Intelligence Theories and Methods

1083 Research on the Evaluation of the Social Influence of Academic Achievements in Philosophy and Social Sciences in the Open Science Environment: Theoretical Framework and Implementation Path Hot! Chen Ming, Ye Jiyuan DOI: 10.3772/j.issn.1000-0135.2026.08.001 In the open science environment, the degree of openness and transparency of academic achievements has increased, and these achievements have received more attention from society. The evaluation of academic achievements has also shifted from a focus on results to a data- and research process-oriented approach, providing a large amount of data for comprehensively evaluating the social impact of academic achievements in philosophy and social sciences. Based on the development trends of the open science environment, this paper explores evaluation ideas and plans for the social impact of academic achievements. Grounded in the theory of knowledge transfer and transformation, it establishes a theoretical framework for evaluating the social impact of academic achievements in philosophy and social sciences in the open science environment, and explores the implementation paths for such evaluation. The research holds that the evaluation subjects need to be diversified, the evaluation objects need to be ecologically expanded, and the evaluation purpose should take social value orientation as its core. The evaluation indicators are set across three evaluation standard dimensions: the scope evaluation of social impact, the in-depth evaluation of social impact, and the relevance evaluation of social impact. Five to six secondary qualitative and quantitative indicators reflecting the social impact of academic achievements in philosophy and social sciences are each set, respectively. The evaluation method adopts a combination of qualitative and quantitative methods. The evaluation system takes systematicity, operability, and value orientation as its core principles. Through a multi-level system design, it ensures the deep coupling of scientific research value and social needs. 2026 Vol. 45 (8): 1083-1091 [Abstract] ( 38 ) HTML (78 KB)  PDF (1271 KB)  ( 30 )

1092 Research on Measurement of High-order Topology Structure of Patent Cooperation Based on Simplex and Its Association with Innovation Output Hot! Guo Jianming, Yang Alex Jie, Song Xinyu, Liu Qian, Deng Sanhong DOI: 10.3772/j.issn.1000-0135.2026.08.002 In an innovation environment where teamwork and collaboration are deepening, describing the patent cooperation network from the structural level and revealing its influence on innovation output have become important topics in innovation intelligence research. Existing research is mostly based on the perspective of binary relationships; therefore, it is difficult to describe the common group cooperation structure in patent cooperation. This paper introduces a high-order network representation method in algebraic topology, systematically describes the multi-agent interaction structure in the patent cooperation network, and examines the association between its higher-order structural features and inventors’ innovative outputs. Based on more than 7 million patent records in the PatentsView database, this paper constructs a simplex complex model of the patent cooperation network and uses a topological quantization method to measure the high-order structural characteristics of the network. Among them, the number of disconnected components β? is used to characterize the basic connectivity of the cooperative network, and the high-order loop β? is used to characterize the group cooperation mode that cannot be filled by low-order cooperation. On this basis, this study constructs an inventor-application year panel dataset and employs negative binomial fixed-effects regression models to systematically assess how higher-order structural features are associated with multiple dimensions of innovative output, including technological productivity, total patent citations, five-year patent citations, and the number of patent claims. The results show that there is an inverted U-shaped relationship between the number of disconnected components β? in the patent cooperation network and the four innovation output indicators; that is, within a reasonable range, an increase in the number of components is positively correlated with the level of innovation output, but when the number of components is high, the correlation with innovation output shows a weakening trend. The high-order loop β? is positively correlated with the four innovation output indicators, suggesting that higher-order collaborative structures are closely linked to knowledge integration capabilities and technological innovation performance. The robustness analysis shows that the above conclusions are stable in different technical development stages and different technical fields and have strong universality. 2026 Vol. 45 (8): 1092-1105 [Abstract] ( 23 ) HTML (250 KB)  PDF (2904 KB)  ( 32 )

