西安交通大学学报（社会科学版）2026，Vol.46Issue(4)：23-34,12.DOI:10.15896/j.xjtuskxb.202604003
社会计算的"三端协同"如何重塑知识生产?
Reshaping Knowledge Production in the Social Sciences:A"Tripartite Synergy"Framework for Computational Social Science
摘要
Abstract
Along with the rapid advancement of information technology,computational social science is comprehensively transitioning from its early"data-intensive"exploratory phase into an"intelligence-intensive"new stage.However,existing research predominantly focuses on conceptual definitions or the application of localized tools,lacking systematic theoretical integration regarding the core technological elements underpinning this paradigm shift—namely,big data,cloud computing,and large language models.This paper aims to construct an innovative"three-terminal synergy"analytical framework,conceptualizing big data as the empirical terminal,cloud computing as the environmental terminal,and large language models as the analytical terminal.It deeply reveals the structural coupling of these three elements at the methodological level and systematically explores how this coupling fundamentally reshapes the logic of knowledge production in the social sciences. Employing a methodology that combines theoretical construction with comparative analysis,this study dissects the fundamental divergences between traditional quantitative paradigms and emerging computational approaches.The traditional paradigm relies heavily on"hypothesis-driven"logic and small-sample structured data,following a top-down deductive pathway.Conversely,the computational paradigm shifts toward a"data-driven"approach utilizing full-sample unstructured data,presenting a bottom-up inductive trajectory.Building on this,the paper elucidates that the maturation of computational social science is the systemic outcome of"three-terminal synergy".Specifically,the empirical terminal reconstructs the empirical object of the social sciences by integrating multi-source heterogeneous data from social networks,mobile trajectories,and IoT sensors,enabling the micro-mapping of macro-social structures through multimodal,holographic tracking.The environmental terminal,leveraging distributed system architectures and elastic underlying computing power,effectively dissolves the physical bottlenecks of storing and processing massive data,providing indispensable physical infrastructure for large-scale panoramic research.Meanwhile,the analytical terminal,centered on large language models,transcends early shallow models.Through cutting-edge capabilities like deep semantic understanding and zero-shot classification,it achieves true"algorithmic democratization",empowering researchers to efficiently handle sentiment computation,intent recognition,and automated theoretical coding of massive unstructured texts.Together,these three terminals form a deeply interactive closed loop of"underlying data supply,computing environment support,and top-level intelligent parsing". However,this paper prudently highlights that while the"three-terminal synergy"vastly unleashes methodological potential,it inherently harbors structural tensions and epistemological crises.At the empirical terminal,the facade of full-sample big data conceals deeper biases stemming from the digital divide and traps studies in the pursuit of superficial correlations at the expense of causality.At the environmental terminal,research barriers created by the computing power divide,high deployment costs,and the data security risks associated with commercial public clouds are becoming increasingly severe.At the analytical terminal,the inherent algorithmic black boxes of complex parameter models,the factual hallucinations of large language models during reasoning,and systemic biases resulting from polluted pre-training corpora pose severe threats to the validity and reliability highly valued in social science research. In response to these multi-dimensional tensions,this paper proposes three pathways that possess both theoretical depth and operational feasibility.First,at the infrastructure level,it calls for national entities or top research institutions to spearhead the construction of autonomous and controllable public cloud platforms and high-quality academic corpora,thereby breaking the data silos and computing monopolies.Second,at the talent cultivation level,there is an urgent need to build an interdisciplinary educational system bridging data literacy,algorithmic logic,and sociological imagination.Third,at the research standardization level,it strongly advocates for localized deployment based on open-source large language models and the integration of mixed methodologies to ensure reproducibility.The core contribution of this paper lies in pioneering the elevation of cloud computing platforms and large language models to the status of independent and equal sociological methodological elements.Future research should seek a dynamic balance between technological propulsion and theoretical guidance within a human-machine collaborative"abductive reasoning"loop,thereby upholding the ultimate pursuit of causal inference and deep mechanism explanation in the social sciences.关键词
计算社会科学/知识生产/科学发现范式/大数据/云计算/AI大模型/算法/人机协同/数据驱动Key words
computational social science/knowledge production/paradigm of scientific discovery/big data/cloud computing/large AI model/algorithm/human-machine collaboration/data-driven分类
社会科学引用本文复制引用
龚为纲,陈斯忆,王天健..社会计算的"三端协同"如何重塑知识生产?[J].西安交通大学学报（社会科学版）,2026,46(4):23-34,12.基金项目
国家社会科学基金项目(22BSH024). （22BSH024）
