[Purpose/significance] The rapid development of Generative Artificial Intelligence (GAI) technology, cou⁃pled with intensifying technological competition between China and the United States, has posed severe challenges to user privacy and data security. The study compares the privacy policies of GAI platforms in both countries, aiming to analyze their characteristics and differences, and provide useful insights for the sustainable and healthy development of AI applications in China. [Method/process] The study selects 20 representative privacy policy texts from GAI plat⁃forms in China and the U.S., employing LDA topic modeling and the PMC index model for quantitative analysis.Through topic extraction and consistency evaluation, it compares the strengths and features of privacy policies in both countries to inform policy optimization in China. [Result/conclusion] The findings show that Chinese platform policies
emphasize government-led design and full-process regulation, while U.S. policies focus more on market orientation, us⁃er autonomy, and legal remedies. Based on this, the study suggests that Chinese AI platforms should shift from a "defen⁃sive compliance" approach to "proactive rule-making", strengthen data lifecycle management, build a sound regulatory framework, and improve mechanisms for protecting user rights and addressing complaints, thereby enhancing the effec⁃tiveness of privacy governance and platform competitiveness.

[Purpose/significance] Reveal the generation mechanism of privacy leakage risks among users of medical e-commerce platforms and propose corresponding control strategies to enhance user privacy security in the medical ecommerce sector and safeguard user privacy rights. [Method/process] In-depth interview is conducted to collect ini⁃tial data. Using thematic analysis, the MOA and SOR frameworks are integrated to extract the influencing factors of pri⁃vacy leakage risks in medical e-commerce. The interrelations among these factors are mapped out, and a model for the generation and control mechanism of these risks is constructed. [Result/conclusion] Privacy disclosure motivation, pri⁃vacy protection environment, and privacy literacy, as stimuli, all indirectly influence users' privacy risk control behav⁃iors through the mediation of perceived privacy leakage risk. Furthermore, users' privacy cognitive psychology impacts their perception of privacy leakage risk, moderating the relationship between stimuli and perceived privacy leakage risk, as well as between perceived risk and risk control behaviors. Given this, the research proposes a range of data gov⁃ernance and privacy protection strategies from the dual perspectives of government and platform, including establishing a full lifecycle privacy protection mechanism and implementing graded labeling of privacy information.

编者按：2025年底，中国人民大学书报资料中心信息资源管理系列刊、图情档39青年学者沙龙与黑龙江大学信息管理学院共同发起了“2025年度中国信息资源管理学界学术‘热点’与‘冷点’”评选活动。40余家专业期刊媒体联合发布了本年度信息资源管理领域的学术“热点”与“亟待加强的研究议题”（“冷点”）评选结果。该活动围绕本领域年度学术前沿动态与潜在议题展开系统梳理与评议——其中，“热点”聚焦于领域内受到集中关注的研究趋势，而“冷点”则指向那些值得深入研究，但未引起学界充分重视，或未能深入研究的涉及知识体系建构的学科基础、深层问题，或社会实践需要理论支撑、解决方案的重要议题。“热点”彰显时代关切，“冷点”关乎学科根基，“冷”“热”交融共进，共同为中国自主的信息资源管理知识体系建构注入新动能与新活力。

[Purpose/significance] The conceptual connotation of the information cocoon, which is both tangible and in⁃tangible, has led to the bottleneck of inconsistent research conclusions and inability to integrate research results in quantitative methods based on conceptual descriptions. Therefore, abstracting the concept of information cocoon and gradually improving the quantitative methodology system becomes a reasonable choice that conforms to the laws of sci⁃entific development. [Method/process] Starting from the important issue of analyzing the form of information cocoons,this article reviews and summarizes the research results at home and abroad. From the perspectives of metaphorical con⁃cept setting, the possibility of multidimensional forms, ecological fallacies and reductionism in empirical research, and the inability to track the "behavior" of social media users, the reasons for the "formal but intangible" nature of informa⁃tion cocoons and related issues in the field are sorted out, and the necessity of abstracting the concept of information co⁃coons is explained. On this basis, based on the process characteristics of information cocoon formation, a conceptual ab⁃stract model of information cocoon is constructed using complex network domain knowledge, and the method of quanti⁃fying information cocoon through complex network domain knowledge is interpreted from three levels: selection homog⁃enization, content homogenization, and group homogenization. [Result/conclusion] This conceptual abstract model aims to innovatively achieve the first abstraction of the concept of information cocoon based on the connotation of the concept of "being tangible but intangible", forming a scientific, stable, and strongly generalized quantitative method for information cocoon.

