Ai and social sciences: new AI, new studies, and new social sciences

The rapid development of Artificial Intelligence (AI) is reshaping the ways in which Social Sciences (SS) conduct observation, modeling, simulation, and governance. At the same time, a series of key challenges faced by AI, including interpretability, controllability, legitimacy, and value alignment, can no longer be resolved through technical pathways alone. This paper argues that the intersection of AI and SS should be understood as a bidirectional co-evolutionary paradigm rather than a one-way application relationship. Based on this view, this paper introduces Parallel Intelligence (PI) and its related theoretical extensions, and proposes an AI × SS framework that connects AI for SS (AI4SS) and SS for AI (SS4AI) within a unified system architecture. Under this framework, the current research landscape is further organized into four key methodological pathways, namely data representation, analysis and modeling, simulation and experimentation, and decision and optimization, and a systematic review is conducted around representative studies along each pathway. The paper then further discusses how these methodological pathways are implemented in representative domains such as finance, communication, management, art, and education. By integrating recent advances in Generative AI, foundation models, Agentic AI, and Computational Social Science, and by incorporating theoretical perspectives from SS on institutions, governance, and human–machine collaboration, this paper characterizes, at the system level, the new research structure emerging from the intersection of AI and SS.