The Ideological and Political Education Large Model: Generative Logic, Core Essentials, and Practical Application

Since 2022, generative artificial intelligence represented by ChatGPT has swept the globe. In January 2025, the domestic large model DeepSeek emerged, once again attracting widespread global attention. This type of artificial intelligence, based on a large model architecture, exhibits an unprecedented "intelligence emergence," regarded as a milestone in the history of AI development, marking a leap forward in artificial intelligence. Large-model AI holds great potential in the field of education and is increasingly advancing toward specialization, thereby providing high-quality intelligent services for various types of education. For example, NetEase Youdao's HiEcho is a virtual English oral private tutor based on the "Confucius Says" education large model; the Chinese Academy of Educational Sciences, in collaboration with iFlytek, released a high school mathematics intelligent teacher system based on the iFlytek Spark large model technology. Addressing the specific needs of ideological and political education, and supported by large model technology, extending human intelligence in the field of ideological and political education into the realm of artificial intelligence can generate a specialized ideological and political education large model (思政大模型). An in-depth exploration of the generative logic, core essentials, and practical application of the ideological and political education large model holds significant value for advancing it from theoretical conception to practical operation.

2022年以来，以ChatGPT为代表的生成式人工智能席卷全球。2025年1月，国产大模型DeepSeek的横空出世，又引发全球的广泛关注。这种以大模型为基础架构的人工智能呈现出前所未有的“智能涌现”，被视为人工智能发展史上的里程碑，标志着人工智能的跨越式发展。大模型人工智能在教育领域具有巨大潜能，并日益朝向专业化的方向精进，从而为各类教育提供优质的智能服务。例如，网易有道推出的HiEcho即是以“子曰”教育大模型为基础的虚拟英语口语私教、中国教科院携手科大讯飞基于讯飞星火大模型技术发布的高中数学智能教师系统等。针对思想政治教育的特定需求，以大模型技术为支撑，将人在思想政治教育领域的智能延展到人工智能领域，可以生成一种专门服务于思想政治教育的思政大模型。深入探讨思政大模型的生成逻辑、内核要义以及实践运用，对于推动其从理论构想走向实践操作具有重要价值。

I. How Is the Ideological and Political Education Large Model Generated?

一、思政大模型何以生成？

The ideological and political education large model is a product of the integration of ideological and political education with large models. Its generation process relies on technical components such as data, algorithms, and computing power of large models, takes the performance of ideological and political education as its core, and ultimately serves ideological and political education.

思政大模型是思想政治教育与大模型融合而成的产物，其生成过程有赖于大模型的数据、算法和算力等技术构件，以思想政治教育性能为核心，最终服务于思想政治教育。

1. Based on Big Data of Ideological and Political Education

1.以思想政治教育大数据为基础

Data is the "raw material" for generating large models. A large model is an artificial neural network constructed through deep learning on massive data, with hundreds of millions or trillions of parameters, and "based on its ultra-large-scale language model, it achieves an exponential increase in its own intelligence." In other words, the transcendent intelligent characteristics of large models arise because they absorb a sufficiently large amount of data and have a sufficiently large model size, thus being better able to "understand" human natural language. Therefore, the ideological and political education large model should logically take the big data of ideological and political education as its "raw material." On the one hand, the "understanding" of a large model is limited by its data content. The ideological and political education large model needs to "learn" data from the field of ideological and political education in order to "understand" ideological and political education. Machine learning is like human learning: to understand a field, one must first study the content materials of that field to form an "understanding" of it; otherwise, it can only be "unable to answer any question." On the other hand, data has an ideological attribute. For the ideological and political education large model to follow the value orientation of ideological and political education, it must absorb data with ideological and political education value. Data is not an objective product of the natural world; especially when humans are the subjects of data production, data is an external manifestation and extension of the ideology of the producing subject, possessing a distinct value orientation. Ideological and political education has a distinct ideological character. Since the ideological and political education large model is to serve ideological and political education and be guided by mainstream ideology, it needs to absorb data from the field of ideological and political education, integrate the principles, standpoints, viewpoints, and methods of ideological and political education contained in this data into the model, and thereby endow the model with the value attributes of ideological and political education inherent in the data. It should be noted that the ideological and political education large model must not only learn the data content of the ideological and political education field but also ensure the correctness of the value of data in this field, eliminating data that does not conform to the value orientation of ideological and political education, so as to guarantee that the ideological and political education large model has the correct value orientation.

数据是大模型生成的“原材料”。大模型是基于对海量数据进行深度学习后构建起来的具有千万亿个参数的人工神经网络，并“基于其超大规模语言模型，来实现自身智能的指数级上涨”。换句话说，大模型超越性的智能特征是因为其吸收的数据量足够多、模型体量足够大，所以更能“理解”人类自然语言。为此，思政大模型理应以思想政治教育大数据为“原材料”。一方面，大模型的“认识”受限于其数据内容。思政大模型需要“学习”思想政治教育领域的数据，才能“理解”思想政治教育。机器学习就像人类学习一样，要了解一个领域需要先学习这个领域的内容资料，才能形成对这个领域的“认识”，否则只能是“一问三不知”。另一方面，数据是具有意识形态属性的。思政大模型要遵循思想政治教育的价值导向，就必须吸纳具有思想政治教育价值的数据。数据并非自然世界的客观产物，尤其是当人作为数据生产主体时，数据是生产主体意识形态的一种外显与延伸，具有鲜明的价值导向性。思想政治教育具有鲜明的意识形态性。思政大模型既然要服务于思想政治教育、以主流意识形态为引领，就需要吸纳思想政治教育领域的数据，将这些数据中蕴含的思想政治教育原理、立场、观点、方法融入模型，进而把数据蕴含的思想政治教育价值属性赋予模型。需要注意的是，思政大模型不仅要学习思想政治教育领域的数据内容，更要保障思想政治教育领域数据的价值正确，剔除不符合思想政治教育价值指向的数据，才能保障思政大模型具有正确的价值导向。