[Purpose/significance] This article designs targeted scene governance strategies to respond to complex gov⁃ernance needs. It improves the refinement and scenarization level of the governance of deepfake events public opinion information. It breaks through the limitations of the vertical depth and adaptation accuracy of existing governance strate⁃gies. [Method/process] By integrating risk management theory and scenario theory, this article takes the perception of public opinion information risk in deepfake events as the dependent variable, and forgery technology, public aware⁃ness, media dissemination, degree of harm, scope of impact, response capability, and regulatory norms as independent variables. It has constructed a model of influencing factors and causal driving factors for the perception of public opin⁃ion information risk in the AIGC era of deepfake events. [Result/conclusion] In personal privacy events, none of the variables constitute a necessary condition for high-risk perception. This article has obtained 16 configuration paths for
high-risk perception of public opinion information in deepfake events. It has refined three governance models: risk identification scene, risk assessment scene, and risk response scene, and proposed scene governance strategies for deepfake events public opinion information.

[Purpose/significance] False short videos integrate multimodal content such as images, text, and audio,with their varying content structures potentially influencing the dissemination of short videos by manipulating user trust. Investigating the impact of multimodal content structure manipulation of false short videos on users′ sharing inten⁃sion aids in identifying false short videos and enhances platforms′ capacity to govern false information. [Method/pro⁃cess] Drawing upon SOR theory, this study constructed a research model in the context of false short videos: "informa⁃tion structure-perceived credibility-user attention-user sharing intention". Data was collected through a mixed-meth⁃ods approach combining questionnaires, eye-tracking experiments, and semi-structured interviews. [Result/conclu⁃sion] Users directed greater attention towards text areas than image areas, whilst exhibiting greater variability in atten⁃
tion distribution within image areas. Based on the structural manipulation of false short videos, this study found that the primary-secondary and parallel relationships exert a significant positive influence on perceived accuracy and consis⁃tency, while causal relationships exert a significant negative influence on perceived accuracy. Perceived accuracy ex⁃erts a significant positive influence on text fixation counts, image fixation counts, and image fixation duration. Text fixa⁃tion counts, image fixation counts, and image fixation duration exert a significant positive influence on users′ sharing in⁃tension.

[Purpose/significance] Analyzing the complexity of network public opinion in the context of generative arti⁃ficial intelligence and exploring the risk factors of intelligent generation network public opinion are of great signifi⁃cance for the governance and decision-making of network public opinion risks. [Method/process] This article uses the WSR methodology to construct an indicator system for network public opinion risk elements in three dimensions: physi⁃cal, rational, and human. At the same time, it combines fuzzy DEMATEL and AHPsortⅡ methods to measure and ana⁃lyze network public opinion risk, and uses the Chengdu demolition rumor incident as an empirical case for in-depth ex⁃ploration. [Result/conclusion] Based on empirical data results, this article proposes targeted risk management strate⁃gies for network public opinion from three levels: physical, physical, and human, aiming to provide valuable references for relevant departments and the public to enhance their ability to respond to network public opinion risks.

[Purpose/significance] Exploring quality improvement strategies adapted to the characteristics of data fac⁃tors from the perspective of multiple stakeholders, ensuring the circulation and application of high-quality data factors,thereby promoting the high-quality development of the data factor market. [Method/process] From the perspective of market self-governance, this study investigates the strategic choices of the data trading platforms, data suppliers, and data demanders under changing costs and benefits. By constructing an evolutionary game model, the study analyzes the interactions among the three stakeholders and their influencing factors. Finally, MATLAB software is used to simulate the dynamic process of strategy evolution under different conditions. [Result/conclusion] The active regulation of data trading platforms is influenced by both explicit and implicit benefits and regulatory costs. The decision-making of data suppliers regarding the provision of high-quality data depends on cost differentials as well as the platform's incentive and penalty mechanisms. The feedback costs and incentive intensity are key factors affecting the feedback behavior of data demanders. Effective platform regulation, a reasonable incentive and penalty mechanism, and the feedback from data demanders together contribute to a virtuous cycle that promotes the circulation of high-quality data elements.