1106 Technical Clustering Based on Patent Synthetic Citations: Case Study in Artificial Intelligence Hot! Gan Jingxian, Tian Chen, Wang Yu DOI: 10.3772/j.issn.1000-0135.2026.08.003 To address the limitations inherent in actual citation data for patent technology topic clustering, specifically data sparsity and inadequate representation of technical relevance, a “Synthetic Citation” construction algorithm based on full-text similarity was designed in this study. Furthermore, by integrating the logic of bibliographic coupling from bibliometrics, “Synthetic Citation Bibliographic Coupling” clustering model was developed. This approach was aimed at reconstructing the technical citation network of patents in a virtual semantic space by mining latent semantic associations. The findings revealed that, compared to actual citation networks, the synthetic citation network effectively captured latent technical associations overlooked by conventional methods, and also significantly enhanced the measurement precision of technical relevance between patents. The empirical results demonstrated that the clustering outcomes obtained using this method can achieve substantial improvements in both the integrity of topic identification and classification accuracy. 2026 Vol. 45 (8): 1106-1125 [Abstract] ( 21 ) HTML (249 KB)  PDF (6727 KB)  ( 28 )

1126 Relationship Between Division of Scientific Labor and Research Performance—the Moderating Role of Geographical Proximities Hot! Lu Chao, Zhou Chenyu, Xiao Chengrui, Zhao Yi, Zhang Chenwei DOI: 10.3772/j.issn.1000-0135.2026.08.004 Because team-based collaboration has become a prevailing mode of production of scientific knowledge, the relationship between division of scientific labor and research performance remains insufficiently understood. Using 148,493 articles from PLoS, we construct the Degree of Division of Scientific Labor (DDoSL) index based on author-contribution statements and the Herfindahl-Hirschman Index (HHI). We use five-year citation counts as a proxy for research performance to examine the association of DDoSL with research performance and explore the moderating roles of multiple proximity dimensions. Research performance is found to be positively associated with DDoSL, and geographical proximity is found to exert a positive moderating effect on the DDoSL-performance relationship. In contrast, social and institutional proximity have no significant association with research performance, and cognitive and organizational proximity do not have any clear moderating effects on the DDoSL-performance relationship. Moreover, positive DDoSL-performance association and positive moderation of geographical proximity are more pronounced in small teams. Overall, the findings suggest that deeper division of scientific labor is associated with higher research performance, while geographic separation may increase coordination costs and attenuate the performance gains from labor division. Thus, small teams may benefit from deepening labor division while minimizing remote collaboration to enhance research performance. 2026 Vol. 45 (8): 1126-1141 [Abstract] ( 22 ) HTML (228 KB)  PDF (2510 KB)  ( 36 )

1142 Identifying Key Technologies in Controlled Games Based on Control Paths in Commerce Control List Hot! Guo Hongmei, Chen Xingyu, Chen Yubei, Zhao Qing, Li Guangjian DOI: 10.3772/j.issn.1000-0135.2026.08.005 The identification of key and core technologies is crucial to national technological security and industrial development. We utilized the U.S. Commerce Control List (CCL) as the core data source. This study designed a technical element extraction agent and inspection evaluation agent to accurately extract technical concepts, attribute features, and various relational patterns such as hierarchical structures, associated control relationships, and technical dependencies from complex texts. The main-path analysis method was employed to identify critical control paths within the control network while considering the key nodes along them as game-controlled critical technologies. An empirical analysis of quantum computer controls provides a novel perspective on the logic behind external controls and key technological focuses, and offers informational references for building an autonomous and controllable technology system in China. 2026 Vol. 45 (8): 1142-1153 [Abstract] ( 20 ) HTML (90 KB)  PDF (5994 KB)  ( 25 )

