AI may propel a research paradigm shift. Photo: TUCHONG

In early 2025, a breakthrough version of DeepSeek was released. With its high degree of accuracy and low cost of use, it has been widely adopted in social science research, propelling social science research toward a “fifth paradigm.” The digital transformation of research objects, the intelligent upgrading of research tools, and the innovative shift in analytical paradigms are exerting broad and far-reaching influence on the disciplinary system, theoretical system, and knowledge system of the social sciences. At the same time, the challenges and social ethical concerns brought by the new artificial intelligence (AI) warrant close attention.

Empowering social science research

Compared with traditional AI, interactive AI represented by DeepSeek enables dynamic human–machine interaction. It not only provides real-time feedback based on researchers’ needs but also flexibly refines generated outputs through ongoing learning and optimization. The innovative advances of DeepSeek have lowered the threshold for social science researchers to use interactive AI and significantly enhanced research efficiency.

First, real-time collaboration and intelligent feedback foster research innovation. Unlike traditional AI models that merely analyze data and return responses based on preset parameters, interactive AI continually adjusts its analytical pathways in response to researchers’ evolving questions and hypotheses, generating new ideas and research directions. In traditional social science research models, researchers typically rely on their professional expertise and experience, drawing on large volumes of literature and data for reasoning and analysis. While this approach boasts a relatively high degree of internal consistency, it is often difficult to rapidly obtain precise insights when facing complex social phenomena and massive datasets. A core advantage of interactive AI lies in its ability to engage in deep interaction with researchers, respond to research needs in real time, and provide intelligent feedback that departs from the one-way processing model of traditional AI. More importantly, interactive AI is capable of continuous learning and self-optimization. In the course of interacting with researchers, it continues to “think” and “l(fā)earn,” refining its algorithms and models to offer increasingly accurate and insightful analyses.

Second, the automation of hypothesis generation and verification deserves emphasis.

In traditional social science research, generating and verifying hypotheses is often a time-consuming process involving repeated reasoning, hypothesis formulation, and testing. Researchers typically propose hypotheses on the basis of existing theoretical frameworks, literature reviews, and data analysis, and then verify them through experimental design or data collection. This process relies heavily on subjective judgment and is constrained by human cognitive limits.

However, with the rise of interactive AI, hypothesis generation and validation have become more automated, precise, and efficient. Through big data analytics and machine learning algorithms, interactive AI can help researchers discover potential relationships and patterns from large volumes of unlabeled data and propose research hypotheses. Once a hypothesis is formed, interactive AI can also assist in rapid verification. This data-driven automated process overcomes the limitations of traditional manual reasoning and accelerates the transition of social science research from theory-driven to data-driven and algorithm-driven paradigms.

Reshaping disciplinary and theoretical systems

With the advent of the interactive AI revolution and the widespread popularization of related technologies in social science research, the disciplinary system, theoretical system, and knowledge system of the social sciences will undergo a profound transformation.

First, disciplinary integration and disciplinary system restructuring will play a central role. The evolution of disciplinary systems has always mirrored advances in social technology and the division of labor. With the arrival of the AI revolution and the era of big data, integrating traditional disciplinary systems with new technologies has become inevitable.

In the AI era, as research demands grow in complexity, the knowledge and tools of a single discipline are increasingly insufficient. Integrating disciplinary knowledge with AI has therefore become imperative to guide disciplinary development towards an AI-oriented direction.

Natural sciences were the first to integrate with AI and achieved substantial outcomes, while the social sciences initially lagged behind. However, the rapid development of generative AI has created favorable conditions for integration, enabling new interdisciplinary fields such as “economics + AI,” “sociology + AI,” and “political science + AI.” These interdisciplinary developments broaden research horizons and provide new methods and perspectives for addressing complex problems in social sciences. Interactive AI not only enhances innovative research through intelligent tools and interactive reasoning but may also gradually replace not only repetitive tasks but certain forms of creative labor. With AI’s continued iteration and development, it poses new challenges to the curriculum system, knowledge system, and overall disciplinary frameworks of social science disciplines.