[Purpose/significance] The information cocoon is a stubborn ailment that must be cracked in the digital era. Information acquisition through generative artificial intelligence can significantly enhance users' information-collection efficiency and help expand information; yet it also harbors the risk of information narrowing, which may lead to cognitive closure. Its latent double-edged-sword effect urgently demands systematic exploration. [Method/process] From an individual perspective, based on cognitive-offloading theory and integrating the elaboration likelihood model with dual-process theory, this study proposes the positive and negative effect paths of "breaking the cocoon" and "spinning the cocoon", constructs a model according to theoretical hypotheses, and conducts empirical testing using PLS-SEM. [Result / conclusion] Generative artificial intelligence information acquisition presents a double-edged-sword effect on information cocoons; the cognitive offloading triggered by generative artificial intelligence information acquisition can, by stimulating users' self-initiated behavior, broaden information boundaries and break cocoons, yet it can also, by fostering behavioral inertia, exacerbate information narrowing and create cocoons; critical thinking, as a key moderating variable, can weaken the negative effect and curb cocoon formation.

[Purpose/significance] Generative AI search engines are becoming the most frequently used and most dis⁃cussed search platforms. However, there remains limited understanding regarding the effectiveness of their system de⁃sign, their ability to support diverse task types, and their capacity to enhance search outcomes for individuals with dif⁃ferent learning styles. This study addresses these gaps to supplement research on search behavior and effectiveness within generative AI search environments in the field of interactive information retrieval. [Method/process] Based on a hierarchical interaction model, we used the New Bing website as our experimental system to investigate the relation⁃ship between system interactions (information resources, page presentation, and technical characteristics) and search effectiveness. We also examined how different task types (factual and exploratory) and learning styles (divergent, con⁃vergent, assimilative, and accommodative) influence user search behavior and search outcomes. [Result/conclusion] Most system interaction metrics correlate positively with search effectiveness at the user level, while cost- benefit search effectiveness relates solely to page layout and typography. Participants demonstrated richer search behaviors and superior search outcomes during factual tasks. Furthermore, significant differences in search behaviors emerged across the four learning styles, primarily reflected in query restructuring frequency and types. Search effectiveness vari⁃ations were equally pronounced, with participants exhibiting divergent learning styles reporting superior experiences in both quality satisfaction and perceived value.

[Purpose/significance] This study aims to clarify the concepts and concept relations between data manage⁃ment and data governance, address the current problems of confusion and misapplication of concepts and provide refer⁃ences for building and improving standard systems for data management and data governance in China, facilitating the adoption and appropriate application of international standards and promoting China's active participation in the formu⁃lation of international rules and standards in the digital field. [Method/process] This paper adopts the text content analysis method and analyzes the definitions of data management and data governance from online databases of ISO,IEC, and ITU-T, in accordance with principles of concept system building in ISO 704:2022. From these definitions, the concept objects, concept characteristics, and types of concept relations of data management and data governance are identified, respectively, and core concept relationship diagrams are constructed. The similarities and differences of con⁃
cepts between the two terms are compared across four dimensions: concept objects, concept characteristics, concept re⁃lations and application levels. [Result/conclusion] Data management involves micro and meso levels and encompasses comprehensive activities including data processing, processes management and value realization; data governance cov⁃ers micro, meso, and macro levels, emphasizing quality management, definition of rights and responsibilities, strategic coordination and policy formulation; both share common concerns about process elements and the realization of data value, presenting an associative relation.