2. Supported by Appropriate Algorithms

2.以适切算法为支撑

Algorithms are the core support for model generation and a key factor in transforming data into a large model. In essence, "an algorithm is code input into a computer to solve a specific problem, which outputs results through specific computations, data processing, and automated reasoning, and is the foundation of computer science." Algorithms dictate how machines operate, giving machine operations direction and standards. An important factor contributing to the birth of large models is precisely the progress in algorithms, which has optimized the training speed, resource utilization rate, and architecture of large models, mainly relying on breakthroughs in deep learning and reinforcement learning algorithms. Deep learning algorithms endow machines with the ability to "learn" like humans: "provide it with sufficient and appropriate data, and through this data, it will discover corresponding knowledge; give it a video stream, and it will watch; give it a library, and it will read; give it results of physics experiments, and it will discover physical laws." Reinforcement learning algorithms provide effective guidance for machine learning, enabling it to learn in the "direction" specified by humans. In short, under the technological revolution of learning algorithms, machines can learn and understand data more efficiently and "obediently." Empowered by deep learning algorithms, the large model can achieve a comprehensive learning of the essence, functions, processes, elements, and content of ideological and political education. Coupled with reinforcement learning algorithms, the large model can conduct guided learning oriented toward ideological and political education. Based on this, the ideological and political education large model can "understand" the content coverage, value functions, goals and tasks, and educational processes of ideological and political education, "master" basic concepts such as fostering virtue through education (立德树人) and comprehensive development, and content systems including ideological education, political education, legal education, and mental health education, thereby carrying out appropriate educational actions and implementing the fundamental task of fostering virtue through education. It is particularly important to note that algorithms also have an ideological attribute. As Marx pointed out: "Nature does not construct any machines, no locomotives, railways, electric telegraphs, self-acting mules, etc. They are the products of human industry; natural material transformed into organs of the human will over nature, or of human participation in nature." Algorithms, as products created by humans, are part of the extension of human will. For the ideological and political education large model to serve ideological and political education, the "will" of its algorithms should logically align with the value orientation of ideological and political education, follow mainstream values, adhere to guiding algorithms with mainstream value orientation, design algorithmic logic under the guidance of Marxist theory, and ensure that the large model operates within the value domain of ideological and political education. Shaping the ideological and political education large model requires creating such a deep learning and reinforcement learning type of "ideological and political algorithm."

算法是模型生成的核心支撑，是将数据转化成大模型的关键因素。就其本质而言，“算法是为解决特定问题而输入计算机的代码，经过特定运算、数据处理和自动推理而输出结果，是计算机科学的基础”。算法规定了机器如何运行，让机器运作变得有方向、有标准。促成大模型诞生的一个重要因素正是算法的进步，使大模型的训练速度、资源利用率以及大模型的架构等都得到了优化，其主要依赖深度学习类和强化学习类算法的突破。深度学习类算法赋予了机器像人类一样进行“学习”的能力，“为它提供足够、适当的数据，通过这些数据，它会发现相应的知识；给它视频流，它就会观看；给它图书馆，它就会阅读；给它物理实验结果，它就会发现物理定律”。而强化学习类算法为机器学习提供了有效引导，使其能够按照人类指定的“方向”进行学习。简言之，在学习算法的技术革命下，机器能够更高效且“听话”地对数据进行学习和理解。在深度学习类算法赋能下，大模型能够实现对思想政治教育的本质、功能、过程、要素和内容等的全面学习。加之强化学习类算法，大模型得以进行以思想政治教育为指向的引导式学习。基于此，思政大模型才能“理解”思想政治教育的内容涵盖、价值功能、目标任务、育人过程，“掌握”思想政治教育的立德树人、全面发展等基本理念，思想教育、政治教育、法治教育、心理健康教育等内容体系，进而开展适切的育人行动，落实立德树人根本任务。需要特别注意的是，算法也是具有意识形态属性的。诚如马克思所指出的：“自然界没有造出任何机器，没有造出机车、铁路、电报、自动走锭精纺机等等。它们是人的产业劳动的产物，是转化为人的意志驾驭自然界的器官或者说在自然界实现人的意志的器官的自然物质。”算法作为人所创造的产物，属于人的意志延伸的一部分。而思政大模型要想服务于思想政治教育，其算法的“意志”理应契合思想政治教育的价值取向，以主流价值为遵循，坚持以主流价值导向驾驭算法，在马克思主义理论指导下设计算法逻辑，保障大模型在思想政治教育的价值疆域内运行。思政大模型的塑造，需要打造这样一种深度学习、强化学习类的“思政算法”。

