时表示知情同意，不过，由于互联网企业掌握着制定和修改用户协议的主动权，使得用户的知情同意在很多时候都只具有形式意义。如果按照此种方式采集数据，那么思想政治教育应用人工智能则可能给师生的个人信息权利带来风险。因此，学校在引入互联网企业的智能技术服务时，应该发挥主动性，掌握数据采集和使用上的主动权，而不放任企业随意处置思想政治教育活动数据。第三，各个学校在思想政治教育活动中积累的教学管理经验，以及教师在教学中形成的大量的科研和教学成果，这些成果将在遵守《中华人民共和国著作权法》等相关法律法规的基础上，通过教学展示、教案征集等方式汇集起来，这类优质数据的汇聚是人工智能赋能思想政治教育活动的有力保障。

Determine the access conditions for intelligent products in ideological and political education activities. The debugging of AI data models is a process of continuous trial and error. Currently, most AI on the market has undergone a process of trial—promotion—debugging—upgrade. It is precisely through continuous trial and error and competition among different development routes that the intelligence level of products is continuously improved. Many units are currently piloting the construction of smart campuses, and some AI products related to ideological and political education are being developed and trialed. Ideological and political education activities are highly ideological, involving national ideological security and the physical and mental health of educational subjects, and many aspects cannot tolerate errors. This requires the establishment of specialized AI research and development norms tailored to the characteristics of ideological and political education, including what standards must be met for trials, how to continuously maintain and upgrade during operation, and so on. Currently, influenced by the technical characteristics of AI, its results have, to a certain extent, a “black box” attribute, that is, inexplicability. Ideological and political education, however, requires transparency of technology, and there is a certain tension between the two. Therefore, we cannot completely hand over ideological and political education work to AI, especially some important decisions, which must rely on the judgment of ideological and political educators.

确定思想政治教育活动中智能产品的准入条件。人工智能的数据模型调试是一个不断试错的过程，目前市场上的人工智能大都经历了试用—推广—调试—升级的过程，正是在不断试错中、在不同开发路线的竞争中，产品的智能水平得到不断提升。目前很多单位已经在试点推进智慧校园建设，一部分同思想政治教育有关的人工智能产品正在开发和试用。思想政治教育活动的意识形态性很强，涉及国家意识形态安全和教育对象的身心健康，很多地方是不容许出错的。这就需要针对思想政治教育的特性建立专门的人工智能研发规范，包括达到何种标准可以试用，如何在运行中不断维护和升级，等等。当前，受人工智能的技术特性影响，其得出的结果在一定程度上具有“黑箱”属性，即不可解释性。而思想政治教育则要求技术的透明性，二者之间存在一定的张力。因此我们不能将思想政治教育工作完全交给人工智能，尤其是一些重要决策，必须依靠思想政治教育者的判断。

Improve the review mechanism for content generated by intelligence in ideological and political education activities. Data and algorithms are influenced by human factors, and ideological differences in reality also affect AI. Intelligent products in the field of ideological and political education must adhere to political standards as the primary criterion, uphold the guiding position of Marxism in the ideological domain, maintain the correct political stance in answering every specific question, be familiar with the Party and the state’s lines, principles, and policies, be familiar with the latest conference spirit, ensure accurate interpretation of the Party’s latest theories, and firmly take a stand on key issues. Through testing, the AI currently accessible on the domestic market can basically identify the ideological nature of questions and adhere to the correct political stance. In contrast, some generative AI products developed by overseas enterprises, while claiming to be value-neutral when facing political questions, adopt a seemingly “objective” presentation method such as “some viewpoints hold that… while other viewpoints hold that…” in their answers. However, as long as the questioning method is slightly altered, it can be seen that their answers contain obvious ideological tendencies, and the materials they provide are also specious. This type of AI hides liberal values within algorithms and data. For users of AI products, they face the results presented on the “front end,” but it is difficult to grasp the computational process on the “back end.” Therefore, when developing AI for ideological and political education, it is necessary to lead algorithm design with core socialist values and conduct ideological review of the data resources added to the AI training process.

健全思想政治教育活动中智能生成内容的审核机制。数据和算法会受到人为因素的影响，现实中的意识形态差异亦会影响人工智能。思想政治教育领域的智能产品必须坚持政治标准第一，坚持马克思主义在意识形态领域的指导地位，在每一个具体问题的回答上坚持正确的政治立场，熟悉党和国家的路线方针政策、熟悉最新的会议精神，保证对党的最新理论的准确阐释，在关键问题上站稳立场。经过测试，目前国内市面上可以接触到的人工智能基本上都能识别出提问的意识形态性，并做到坚持正确的政治立场。相比之下，部分境外企业开发的生成式人工智能产品在面对政治问题时虽然总是说自己是价值中立的，在回答问题时会采用“有的观点认为……而有的观点认为……”这种看似“客观”的呈现方式，但只要稍微变换提问方式，就能看出它的答案中包含着明显的意识形态倾向，它所提供的资料也存在似是而非的情况。此类人工智能将自由主义价值观隐藏在算法和数据中，对于人工智能产品的用户来说，他们面对的是“前台”呈现出的结果，而很难把握“后台”的运算过程。因此，在开发思想政治教育类人工智能时要以社会主义核心价值观引领算法设计，对加入人工智能训练过程的数据资源进行意识形态上的审核。