[Purpose/significance] This study aims to systematically explore the impact of the group effects of push pullfactors on individual digital hoarding behavior. [Method/process] Drawing on push-pull theory and the MOA framework, this study developed a research model encompassing multi-dimensional factors such as emotional attach⁃ment, fear of missing out, intolerance of uncertainty, technology empowerment, digital information characterization, and social impact. By combining Necessary Condition Analysis (NCA) and Fuzzy-Set Qualitative Comparative Analysis (fsQCA), this study identified the core pathways and grouping patterns influencing digital hoarding behaviors. [Result/conclusion] The formation of individual digital hoarding behavior is jointly influenced by push-pull factors, with three
main pathways identified: the pull-dominant opportunistic motivation type, the push-induced motivational stress type,and the push-pull synergistic capacity-supporting type. Accordingly, this study proposes promoting digital environ⁃ment governance through a synergistic approach encompassing technological regulation, cognitive adaptation, and emo⁃tional guidance, thereby achieving the dual objectives of individual psychological adjustment and information environ⁃ment optimization.

[Purpose/significance] In response to the strategic demand for building autonomous disciplinary knowl⁃edge systems, this study aims to systematically review the research status and evolutionary trajectory of digital library user experience (UX). The goal is to construct a knowledge system for this specific field, thereby providing a micro-lev⁃el foundation and a practical case for the development of the parent discipline's autonomous knowledge system. [Meth⁃od/process] Following the PRISMA guidelines, a systematic literature review was conducted. A total of 81 documents were selected as the analytical sample from three major Chinese databases. The included literature was systematically coded for research themes, fundamental concepts, research paradigms, theoretical models, and evaluation frameworks.[Result/conclusion] This study identifies a three-stage developmental trajectory in digital library UX research and proposes a three-dimensional analytical framework for its research paradigms. Furthermore, it constructs a comprehen⁃sive knowledge system for digital library UX, centered on the core path of "value-driven - status analysis - practical approach". This work offers a novel pathway for the construction of disciplinary knowledge systems.

[Purpose/significance] Investigating the pathway of user trust in deepfake videos can mitigate blind trust and provide theoretical support for the government and social media platforms to formulate deepfake risk governance strategies. [Method/process] The meta-analysis was used to determine the 12 elements affecting user trust in deep⁃fake videos. The fuzzy ISM and the MICMAC were combined to analyze the correlation pathways and classification rela⁃tionships of user trust in deepfake videos. [Result/conclusion] The pathways to user trust are categorized into literacydriven pathway, platform-environment dependency pathway, and information assessment pathway. Cognitive ability,awareness of deepfakes, and informational cues constitute foundational factors exerting the most profound influence. In⁃ternalization of personal literacy, optimization of the information ecosystem, and reinforcement of external cues are key dimensions for reducing user trust in deepfake videos.

[Purpose/significance] The rapid advancement of large language models（LLMs）has transformed tradition⁃al human-computer interaction. Investigating the characteristics of human-AI conversational interactions among uni⁃versity students in proactive mental health contexts can enhance our understanding of the complexity and diversity of user behaviors in such scenarios. [Method/process] Data were collected using the mobile experience sampling method (mESM). A total of 32 subjects were recruited for a two-week experiment on human-AI conversational interactions.Longitudinal data tracking was implemented through structured diary logs, while participants′ mental health status was assessed before and after the experiment using validated psychological scales. The experimental data were processed and analyzed through a combined approach of open coding and conversation analysis. [Result/conclusion] The study systematically examines the content characteristics of human-AI conversational interactions through three dimensions:user query features, AI response features, and AI interpretability features. It further analyzes behavioral characteristics through conversational structure features and "query-response" behavioral patterns, while also organizing features re⁃lated to mental health dimensions. Additionally, the study proposes three distinct modes of human-AI interaction: diag⁃nostic-consultative interaction, affective-interactive engagement, and cognitive-collaborative coordination.