3. Guaranteed by Powerful Computing Power

3.以强大算力为保障

“Computing power mainly refers to the capacity for data processing” and serves as the “engine” for the operation of large models. The training process in which a large model learns from data through algorithms requires the support of computing power, just as human productive activities also require power. Computing power largely determines the speed at which a large model operates; the more powerful the computing power, the greater the amount of computation the large model can complete per unit of time, and the shorter its learning cycle, which can greatly save the time cost of large model training, thereby enhancing the performance of the large model. The ideological and political large model requires ultra-strong computing power as a supporting guarantee to meet the specific requirements of its generation process, practical application, and performance optimization. Regarding the generation process of the ideological and political large model, it needs to achieve seamless continuous interaction with humans in the field of ideological and political education, which imposes extremely high demands on the model’s parameters, and thus also places high demands on the supply of computing power. At the same time, it is also necessary to purify the professionalism of the large model, absorb more data resources from the field of ideological and political education, and eliminate inappropriate content, which also imposes high demands on the model’s computing power. Regarding the practical application of the ideological and political large model, ideological and political education is a complex practical activity, encompassing many tasks such as knowledge transmission, thought enlightenment, and value guidance, involving multidisciplinary content including political science, law, history, pedagogy, and psychology, and the issues involved are also diverse and complex. Such a complex situation requires more computing power support for the reasoning and generation of the ideological and political large model. Regarding the performance optimization of the ideological and political large model, it needs to constantly update its data domain; it must not only learn the basic principles of Marxism but also update in real time the latest achievements of the Party’s innovative theory. It must not only grasp information about the current world landscape but also promptly supplement new variables, and dynamic updating imposes even higher demands on computing power. In short, strong computing power is an important support for the ideological and political large model; without powerful computing power, the large model is like a car without power and cannot operate.

“算力主要指对数据的处理能力”，是大模型运行的“发动机”。大模型通过算法对数据进行学习的这一训练过程需要算力的支撑，就如同人进行生产活动也需要动力一般。算力在很大程度上决定着大模型运行的速率，算力越强大，大模型在单位时间内能够完成的计算量就越大，其学习周期就越短，可以极大地节省大模型训练的时间成本，进而提升大模型性能。思政大模型需要超强的算力作为支撑保障，以满足思政大模型的生成过程、实践应用和性能优化等方面的具体要求。就思政大模型的生成过程而言，思政大模型需要在思想政治教育领域实现与人的无障碍连续性交互，这对大模型参数提出了极高要求，故而对算力供给也提出了很高的要求。同时，还需要对大模型的专业性进行提纯，吸纳更多的思想政治教育领域的数据资源以及对不当的内容进行剔除，这也对大模型算力提出了较高要求。就思政大模型的实践应用而言，思想政治教育是一项复杂的实践活动，涵盖知识传授、思想启蒙、价值引导等诸多任务，涉及政治学、法学、历史学、教育学、心理学等多学科内容，涉及的问题也具有多样性和复杂性。如此复杂的情况使思政大模型在推理和生成中需要更多的算力支撑。就思政大模型的性能优化而言，思政大模型需要时刻更新数据疆域，不仅要学习马克思主义基本原理，还需要对党的创新理论的最新成果进行实时更新。不仅需要掌握当下的世界格局信息，还需要对新增变量及时补足，而动态化更新对算力又提出了更高的要求。总之，强算力是思政大模型的重要支撑，没有强大的算力，大模型就像一辆没有动力的汽车一样无法运转。

4. Taking the Performance of Ideological and Political Education as the Core

4.以思想政治教育性能为核心

Performance is the ultimate goal of a large model and determines its use value. The emergence of large models is the result of big data, strong algorithms, and large computing power. “Compared with other digital technologies, the transcendence of generative artificial intelligence lies in its generative value.” [2]2 The original Chinese text cites a source here, but the source is not provided in the input. The note is retained to mark the citation. Large models can “understand” the meaning of natural language and generate multimodal content that meets needs, such as text, images, audio, video, and code, thereby improving the efficiency of users in production, work, and learning. Differences in performance are also manifested in this aspect; the better the performance of a large model, the more accurately it can “understand” instruction requirements and generate product materials that better meet user needs. The performance of the ideological and political large model needs to take the attributes of ideological and political education as its core. In terms of basic performance, the ideological and political large model needs to be able to accurately “understand” the discourse semantics of ideological and political educators and educatees, precisely identify educational scenarios, expand ideological and political education knowledge and match resources, interact seamlessly with subjects, ensure the accuracy of services, and avoid situations of “fabricating nonsense.” At the same time, as a large model serving ideological and political education, it should possess distinct characteristics of ideological and political education, have a clear value stance, be able to and only provide content materials of the mainstream ideology, understand the laws of human thought and behavior development, align with the goal orientation of ideological and political education, and master the educational methods in the process of ideological and political education, only then can it effectively promote the digital and intelligent transformation of ideological and political education.

性能是大模型的最终指向，决定着模型的使用价值。大模型的出现是基于大数据、强算法和大算力的结果。“相较于其他数字技术，生成式人工智能的超越性在于其生成性价值”。大模型可以“理解”自然语言的含义，并生成符合需求的多模态内容，如文本、图像、音频、视频、代码等，进而可以提高使用者的生产、工作、学习等效率。性能的差异也表现在这一方面，性能越好的大模型，越能精确“理解”指令要求，并能生成更加符合用户需求的产品素材。思政大模型的性能需要以思想政治教育属性为核心。在基础性能方面，思政大模型需要具备精确“理解”思想政治教育者和受教育者的话语语义，精准识别育人场景，进行思想政治教育知识的拓展和资源的匹配，无障碍地和主体进行互动，保障服务的精准性，避免出现“胡编乱造”的情况。同时，思政大模型作为一种服务于思想政治教育的大模型，应当具备鲜明的思想政治教育特性，具有鲜明的价值立场，能够且只能提供主流意识形态的内容材料，理解人的思想与行为发展规律，契合思想政治教育的目标指向，掌握思想政治教育过程中的育人方法，才可能有效推动思想政治教育数智化变革。

II. How Can the Ideological and Political Large Model Be Understood?

二、思政大模型何以理解？

The ideological and political large model is a practical tool related to ideological and political education, possessing characteristics such as ideological nature, human-likeness, instrumentality, integration, and evolution, and has functions including model-based data integration, highly intelligent continuous dialogue, and intelligent skill assignment.