Intelligence Technology and Application

1154 Research on Medical Information Retrieval Methods Integrating Query Intent and Large Language Models Hot! Han Pu, Liu Senling, Li Xiong, Wang Wei DOI: 10.3772/j.issn.1000-0135.2026.08.006 Medical information retrieval aims to obtain highly relevant information from diverse medical data sources for users’ queries. Traditional information retrieval methods mostly rely on keyword matching or surface semantic similarity modeling, which cannot effectively bridge the semantic gap between colloquial queries and the formal expression of medical knowledge. To address this issue, this paper proposes a medical information retrieval method that integrates query intent with a large language model. First, to enhance the model’s ability to represent medical texts, this paper injects task knowledge into the Chinese-RoBERTa-wwm-ext model based on the recognition of mainstream Chinese medical query intents to accurately identify the user’s query intent. Second, using the intent recognition results, the large language model is guided to perform semantic optimization and reconstruction of the original query to improve the accuracy and stability of the query optimization of the large language model. Third, to avoid losing the original semantics and enhance the semantic information, the original query and the optimized query are encoded and retrieved, respectively, and the top-k candidate documents are obtained through the weighted fusion of similarity scores. Finally, the original query and the optimized query are combined to form a joint representation, which is input into a cross-encoder to re-rank the candidate documents and obtain the final retrieval result. Experimental results on the public datasets CmedqaRetrieval and DXYDiseaseRetrieval show that the proposed method can significantly improve the accuracy and relevance of medical information retrieval and exhibits good generalization and robustness across four large language models. 2026 Vol. 45 (8): 1154-1165 [Abstract] ( 24 ) HTML (161 KB)  PDF (3053 KB)  ( 24 )

1166 Efficient Identification of Mediation Effects Using Exhaustive Meta-analysis Hot! Yang Yang, Lin Weijie, Zhou Wenjie, Wei Zhipeng, Yang Kehu DOI: 10.3772/j.issn.1000-0135.2026.08.007 Quantitative analysis methods in social science research do not adequately evaluate mediation mechanisms between variables. From an evidence-based perspective, this study integrates the difference-in-coefficients (c-c′) approach to mediation identification using an exhaustive meta-analytic framework to address the effect size comparability and endogeneity inherent in conventional mediation tests and their meta-analyses. We propose a complete framework that generates standardized, comparable contextual data by exhaustively combining control variables based on the mathematical logic that the total effect is the sum of the direct and indirect effects and identifies the mediation effects by conducting a meta-regression on the difference (c-c′). The empirical results obtained using information poverty and financial analytics as the contexts for verifying robustness demonstrate that this method enhances the internal validity by avoiding the endogeneity risks associated with conventional estimates of the path coefficients (a, b). Additionally, it enhances the external validity by simulating diverse control settings to evaluate the stability of the mediation effects across the contexts. By bridging these two dimensions, this method improves the internal and external validity of the identified mediation effects and enriches the methodological toolbox for detecting and testing complex mechanisms in the social sciences. 2026 Vol. 45 (8): 1166-1177 [Abstract] ( 24 ) HTML (202 KB)  PDF (1229 KB)  ( 10 )

Intelligence Users and Behavior

1178 Research on Herd Effect Identification in Online Public Opinion Events Hot! Shen Wang, Li Xin, Yan Zhipeng, Li He DOI: 10.3772/j.issn.1000-0135.2026.08.008 This study focuses on identifying the herding effect in online public opinion events, aiming to systematically construct a multidimensional quantitative identification method for it and further explore the dynamic evolution of this effect. First, this study analyzes and deconstructs the herding effect in online public opinion. Then, it constructs a dynamic user network that integrates semantic features and emotional intensity, and employs the Louvain-Blondel algorithm for dynamic community detection among users. Subsequently, it systematically identifies the herding effect across four dimensions: the explosiveness of dissemination, the influence of dominant communities, group aggregation, and the convergence of information content. Empirical research shows that the Herding Index follows a dynamic trajectory of “rapid rise-fluctuation-resurgence-decline,” with changes in modularity serving as a key driving factor. Specifically, in the outbreak stage of public opinion, the modularity decreases significantly, and the dominant community is formed rapidly to promote the aggregation of opinions; at the point of major information disclosure, the decline in modularity again can trigger a nonlinear resurgence of the herd effect, indicating that external stimuli have a significant activating effect on group behavior. In terms of validity verification, the herd effect index is highly synchronized with the change in the number of comments (Pearson correlation coefficient r = 0.9172, p < 0.001). Exploratory Factor Analysis (EFA) demonstrates that the four indicators had good construct validity and can effectively measure the core characteristics of the herding effect. The fitting and prediction results of the grey prediction model (GM (1,1)) show that the index has good trend response and prediction capabilities in different stages, especially showing high prediction accuracy in the stable evolution stage, which verifies the effectiveness of the identification method. 2026 Vol. 45 (8): 1178-1191 [Abstract] ( 34 ) HTML (163 KB)  PDF (2674 KB)  ( 29 )