[Purpose/significance] Exploring multimodal evidence reorganisation and association patterns from a re⁃source management perspective facilitates the formation of an evidence-based logic loop of "evidence fragments—evi⁃dence chains—factual verification", driving the transformation of information services toward data-driven evidencebase dapproaches. [Method/process] First, following the digital humanities research process, this study proposes an evidence-based framework for Hanfu digital resources based on knowledge graphs. Second, this study ensures the reli⁃ability of the evidence-based process and the verifiability of results through four aspects: evidence extraction, evidence association, cross-modal evidence fusion, and multi-source mutual verification of conclusions. Finally, using Song Dy⁃nasty women's clothing resources as an example, this study conducts evidence chain construction and scenario applica⁃tions to validate the scientific validity and feasibility of the proposed evidence-based framework. [Result/conclusion]By systematically analysing the eight evidence categories, five evidence levels, and seven evidence association patterns in the evidence-based context of Hanfu literature, it can achieve a pathway integrating "data preparation—evidence or⁃ganisation—conclusion verification", thereby validating the scientific validity of the evidence-based framework for Hanfu digital resources mentioned in this paper. This provides methodological support for evidence-based research on cultural heritage resources, with Hanfu evidence-based research as a representative example.

[Purpose/significance] Classifying research topics into their respective disciplines is fundamental for inter⁃disciplinary research, such as measuring interdisciplinary integration and identifying cross-disciplinary themes. Only by determining the disciplinary category of each research topic can we assess whether it represents an interdisciplinary intersection. [Method/process] This study proposes a framework for classifying research topics into disciplines using large language models. Building on base models such as Llama3-8B-Instruct, Qwen2.5-7B-Instruct, and DeepSeek-R1-Distill-Qwen-7B, a two-stage optimization strategy of "domain-adaptive pretraining + supervised fine-tuning" was implemented. A total of 116192 academic papers were used for domain-adaptive pretraining to enhance the model′s semantic understanding of scientific literature. Author keywords were used to represent research topics, and a manual⁃ly annotated dataset of 126919 "research topic-disciplinary label" pairs was employed for supervised fine-tuning to op⁃timize the model′s classification performance. [Result/conclusion] While large language models possess zero-shot dis⁃ciplinary classification capabilities, their precision and F1-scores remain below 50% when relying solely on prompt de⁃sign, which is insufficient for practical applications. In contrast, the proposed framework achieves a precision of 93.61% and an F1-score of 83.09%, significantly improving the accuracy of disciplinary classification for research top⁃ics.

[Purpose/significance] Artificial intelligence is profoundly driving the paradigm transformation, content op⁃timization, and efficiency leap of public digital cultural services. By constructing vertical domain AI agents, an attempt is made to provide a feasible path for enhancing the efficiency of public digital cultural services through the empower⁃ment of artificial intelligence. [Method/process] Point out the specific aspects of how vertical domain AI agents em⁃power the efficiency improvement of public digital cultural services, design the construction path of vertical domain AI agents for public digital cultural services, and conduct empirical tests on the proposed path by taking the construction of vertical domain AI agents for digital cultural services at the 2025 Asian Winter Games in Harbin as an example. [Re⁃sult/conclusion] Through empirical tests, it is proved that the construction path and application scenarios of the verti⁃cal domain AI agent for public digital cultural services proposed in this study are feasible, thereby exploring an effec⁃tive solution for enhancing the efficiency of public digital cultural services through artificial intelligence based on the construction of vertical domain AI agents.

[Purpose/significance] As generative artificial intelligence (GAI) grows more dependent on training data,super platforms have used their structural advantages to dominate model training, creating a "platform-model" symbiot⁃ic relationship. However, existing research has paid little attention to the operational mechanisms of platform control and the associated governance dilemmas within the training data phase. [Method/process] Using normative analysis and institutional comparison, this paper dissects the unique power structure and risk types inherent to the "platformmodel" symbiosis and reveals their risk formation and governance pathways. [Result/conclusion] The study finds that training data risks under this symbiotic relationship are endogenous and amplified, challenging traditional regulatory models. To address this, it proposes a multi- stakeholder collaborative governance system featuring: a principled,tiered, government-led source documentation framework to manage privacy leakage and data crossover risks; a verifi⁃able platform self-assessment system to dismantle governance barriers; and superplatform-led construction of trusted data spaces to mitigate data circulation risks.