思政大模型是一种关于思想政治教育的工具实物，具备意识形态性、类人性、工具性、集成性、进化性等特征，且有着模型化的数据整合、高智性的连续对话和智能化的技能赋值等功能。

1. The Connotation of the Ideological and Political Large Model

1.思政大模型的内涵

The so-called ideological and political large model refers to a specialized large model generated on the basis of big data in ideological and political education, appropriate algorithms, and powerful computing power, which embodies the positions, viewpoints, values, and other attributes of ideological and political education, can serve the education and guidance of mainstream ideology, and promotes the digital and intelligent development of ideological and political education. To understand this connotation, attention should be paid to the following aspects: First, the ideological and political large model is the result of the effective combination of the “intelligence emergence” of large models with ideological and political education. “Intelligence emergence” refers to the transcendent intelligence exhibited by large models. After the release of the ChatGPT-3.5 large model, its monthly user count exceeded 100 million, making it the fastest-growing application in history. “The reason ChatGPT’s user growth could be so fast is that it possesses super-strong functions that no previous artificial intelligence had.” [3]3 The original Chinese text cites a source here, but the source is not provided in the input. The note is retained to mark the citation. These super-strong functions are precisely the result of “intelligence emergence” in large-parameter models, including the emergence of new capabilities such as language generation and reasoning, and the exponential improvement in data recognition and utilization performance. This transcendent intelligence is key to constructing the ideological and political large model. Second, the ideological and political large model is a large model that integrates general intelligence and professional intelligence in ideological and political education, with the ultimate goal of serving ideological and political education. Based on the requirements of the complex practical activity of ideological and political education, the ideological and political large model needs to have general intelligence logic, relying on abilities such as understanding, generation, logic, and memory to accurately understand the various needs and diverse environmental information of ideological and political educators and educatees, and to solve complex problems. And based on the professionalism of ideological and political education, the ideological and political large model also needs professional intelligence logic, able to form an ideological and political education content system, transmit mainstream ideology, etc., and even possess “expert-level ideological and political wisdom.” Third, the ideological and political large model is an important support for the digital transformation and intelligent evolution of ideological and political education. The ideological and political large model can become a bridge for the mutual integration of ideological and political education and digital technology, becoming the “singularity” of the digital and intelligent transformation of ideological and political education. Finally, the ideological and political large model is essentially a practical tool dependent on humans. The “intelligence emergence” exhibited by the ideological and political large model makes it appear to be a “new subject” with agency that can even replace educators in completing some tasks, but this is not the case. The ideological and political large model always requires human drive as its “switch”; it is not an autonomous and active individual, nor can it completely replace the role of educators. “In the practice of ideological and political education in the era of artificial intelligence, only by maintaining the basic requirement of ‘people first’ can the risk of ideological and political education being ‘backlashed’ by intelligent technology be avoided.” [4]4 The original Chinese text cites a source here, but the source is not provided in the input. The note is retained to mark the citation.

所谓思政大模型，是指在思想政治教育大数据、适切算法和强大算力基础上生成的体现思想政治教育立场、观点、价值等属性的，且能够服务于主流意识形态教育和引导的，推动思想政治教育数智化发展的专业化大模型。对这一内涵的理解，需要注意以下几个方面：首先，思政大模型是大模型实现“智能涌现”后与思想政治教育有效联合的结果。“智能涌现”是指大模型展现出的超越性智能。ChatGPT-3.5大模型发布后月用户数超过1亿人次，成为历史上用户增长最快的应用程序。“ChatGPT用户增长之所以能够这么快，是因为它具有以往任何人工智能都不具备的超强功能。”这些超强功能正是大参数模型产生“智能涌现”的结果，包括语言生成能力、推理能力等新能力的出现，数据识别和利用性能的指数级提升等。这种超越性智能是构建思政大模型的关键。其次，思政大模型是集通用智能和思想政治教育专业智能于一体的大模型，其最终目的是为了服务思想政治教育。基于思想政治教育这一复杂实践活动的要求，思政大模型需要具有通用智能逻辑，依托理解、生成、逻辑、记忆等能力精准理解思想政治教育者、受教育者的各类需求和多元环境信息，并解决复杂问题。而基于思想政治教育的专业性，思政大模型也需要专业智能逻辑，能够形成思想政治教育内容体系、传导主流意识形态等，甚至具备“专家级思政智慧”。再次，思政大模型是思想政治教育数字化转型和智能化演进的重要依托。思政大模型可以成为思想政治教育与数字技术相互融合的桥梁，成为思想政治教育数智化转型的“奇点”。最后，思政大模型本质上是一种依赖于人的实物工具。思政大模型呈现的“智能涌现”使其看似是一个具有能动性甚至能代替教育者完成部分任务的“新主体”，实则不然。思政大模型始终需要人的驱动作为“开关”，并非是一个自主能动的个体，更不能完全替代教育者的角色。“在人工智能时代的思想政治教育实践中，唯有保持‘人本至上’的基本要求，才可能规避思想政治教育被智能技术‘反噬’的风险。”