1192 Impacts of Task Types on Generative Artificial Intelligence Interaction Outcomes from the Emotional Experience Perspective: Evidence from DeepSeek Hot! Zhang Min, Zhang Dongxin, Zhang Ke, Qin Kenan, Yang Han DOI: 10.3772/j.issn.1000-0135.2026.08.009 From the emotional experience perspective, exploring the impact of task types on the interaction outcomes of generative artificial intelligence (GAI) is highly significant for enhancing users’ intelligent interaction behavior and improving the design and service quality of human-AI interactions. This study uses generalized structural equation modeling, GAI-assisted coding, and interpretable machine learning with shapley additive explanations to conduct an in-depth analysis of 5560 app reviews related to DeepSeek’s text-generation application. The findings reveal that users predominantly use DeepSeek for instrumental tasks, followed by emotional and creative tasks. Both incidental emotions and overall satisfaction perform better in emotional and creative tasks, and task attributes significantly affect overall satisfaction, with high emotional exposure tasks more likely to elicit positive emotions and higher satisfaction. Moreover, incidental emotions exert a cumulative effect on overall emotions; however, the two differ in formation mechanisms: incidental emotions are shaped by multiple technical factors, whereas overall satisfaction is mainly constructed through functional features and emotional resonance. This study highlights that the GAI interaction design should prioritize emotionally intelligent responses across task types while guiding users to adjust their interaction strategies based on task characteristics, thus systematically improving the overall quality of interaction experiences. 2026 Vol. 45 (8): 1192-1205 [Abstract] ( 30 ) HTML (155 KB)  PDF (2344 KB)  ( 42 )

Intelligence Reviews and Comments

1206 A Review of Scientists' Interdisciplinary Research Behavior Hot! Chen Yifan, Hu Wei, Xie Ruixia, Cui Guirong, Zhang Zhiqiang DOI: 10.3772/j.issn.1000-0135.2026.08.010 Identifying the common behavioral characteristics of scientists engaged in interdisciplinary research provides essential support for optimizing China’s interdisciplinary talent cultivation system, thereby improving the efficiency of interdisciplinary collaboration within research institutions. In this study, we adopted a behavioral science perspective to define all observable and measurable purpose-driven scientific activities that scientists undertake during interdisciplinary research as “Interdisciplinary Research Behavior” (IDRB). Drawing on the classic theoretical paradigms of the behaviorist school, we constructed a systematic analytical framework encompassing three stages: antecedents, processes, and outcomes. Accordingly, two types of actors—individual scientists and research teams—were distinguished, and the current state of IDRB research at each stage was systematically reviewed, synthesizing the existing findings and proposing future directions. The analysis revealed several shortcomings: Chinese research tends to be narrow in scope, emphasizing macro-level measurements while lacking an in-depth exploration of micro-level mechanisms; studies on top scientists’ IDRB lack comprehensiveness; a causal explanation chain for the dynamic process of IDRB is yet to be established; and interdisciplinary measurement indicators require further refinement in terms of stability and innovation. Future research should address these gaps through a more in-depth analysis. 2026 Vol. 45 (8): 1206-1222 [Abstract] ( 22 ) HTML (210 KB)  PDF (1967 KB)  ( 35 )