[Purpose/significance] In the context of sudden emergency events, content on social media poses severe challenges to online environments and emergency management due to the amplification effect of public opinion dissemi⁃nation. Traditional detection methods suffer from issues such as imbalanced classification data and low detection accu⁃racy, necessitating efficient solutions. [Method/process] This study constructs a dedicated dataset of content from Wei⁃bo comments in sudden emergency events. Generative Artificial Intelligence (GenAI) technology is employed to gener⁃ate semantically equivalent pseudo-toxic content through few-shot prompt learning, balancing the sample distribution.Furthermore, the MACBert-Att model is proposed by integrating the MACBert pre-trained model with a global atten⁃tion mechanism, enhancing the semantic capture capability for domain-specific terms and emotional expressions. [Re⁃sult/conclusion] Experiments demonstrate that with GenAI data augmentation, the MACBert-Att model achieves an F1 score of 0.95, representing a 15% improvement over the baseline Bert model and significantly outperforming tradi⁃tional augmentation methods like SMOTE. This validates the collaborative effectiveness of GenAI-based semantic-lev⁃el data augmentation and the model architecture.

[Purpose/significance] The unstructured and highly context-dependent nature of emotional information
poses challenges to traditional intelligence processes. Its deep integration with the emotional labor of practitioners
makes the synergy between "value extraction, labor expenditure, and ethical compliance" a core dilemma. Systematical⁃
ly resolving this dilemma is crucial for enhancing intelligence work efficiency. [Method/process] This study construct⁃
ed an "Emotional Information Intelligence Governance Loop" (EIG-Loop) model that incorporates emotional labor regu⁃
lation, based on intelligence cycle theory. Empirical validation was conducted through a mixed-method study involving
30 open-source intelligence analysts and 400 commercial customer service representatives. [Result/conclusion] The
research identified a four-stage information chain fracture mechanism ("collection-processing-decision-feedback") in
traditional processes and confirmed a significant positive correlation between information entropy and emotional labor
intensity. The EIG-Loop model, via its dual-closed-loop design of "information flow-labor regulation", effectively im⁃
proved intelligence conversion efficiency, reduced emotional labor intensity, and enhanced ethical compliance.

[Purpose/significance] This study aims to clarify the core impact factors and pathways in data value realiza⁃tion, thereby promoting the effective release and realization of data value. [Method/process] A multi-chain system for data value realization is constructed. The CFCS-DEMATEL method is employed to identify causal relationships among factors, and fuzzy cognitive maps are used to simulate single-chain, double-chain, and triple-chain models to identify core factors within and across chains and analyze key impact paths. [Result/conclusion] A single-chain provides limit⁃ed momentum; a double-chain improves data flow efficiency; and triple-chain collaboration significantly enhances val⁃ue realization performance. The core impact factors differ across the single-, double-, and triple-chain models, and the impact paths within and across chains vary with the stage of data flow. Data value realization can be improved by strengthening single-chain capabilities, enhancing cross-chain collaboration, and advancing triple-chain collabora⁃tive governance.

[Purpose/significance] Addressing critical needs in digital economy development, this study establishes a theoretical framework for data trading platforms and investigates their construction modes, providing theoretical and practical guidance for platform operations. [Method/process] Applying the LDA model to policy texts for topic model⁃ing, this study constructed the framework of data trading platforms, followed by comparative case studies of representa⁃tive government-led and enterprise-led platforms through systematic platform investigations. [Result/conclusion]This study established a five-in-one theoretical framework for data trading platforms. Government-led platforms pay more attention to platform ecological construction and compliance requirements, which adopt the constructing mode of "policy orientation+ecological synergy+normative guidance". Enterprise-led platforms focus on flexible service and value realization, which adopt the construction mode of "flexibility and autonomy+precision marketing+diversified ser⁃vice". The two types of platform construction basically conform to the policy orientation. The government-led platforms should strengthen technological innovation, promote data supply and enhance market-oriented operation ability. Enter⁃prise-led platforms should deepen cooperation, improve compliance mechanism and strengthen whole process manage⁃ment.