2. Characteristics of the Ideological and Political Large Model

2.思政大模型的特征

As a new phenomenon combining ideological and political education with large model technology, the ideological and political large model derives its characteristics from these two aspects, organically integrating their features to produce new traits. First, ideological nature. Ideological nature is the fundamental characteristic of the ideological and political large model. Ideological nature is the fundamental attribute of ideological and political education; the ideological and political large model should prioritize “ideological and political considerations first.” All its empowerment of ideological and political education can only be realized on the premise that its value stance aligns with mainstream value guidance. Therefore, the ideological and political large model must adhere to the guiding position of Marxism and uphold the basic principles of socialism, among other things. Second, human-like intelligence. The ideological and political large model possesses a “brain” far exceeding the limits of human memory and can engage in continuous dialogue interactions, even conceptual interactions, “like a human” with ideological and political educators and educatees, providing various material resources for ideological and political education. More importantly, the ideological and political large model can “understand” the meaning and logic of natural language, generate new content conforming to instructions based on raw data in the database, and realize human-machine dialogue, thereby “playing the role” of an ideological and political educator or educatee to conduct virtual educational activities. Third, specialized intelligent tool nature. As an intelligent tool, the ideological and political large model can significantly enhance capabilities in data analysis, graphic and textual creation, etc., becoming a lesson preparation tool, teaching tool, evaluation tool, and management tool, thereby improving the effectiveness of ideological and political education. At the same time, as a media tool, its intelligent generation capability and continuous dialogue capability enable it to empower the dissemination of ideological and political education content. Thus, the ideological and political large model serves as a specialized intelligent tool specifically for ideological and political education, markedly distinguishing it from general-purpose large models. Fourth, learning progressiveness. Learning progressiveness is the technical characteristic of the ideological and political large model. Learning is a form of evolution; with the support of deep learning and reinforcement learning algorithms, the ideological and political large model “continuously engages in autonomous learning, achieving self-renewal, thereby becoming increasingly ‘intelligent,’” promptly grasping the latest dynamics of the ideological and political education environment, trends in content composition, emerging changes at the methodological level, and new manifestations of subject characteristics and other information and experiences. This drives the emergence of more advanced versions such as the ideological and political large model 2.0 and 3.0, further promoting the digital and intelligent development of ideological and political education.

思政大模型作为思想政治教育和大模型技术相结合的新鲜事物，其特征也源于这两个方面，在吸收二者特性的基础上有机融合出新的特征。一是意识形态性。意识形态性是思政大模型的根本特征。意识形态性是思想政治教育的根本属性，思政大模型应以“思政为先”，它对思想政治教育所有的赋能都是在其价值立场契合主流价值导向前提下才能实现。因此，思政大模型必须坚持马克思主义的指导地位，坚守社会主义的基本原则等。二是类人智慧性。思政大模型具有一个远超人类记忆极值的“大脑”，并能够“像人一样”同思想政治教育者、受教育者开展连续性对话交互甚至是观念交互，可以为思想政治教育提供各种素材资源。更为重要的是，思政大模型可以“理解”自然语言的含义和逻辑，可以基于数据库中的原始数据生成符合指令的新内容并实现人机对话，从而“扮演”思想政治教育者或受教育者的角色，开展虚拟的教育活动。三是专业智能工具性。思政大模型作为智能工具，可以大幅提升数据分析、图文创作等能力，成为备课工具、教学工具、评价工具、管理工具，进而提升思想政治教育的有效性。同时，作为媒介工具，其智能生成能力和连续性对话能力使其可以赋能思想政治教育内容的传播。由此，思政大模型是作为一种专门服务于思想政治教育的专业化智能工具，显著区别于通用大模型。四是学习进阶性。学习进阶性是思政大模型的技术特征。学习就是一种进化，思政大模型在深度学习和强化学习类算法的加持下，“不断地自主学习，做到自我革新，从而变得越来越‘聪明’”，及时地掌握思想政治教育环境的最新动态、内容构成的革新趋势、方法论层面的新兴变迁以及主体特性的新型展现等信息和经验，从而推动思政大模型2.0版本、3.0版本等更高级版本的出现，进一步推动思想政治教育数智化发展。

3. Functions of the Ideological and Political Large Model

3.思政大模型的功能

Based on the general logic of large models, the ideological and political large model can be integrated into various scenarios and processes of ideological and political education, creating a human-machine co-presence situation within the field of ideological and political education, thereby achieving multiple functional empowerments. First, model-based data integration function. The large model technology on which the ideological and political large model relies makes the data storage logic manifest as “harmonious integration and commonality.” Unlike the individualized logic of the internet era—characterized by platform differentiation, webpage segmentation, and independent databases—the large model integrates all big data of ideological and political education through deep learning and reinforcement learning, storing it in the form of a model. Accordingly, barriers between various databases of ideological and political education are broken, and the links between data across dimensions such as subject, object, mediator, and environment are unprecedentedly strengthened. Based on this, the scale effect of data is more deeply reflected, making it easier to grasp the laws of practical education in ideological and political education, the mutual mapping of thought and behavior, and the influence of the environment. Second, highly intelligent continuous dialogue function. Generative artificial intelligence can provide continuous dialogue interactions “like a human,” giving rise to a new paradigm of sustained, simulated “human-machine interaction” in ideological and political education. Such interaction is not shallow mechanical dialogue but, based on the “human-like intelligence” attribute of the large model, constructs situational dialogues by displaying specific knowledge backgrounds and behavioral patterns, presenting different role images when interacting with users, thereby endowing it with a highly intelligent continuous dialogue function. Third, intelligent skill assignment function. Large models not only possess human-like traits such as calculation, understanding, memory, and creativity but also professional skills like data processing, view design, and sentiment analysis. The ideological and political large model can assign these skills to ideological and political educators. In the teaching, learning, management, and evaluation processes, educators need only express work requirements to achieve “one-click generation” of behavior monitoring, data analysis, and thought prediction for educatees, thereby enhancing the precision and effectiveness of ideological and political education.