Second, theoretical renewal and knowledge system reconstruction are urgently required. Interactive AI, with its powerful interactive, deep thinking, and reasoning capabilities, differs from all previous technological revolutions and introduces a wide range of new questions for social science research. As AI becomes a new object of inquiry, social science theories must be renewed and expanded. Historically, each technological revolution has reshaped economies and societies at multiple levels. Unlike earlier technologies that mainly focused on replacing manual labor, AI marks a shift toward the substitution of human cognitive abilities. This qualitative transformation calls for innovation and development in social science research. For example, economics must now address how the replacement of high-skilled workers with interactive AI affects areas such as the labor market, employment structures, and income distribution. Sociology must respond to the severe challenges AI poses for social governance and explore how to reconstruct existing governance systems to adapt to the AI era. Existing theoretical systems have not yet accounted for AI-driven societal transformation. Facing these new real-world problems, it is urgent to innovate and develop, continuously improving the knowledge system of social science to meet new challenges and demands.

Posing multi-level challenges

While AI’s rapid iteration and upgrading has greatly facilitated social science research, it has also introduced a series of explosive new challenges to social development that demand careful attention.

First, data security and privacy protection must be prioritized. The accuracy of AI-generated information still requires verification. For social science researchers, the learning mechanisms behind large language models present a high degree of opacity due to their vast parameter sets and complex architecture. This opacity creates challenges for transparency and explainability in research and raises ethical concerns around the application of AI in the social sciences. Privacy protection presents an additional challenge. The training and optimization of AI systems depend on vast datasets that often include sensitive personal information such as identity and behavioral data. Effectively protecting personal privacy throughout the entire life cycle of data collection, storage, and processing, as well as preventing data leakage, abuse, or unauthorized access, has become a key challenge requiring urgent attention. This not only involves designing technical security mechanisms but also touches upon complex issues such as legal norms, ethical standards, and social trust.

Second, identifying and protecting intellectual property has become a growing concern. Interactive AI can generate poetry, novels, and academic papers according to user requirements and with remarkable efficiency, offering new creative tools for cultural creation and academic research. However, the widespread application of such technologies has also sparked numerous controversies, particularly as the issue of copyright ownership remains unclear. Who exactly holds the copyright for the generated content—the user, the developer, or the model itself—remains an unresolved legal and ethical conundrum. For instance, if a user merely provides simple instructions while the model independently does most of the work, should the user still be considered the legitimate copyright holder? Moreover, the vast amount of data used in model training may involve copyrighted works of others, raising the question of potential infringement. These issues not only affect the legality of technology application but also pose challenges to academic norms and ethical standards in the social sciences. Addressing these problems urgently requires interdisciplinary collaboration across legal, ethical, and technological fields to establish a copyright governance framework suitable for the AI era.

The new wave of AI development is driving the social sciences toward a transformative “quantum leap.” This transformation is not a mere upgrade of research tools but a systemic restructuring from ontology to methodology. Future social science researchers must possess three core qualities: technological literacy to maintain awareness of tools, humanistic insight to safeguard value rationality, and interdisciplinary integration to navigate complex systems. Throughout this transformation, the dialectical unity of actively embracing technological evolution while upholding a humanistic spirit will be key to the survival and development of the social sciences. When technological systems are able to autonomously generate social science theories, the distinctive value of human researchers will lie in their ability to pose “genuine questions” and maintain ethical awareness in value judgment. This fundamental paradigm shift will ultimately lead to a new form of “AI social sciences” characterized by stronger explanatory and predictive power.

Wang Shouyang is a research fellow from the Academy of Mathematics and Systems Science at the Chinese Academy of Sciences. Guo Dongmei is a professor from the School of Economics at the Central University of Finance and Economics. Li Mingchen is an associate professor from the School of Economics and Management at China University of Mining and Technology.

Edited by ZHAO YUAN