[Purpose/significance] Conflict health information is an important interfering factor in the health informa⁃tion dissemination for the new elderly people. Clarifying the influencing mechanism and role path of conflict health in⁃formation on resistance to persuasion among the new elderly people has important reference value for designing effec⁃tive health information dissemination intervention measures. [Method/process] Based on the person-environment fit theory, this study constructed a multiple serial mediation model to examine the effects of conflict health information on resistance to persuasion. Empirical analysis was conducted through a 2 (conflict health information: conflict vs. nonconflict)×2 (evidence type: non-narrative vs. narrative) between-subjects design, while multi-group analysis and posthoc test was employed to explore the differential effects across different evidence types. [Result/conclusion] Conflict health information indirectly influences resistance to persuasion among the new elderly people; conflict health informa⁃
tion positively affects the information supplies-information needs mismatch and the information demands-information abilities mismatch, and two types of mismatches both result in cognitive dissonance and perceived low efficiency among the new elderly people, which in turn result in resistance to persuasion; compared with non-narrative conflict health in⁃formation, narrative conflict health information result in stronger resistance to persuasion among the new elderly peo⁃ple, and the post-hoc test further showed that narrative conflict health information directly result in active resistance among the new elderly people.

[Purpose/significance] The integration of the digital economy and generative AI has exacerbated the trust crisis in data circulation, making the construction of a trusted data space a key path to break through the predicament.Systematically analyze libraries, archives, and data intermediaries institutions′ differences and complementarities are of significant importance for optimizing disciplinary practices and building a collaborative governance system. [Meth⁃od/process] This paper centers on the construction of trusted data space within the realm of information resource man⁃agement. It compares the differences among libraries, archival institutions, and data intermediaries institutionsin as⁃pects including data trustworthiness, data governance maturity, data circulation efficiency, and social credibility. [Re⁃sult/conclusion] Three types of subject institutions have varying degrees of differences in the standards for building a trusted data space. The governance strategies proposed include leveraging disciplinary advantages to make up for tech⁃nical deficiencies, establishing a division of labor and collaboration mechanism centered on scenarios, and constructing a governance framework that is incentive-compatible and dynamically coordinated. These strategies provide a theoreti⁃cal framework and path selection for the construction of a trusted data space in the AIGC era.

[Purpose/significance] This study aims to reveal the characteristics and influencing factors of mobile APP users' privacy risk retention behavior from the perspective of privacy uncertainty, explaining the paradoxical phenome⁃non where users choose to bear losses caused by privacy risks rather than actively adopt privacy protection measures. It contributes to strengthening user privacy security protection and optimizing the mobile application ecosystem. [Meth⁃od/process] Based on grounded theory method, this study explores the characteristics and influencing pathways of mo⁃bile APP users' privacy risk retention behavior. [Result/conclusion] Under the perspective of privacy uncertainty, mo⁃bile APP users' privacy risk retention behavior manifests as three characteristics: active privacy concession, default pri⁃vacy permission, and forced privacy compromise. Such behavior leads to intensified risks, risk addiction, and cognitive rigidity. At the user level, platform level, privacy benefit level, and social environmental level, ten factors influence the occurrence of privacy risk retention behavior. User privacy uncertainty plays a mediating or moderating role in the path⁃ways through which these factors affect privacy risk retention behavior.

[Purpose/significance] This study aims to develop an automated prediction framework for identifying poten⁃tially high-impact papers by integrating multidimensional feature index with deep learning methods, enhancing the comprehensive characterization and accurate prediction of academic value and impact, thereby providing methodologi⁃cal support and technical pathways for optimizing research evaluation, allocating academic resources, and informing science and technology decision-making. [Method/process] In order to realize the accurate prediction of potentially high-impact papers in the literature of social science field, this study firstly constructs a three-dimensional feature in⁃dex system of scientific and technological papers from the three dimensions of the papers' own features, content fea⁃tures and cited features. Then, this study designs and realizes the automatic prediction framework of potentially highimpact papers, which integrates the three-dimensional feature index of papers and deep learning model, to realize the deep learning prediction of potentially high-impact papers in the massive literature from the major social sciences dis⁃cipline named economics. Finally, the differences and advantages in features of potentially high-impact papers are sys⁃tematically compared and analyzed. [Result/conclusion] There are significant differences among various dimensional index in scientific and technical literature. The constructed three-dimensional feature index system and feature vector space, combined with deep learning prediction models, can comprehensively measure the value and influence of scien⁃tific and technological papers. At the same time, the superior prediction effect can promote the automatic prediction and recommendation application of potentially high-impact papers in massive literature. Artificial neural networks and TabNet perform well in prediction accuracy and precision, but are inferior to traditional machine learning models in metrics such as recall, P-R area, and AUC value. Using citation features of papers outperforms using the paper's own features or content features to predict high impact papers. Potentially high-impact papers exhibit significant advantag⁃es in multidimensional feature index, including papers' own features, thematic features, citation dynamics, etc.