基于大模型的通用逻辑，思政大模型可以融入思想政治教育的各个场景和过程中，在思想政治教育场域中营造出人机共在情景，从而实现多种功能性赋能。一是模型化的数据整合功能。思政大模型依托的大模型技术，使数据储存逻辑呈现为“和合共通”。与互联网时代的平台区分、网页区隔、数据库独立的个体化逻辑不同，大模型通过深度学习和强化学习将思想政治教育大数据全部融合且使其以模型的形式储存。依此，思想政治教育各类数据库之间的壁垒得以打破，主体、客体、介体和环境等各维度数据之间的链接得到前所未有的增强。基于此，数据的规模效应得以更深层次体现，思想政治教育实践育人规律、思想与行为的相互映射、环境的影响等更容易被把握。二是高智性的连续对话功能。生成式人工智能能够“像人一样”提供连续性的对话交互，使思想政治教育衍生出持续性、模拟性的“人机交互”新范式。这样的交互并非是浅层的机械性对话，而是在大模型具备的“类人智慧”属性基础上，通过展示特定的知识背景和行为模式等特征，在与用户交互时呈现出不同的角色形象而构建出的情景对话，从而使之具备高智慧性的连续对话功能。三是智能化的技能赋值功能。大模型不仅具有计算、理解、记忆、创造等类人性特征，还具有数据处理、视图设计、情感分析等专业性技能。思政大模型可以将这些技能赋予思想政治教育者。思想政治教育者在教、学、管、评各环节中仅需要表达工作需求，即可“一键生成式”地实现对受教育者的行为监测、数据分析和思想预测等，从而提升思想政治教育的精准性与实效性。

III. How to Apply the Ideological and Political Large Model?

三、思政大模型如何运用？

As a specialized tool serving ideological and political education, the key is to rely on its unique functions to integrate with ideological and political education, empowering innovations in resource utilization, interaction paradigms, and teaching methods of ideological and political education, thereby maximizing its effectiveness.

思政大模型作为一种服务于思想政治教育的专业化工具，关键是要依托其特有功能与思想政治教育相结合，赋能思想政治教育的资源运用、作用范式、施教方法等创新，从而发挥其最大功效。

1. Integrating Ideological and Political Education Data Resources Based on the Data Integration Function

1.基于数据整合功能融通思想政治教育数据资源

Currently, the application of big data in the field of ideological and political education has begun to emerge, with the establishment of digital platforms or large databases such as the National University Ideological and Political Work Network and the CPC Ideological and Theoretical Resource Database, achieving many accomplishments in practical application. However, these databases remain constrained by platform thinking, with barriers between different databases making it difficult to achieve overall resource integration, and the scale effect of resources cannot be fully realized. The ideological and political large model, based on deep learning and reinforcement learning algorithms, has a underlying logic of “harmonious integration and commonality.” By integrating all data into the large model through deep learning and reinforcement learning, it can break the boundaries between resource libraries or platforms, thereby constructing an integrated intelligent resource library existing in the form of a large model, and this resource library is the ideological and political large model itself. In other words, the data integration function of the ideological and political large model can greatly promote the integrated use of ideological and political education data resources.

当前，大数据在思想政治教育领域的应用已经崭露头角，建成了全国高校思想政治工作网、中国共产党思想理论资源数据库等数字平台或大型数据库，并在实践运用中取得了诸多成就。但是这些数据库仍然受制于平台思维，不同数据库之间存在壁垒，难以实现整体的资源融通，资源的规模效应难以充分发挥。思政大模型以深度学习类和强化学习类算法为基础，其底层逻辑是“和合共通”，通过深度学习和强化学习将数据全部融入大模型中，可以打破资源库或者平台之间的界限，进而构建以大模型形态存在的一体化智能资源库，而这个资源库就是思政大模型本身。换言之，思政大模型的数据整合功能可以极大地推动思想政治教育数据资源的贯通使用。

This is specifically manifested in the following aspects: First, using the ideological and political large model to enhance the integration of ideological and political education data resources. The ideological and political large model is a huge data repository, capable of efficiently absorbing large-scale data with the support of deep learning and reinforcement learning algorithms, distributed storage systems, and high computing power. Using the ideological and political large model to feed and integrate resources from ideological and political education-related databases—such as theoretical databases, policy documents, news reports, activity cases, and other diverse data, even multi-module data from different schools, educational stages, and subjects—aims to achieve as comprehensive a fusion and integration of all data in the field of ideological and political education as possible. Second, enhancing the effectiveness of using ideological and political education data resources through the ideological and political large model. Traditional database integration methods simply stack and summarize ideological and political education resources, whereas the model-based storage mode of the ideological and political large model represents a deep integration application, dedicated to uncovering the ideological and political imprints behind data resources. Using the ideological and political large model to process ideological and political education data through deep learning, detailed classification, and logical connection, it links related data fields, enhances the intrinsic relevance of data, and then mines from these data resources deeper content such as thought evolution, behavioral patterns, moral shaping, and value tendencies that can be used for ideological and political education, assisting in the conduct of ideological and political education activities. Third, realizing embodied services of ideological and political education data resources through the ideological and political large model. Model-based data storage, through precise screening and reorganization mechanisms, can provide highly customized resource retrieval services based on users’ specific needs. Subjects such as ideological and political educators and educatees can, leveraging the data relevance characteristics of the ideological and political large model, precisely locate resource information that best matches their own needs from the vast sea of ideological and political education big data. Moreover, under the rigorous deconstruction, reorganization, and other deep screening and sorting processes of the ideological and political large model’s algorithms, they can obtain mature content generated from the integration and reorganization of existing ideological and political education data resources, thereby meeting their own ideological and political education needs.