Formulate norms for the application of artificial intelligence in ideological and political education based on laws and regulations. To ensure that artificial intelligence can always recommend appropriate educational resources based on educators’ needs and provide personalized instruction tailored to the needs of educational objects, the application of artificial intelligence in ideological and political education activities requires the collection of users’ personal information, which will bring risks of infringing on user privacy and personal information. Ideological and political education needs to formulate norms for the application of artificial intelligence based on current laws and regulations and its own characteristics. The current relevant laws and regulations mainly include the Cybersecurity Law of the People’s Republic of China, the Personal Information Protection Law of the People’s Republic of China, the Data Security Law of the People’s Republic of China, and other laws, as well as the Interim Administrative Measures for Generative Artificial Intelligence Services, the Regulations on the Administration of Students in Ordinary Institutions of Higher Education, and other regulations. Complying with these laws and regulations in the use of artificial intelligence is both a basic requirement for adhering to law-based governance and provides a legal basis for ideological and political educators to apply artificial intelligence. Taking the Data Security Law of the People’s Republic of China as an example, Article 21 stipulates: “The state establishes a system for classified and graded protection of data, implementing classified and graded protection of data based on the importance of data in economic and social development, and the degree of harm to national security, public interests, or the legitimate rights and interests of individuals and organizations once the data is tampered with, destroyed, leaked, or illegally obtained or utilized.” [10]10 The Data Security Law of the People’s Republic of China, Article 21, as cited in the original text. It also stipulates: “All regions and departments shall, in accordance with the system for classified and graded protection of data, determine the specific catalogues of important data for their regions, departments, and related industries and fields, and provide key protection for the data included in the catalogues.” [10]10 The Data Security Law of the People’s Republic of China, Article 21, as cited in the original text. Currently, the rules for the classification and grading of data in ideological and political education activities are not yet complete, and since ideological and political education concerns national ideological security and the healthy growth of students, data protection norms should be established at the outset of applying artificial intelligence and continuously improved in practice. Due to the particularity of ideological and political education, the author believes that data can be divided into four types: first, data related to national ideological security; second, data related to students’ ideological and political concepts; third, data related to students’ physical and mental health; and fourth, data related to students’ diverse lifestyles. Different data will require different handling methods.

基于法律法规制定思想政治教育应用人工智能的规范。为了保证人工智能始终能够基于教育者的需要推荐合适的教育资源、针对教育对象的需要进行个性化的教学，思想政治教育活动应用人工智能时就需要采集用户的个人信息，而这将带来侵犯用户隐私和个人信息的风险。思想政治教育需要基于当前法律法规和自身特性制定应用人工智能的规范。目前相关法律法规主要有《中华人民共和国网络安全法》《中华人民共和国个人信息保护法》《中华人民共和国数据安全法》等法律，《生成式人工智能服务暂行管理办法》《普通高等学校学生管理规定》等规章。遵照这些法律法规来使用人工智能，既是坚持全面依法治国的基本要求，也让思想政治教育者应用人工智能有了法律依据。以《中华人民共和国数据安全法》为例，其中第二十一条规定：“国家建立数据分类分级保护制度，根据数据在经济社会发展中的重要程度，以及一旦遭到篡改、破坏、泄露或者非法获取、非法利用，对国家安全、公共利益或者个人、组织合法权益造成的危害程度，对数据实行分类分级保护。”[10]其中还规定：“各地区、各部门应当按照数据分类分级保护制度，确定本地区、本部门以及相关行业、领域的重要数据具体目录，对列入目录的数据进行重点保护。”[10]目前，关于思想政治教育活动中数据的分级分类规则还不完善，而思想政治教育又关乎国家意识形态安全、学生健康成长，因此在应用人工智能之初就应该建立起数据保护规范，并在实践中不断完善。由于思想政治教育的特殊性，笔者认为可以将数据分为四种：一是同国家意识形态安全相关的数据，二是和学生思想政治观念相关的数据，三是和学生身心健康相关的数据，四是同学生多样化生活方式相关的数据，数据不同，处理方式也会有所变化。

(Author’s bio: Chang Yanhui, Associate Professor and Doctoral Supervisor at the School of Marxism, Nankai University)

（作者简介：常宴会，南开大学马克思主义学院副教授、博士生导师）

Online editor: Tongxin

网络编辑：同心