[Purpose/significance] By constructing a multi-level driving model, this study systematically analyzes the
topological relationships and action pathways of driving factors influencing mobile short video users' digital detox be⁃
haviors, aiming to provide a research paradigm with both theoretical explanatory power and practical applicability for
the digital health ecosystem. [Method/process] First, grounded theory analysis was employed to identify key factors af⁃
fecting users' digital detox behaviors. Second, the ADSM was applied to categorize these factors hierarchically and ex⁃
plore their transmission paths, revealing interrelationships among them. Finally, the MICMAC algorithm was used to
classify the influencing factors into clusters and validate the effectiveness and reliability of the constructed model. [Re⁃
sult/conclusion] The results indicate that the counter-dependency model of mobile short video users' digital disen⁃
gagement behaviors can be divided into seven hierarchical levels. The twelve driving factors are further classified into
three clusters: dependent clusters, autonomous clusters, and driving clusters. Among them, self-regulation capability is
identified as the most fundamental driver influencing digital detox behaviors.

[Purpose/significance] Aiming at the problems of single feature dimension and insufficient interpretability in the early identification of current breakthrough papers, this study constructs an interpretable machine learning meth⁃od to improve the identification accuracy and logical transparency, providing methodological support for scientific re⁃search management and innovation layout. [Method/process] Firstly, starting from the connotation of breakthrough pa⁃pers, on the basis of traditional features, innovation attribute measurement is introduced, knowledge innovation vectors are obtained by inducing large language Prompts, and a multi-dimensional feature system of breakthrough papers is constructed; secondly, an information expression system is built based on formal concept analysis (FCA), core features are screened by combining statistical correlation analysis and FCA attribute reduction algorithm, and various machine learning classifiers are used to predict the model identification effect; finally, a two-layer interpretation framework based on FCA concept lattice and SHAP analysis is constructed to form a visual interpretation chain from screening rules to prediction verification. [Result/conclusion] XGBoost model has an F1 value of 0.952 on multi-disciplinary da⁃tasets, which is significantly better than traditional methods; the two-layer interpretation system clarifies the feature combination rules of high innovation attributes and strong knowledge correlation, and quantifies the contribution of sin⁃gle features to the prediction results.

[Purpose/significance] While Large Language Model (LLM) demonstrates substantial potential in academic peer review, its application has also sparked widespread controversy regarding academic ethics and norms. Focusing on reviewers, this study investigates the behavioral processes and mechanisms underlying their collaboration with LLMs in real-world workflows. [Method/process] This study conducted in-depth interviews with 20 reviewers experienced in LLM-assisted peer review and employed thematic analysis for data analysis. [Result/conclusion] This study develops a human-AI collaborative behavior model in the context of peer review, consisting of five elements: collaborative conditions, collaborative motivations, collaborative strategies, collaborative feedback evaluation, and collaborative adaptation. Reviewers' behavioral decisions are moderated by subjective norms and capability foundations, and are driv⁃en by both instrumental and avoidance motivations. Based on the level of cognitive offloading and the degree of impact on the final review opinion, task allocation strategies between reviewers and LLMs can be structured as a continuum of "assistance–augmentation–co-creation agent." This continuum is accompanied by risk management strategies to ensureda⁃tasecurity and information quality. The collaboration between reviewers and LLMs is a continuous process of learning and adaptation. Driven by collaborative feedback evaluation, reviewers make corresponding behavioral adaptations, including the enhancement of LLM usage skills and the adjustment of collaborative strategies. This behavioral model provides new insights into the elements, pathways, and boundaries of human-AI collaboration in the context of peer review tasks.