这具体表现在以下几方面：一是运用思政大模型提升思想政治教育数据资源的集成性。思政大模型是一个巨大的数据收纳库，在深度学习类和强化学习类算法、分布式储存系统和大算力的支撑下可以高效吸纳大规模数据。运用思政大模型对思想政治教育相关数据库进行资源投喂和整合，比如理论数据库以及思想政治教育类政策文件、新闻报道、活动案例等多样态数据，甚至是不同学校、不同学段、不同主体的多模块数据等，以尽可能多地实现思想政治教育领域的全数据融合集成。二是通过思政大模型提升思想政治教育数据资源使用的有效性。传统数据库的整合方式是将思想政治教育资源进行简单的堆叠汇总，而基于模型化存储模式的思政大模型则属于一种深度整合运用，致力于深挖数据资源背后的思政印记。运用思政大模型对思想政治教育数据进行深度学习、细致分类与逻辑联结等处理，串联相关领域的数据，提升数据的内在关联性，进而从这些数据资源中挖掘能够为思想政治教育所用的更深层次的思想演变、行为模式、品德塑造、价值倾向等内容，助力思想政治教育活动的开展。三是借助思政大模型实现思想政治教育数据资源的具身化服务。模型式的数据储存通过精密的筛选与重组机制，能够依据使用者的特定需求提供高度定制化的资源检索服务。思想政治教育者、受教育者等主体能够借助思政大模型的数据关联性特征，于海量的思想政治教育大数据中精确锁定与自身需求最为契合的资源信息，并且在思政大模型算法严谨的解构、重组等深度筛选和整理流程下，获得对现有思想政治教育数据资源整合重组后生成的成熟内容，从而满足自身的思想政治教育需求。

2. Innovating the Interaction Paradigm of Ideological and Political Education Based on the Continuous Dialogue Function

2.基于连续对话功能革新思想政治教育交互范式

Ideological and political education aims at the comprehensive development of the individual, involving education in thought, values, mental health, and other aspects; it is a highly complex practical activity. Therefore, only through exploratory, heuristic, and interactive teaching from the dimensions of thinking cultivation, value shaping, moral nurturing, and political enlightenment can the educational effect be maximized. The unique "human-like intelligence" attribute of the ideological and political large model and its continuous dialogue capability bring a new "human-computer interaction" paradigm to ideological and political education, making real-time interaction of "one student with one teacher" a reality, and the subjectivity of the educated will be unprecedentedly highlighted.

思想政治教育以实现人的全面发展为目标，涉及人的思想、价值、心理健康等多个方面的教育，是一项十分复杂的实践活动。因此，唯有从思维培养、价值塑造、品德涵养和政治启迪等维度对受教育者进行探讨式、启发式、互动式施教，才能使育人效果最大化。思政大模型特有的“类人智慧”属性及其连续对话能力为思想政治教育带来了新的“人机交互”范式，“一个学生配一个教师”的即时交互得以成为现实，受教育者的主体性将得到前所未有的彰显。

Specifically, the ideological and political large model can assist in the practical activities of ideological and political education through the following interaction paradigms. First, achieving exploratory continuous interaction. The ideological and political large model does not simply provide content materials and guidance but can offer dialogues with deep connotations. On the same issue, the large model and the educated can express their own views, even engage in sharp debates, allowing the educated to achieve ideological enlightenment, value guidance, and political inspiration in deep-level logical dialogue during the debate, forming correct values. Second, providing heuristic continuous interaction. Using the ideological and political large model, the teaching process can be optimized based on set teaching themes and goals, designing problem-oriented guided teaching around core objectives. The continuous and segmented dialogue capability of the large model allows for continuous heuristic questioning of the educated, and based on their feedback, personalized cyclical prompts are continuously adjusted, guiding them to draw inferences and analogies, achieving a tightly closed-loop heuristic teaching from problem posing and thought stimulation to goal attainment, ultimately reaching the objectives of ideological and political education. Third, creating simulated continuous interaction. Utilizing the "role simulation" capability of the ideological and political large model, a "simulated interaction object" can be shaped that mimics the linguistic expression, behavior, and emotional responses of the ideological and political education subject, thereby conducting simulated interactions. For the educated, this adds an "intelligent teacher" to the learning process; for the educator, it adds a "virtual object" to the teaching process, allowing for the deduction and simulation of the teaching process. Fourth, conducting scenario-based interaction. "Scenario experience is a key link in scenario-based teaching triggering the generation of emotional attitudes and values in college students." The ideological and political large model, based on real-time continuous dialogue, offers the possibility of simulating the educator role, accompanying the educated into various real scenarios, and conducting ideological and political education based on specific scenario information, enabling the educated to gain thought, value, and emotional inspiration through real-scenario perception.

具体来看，思政大模型可以借助以下交互范式，助力思想政治教育实践活动的开展。一是实现探讨式连续交互。思政大模型并非只是简单地提供内容素材和指导，而是能够提供具有深层内涵的对话。针对同一问题，思政大模型和受教育者可以各抒己见，甚至是开展针锋相对的争论，在争论中让受教育者在深层次的逻辑对话中实现思想启蒙、价值引导和政治启迪，形成正确的价值观。二是提供启发式连续交互。运用思政大模型可以基于设定的教学主题和目标优化教学流程，围绕核心目标设计具有问题导向的引导式教学。思政大模型的连续性和分众化对话能力，可以向受教育者进行连续的启发式提问，再基于其反馈情况不断进行个性化的循环提示，引导其触类旁通、举一反三，实现从问题提出、思维激发到目标达成的严密闭环的启发式教学，最终达成思想政治教育目标。三是打造仿真式连续交互。运用思政大模型所具备的“角色模拟”能力，可以塑造出模拟思想政治教育主体的语言表述、行为举止以及情感反应等多维特征的“仿真交互对象”，进而开展仿真交互。对于受教育者而言，在学习过程中多了一个“智能教师”；对于教育者而言，在教学过程中多了一个“虚拟对象”，可以进行施教过程的演绎和模拟。四是开展场景式交互。“场景体验是场景式教学触发大学生情感态度价值观生成的关键环节。”思政大模型基于即时性连续对话带来了这样一种可能，即运用思政大模型模拟教育者角色，伴随受教育者进入各种现实场景，并基于具体场景信息开展思想政治教育，使受教育者能够在真实场景的感悟中得到思维、价值和情感启迪。

3. Innovating Ideological and Political Education Teaching Methods Based on Skill Assignment Function

3.基于技能赋值功能创新思想政治教育施教方法

In the digital intelligence era, the deep integration of cutting-edge technologies such as big data and the metaverse with ideological and political education has become a common focus in both academia and practice. Although these technologies hold immense value potential, their implementation in the field of ideological and political education still faces many challenges. Among them, the issue of high technical thresholds is particularly prominent. If ideological and political educators do not master the professional skills of these emerging technologies, it is difficult to truly apply technology to method innovation, leading to a dilemma of abundant technological resources but weak application. The ideological and political large model possesses professional skills such as data processing, content interpretation, emotional analysis, and audio-image-view generation. With its significant intelligence enhancement, a large number of digital-intelligence ideological and political education methods can be generated without requiring excessive skill assignment from educators, thereby achieving new development in ideological and political education teaching methods.

数智时代，大数据、元宇宙等前沿技术与思想政治教育深度融合，已成为学术界与实践领域共同关注的热点。尽管这些技术蕴含着巨大的价值潜能，但它们在思想政治教育领域的落地仍面临诸多挑战。其中，高技术门槛问题尤为凸显，思想政治教育者不掌握这些新兴技术的专业技能，就难以将技术真正应用于方法创新，继而陷入一种技术资源充裕但应用乏力的窘境。思政大模型具备的数据处理、内容解读、情感分析、音视图生成等专业化技能，在其显著的智能性加持下，一大批数智思想政治教育方法不需要教育者过多的技能赋值就能产生，从而实现思想政治教育施教方式的新发展。

Specifically, the main digital-intelligence teaching methods that can be applied in ideological and political education are as follows: First, adopting the intelligent learning companion method for teaching. The so-called intelligent learning companion method refers to relying on the ideological and political large model to construct an "intelligent learning companion" for ideological and political education in the form of apps, mini-programs, websites, etc., participating in the entire learning process of the educated. Using the large model to build intelligent interactive programs provides the educated with diverse content materials, responds instantly to their multidimensional needs, engages in deep interaction, guides personalized learning, provides real-time evaluation feedback, and creates an immersive shared learning atmosphere, thereby achieving companionship, observation, and guidance of the educated's learning process. Second, employing the expert system method for teaching. The so-called expert system method refers to a method based on artificial intelligence technology that simulates the decision-making and problem-solving processes of human experts in specific fields. By relying on the ideological and political large model's rich resources of ideological and political education knowledge theory, content materials, and practical cases, a "virtual expert" system in the field of ideological and political education can be constructed. Through various means such as theoretical knowledge supply, practical case reference, survey data analysis, social environment analysis, and systematic practice guidance, it can provide the educated with more refined, professional, and scientific knowledge services, ideological guidance, psychological counseling, and behavioral guidance. Third, applying the multimodal teaching method for teaching. The ideological and political large model possesses powerful intelligent generation capabilities, enabling the translation and conversion of ideological and political education content between "text-audio-image-view," thereby presenting diverse content based on the multiple sensory needs of the educated. Educators can use the large model to convert abstract textual content into intuitive visual presentations of theoretical connotations, match musical melodies to achieve auditory expression of textual emotions, and create diverse narrative agendas to strengthen classroom interactions such as collaboration, discussion, and role-playing among the educated, enhancing their participation and constructing a multimodal teaching model integrating sight, hearing, and speech.

具体来看，可以运用的思想政治教育数智施教方法主要有以下几种：一是采用智能学伴法进行施教。所谓智能学伴法，是指依托思政大模型以APP、小程序、网站等形式构建思想政治教育的“智能学伴”，参与受教育者的学习全过程。运用思政大模型构建智能交互程序，为受教育者提供多样化的内容素材，即时响应受教育者的多维需求，与其进行深度互动，指导受教育者进行个性化学习，提供即时性评价反馈，营造身临其境的共同学习氛围等，以此实现对受教育者学习过程的陪伴、观察和引导。二是借助专家系统法进行施教。所谓专家系统法，是指基于人工智能技术模拟人类专家在特定领域中的决策和问题解决过程的一种方法。依托思政大模型内涵丰富的思想政治教育知识理论、内容素材和实践案例等资源，构建出一个思想政治教育领域的“虚拟专家”系统，可以通过理论知识供给、实践案例借鉴、调研数据解析、社会环境分析和系统实践指导等多种方式为受教育者提供更为精细化、专业化、科学化的知识服务、思想引领、心理疏导及行为指导等。三是运用多模态教学法进行施教。思政大模型具备强大的智能生成能力，可以实现思想政治教育内容“文—音—视—图”的互译转换，进而结合受教育者的多元感官需求进行丰富多样的内容呈现。教育者可以运用思政大模型将抽象的文字内容转换为理论内涵的直观视觉化呈现，将音乐旋律进行匹配，实现文本情感的听觉化表达，创作出形式多样的叙事议程强化受教育者的协作交流、相互讨论和角色扮演等课堂交互，提升受教育者的参与度，构建出视、听、说一体的多模态教学模式。

(The authors are a Hanhong postdoctoral fellow (on-station) and an associate researcher at Southwest University)

（作者系西南大学含弘博士后（在站），副研究员）

Online Editor: Ma Jingren

网络编辑：马京仁