Purpose This study investigates the behavioral patterns, key influencing factors and configurational mechanisms of researchers' information verification behavior toward Generative AI-generated content in human-AI interaction contexts. Design/methodology/approach A mixed-methods design was adopted. First, qualitative interviews and grounded theory identified influencing factors and constructed a theoretical framework. Then, hypotheses were proposed based on coding results and tested using partial least squares structural equation modeling on 506 valid questionnaires. Finally, fuzzy-set qualitative comparative analysis identified multiple paths triggering high-level verification behavior from a configurational perspective of antecedent conditions. Findings The study revealed three patterns of information verification behavior: verification via authoritative sources, platform comparison and interactive optimization verification and social network verification. Structural Equation Modeling showed that information quality, perceived risk, cognitive evaluation, AI literacy, task importance and algorithmic transparency significantly affect verification behavior. Qualitative comparative analysis further identified five equivalent driving paths, with cognitive evaluation and AI literacy as core conditions across all paths. The study also found that social influence, platform reputation and interactive design limitations, while not having significant independent effects, emerged as key configurational elements, highlighting the complexity of multi-factor concurrent effects. Originality/value The widespread use of Generative AI has triggered a severe information credibility crisis, making researchers' verification behavior crucial. This study addresses this challenge, fills the gap in research on AIGC as a “black-box information source” and provides evidence for building trustworthy AI systems and usage guidelines.
PurposeChatbots are increasingly embodied in business and IS contexts to enhance customer and user experience. Despite wide interest in chatbots among business and IS academics, surprisingly, there are no current comprehensive reviews to reveal the knowledge structure of chatbot research in such areas. Design/methodology/approachThis study employed a mixed-method approach that combines systematic review and bibliometric analysis to provide a comprehensive synthesis of chatbot research. The sample was obtained in December 2023 after searching across six databases: EBSCOhost, PsycINFO, Web of Science, Scopus, ACM Digital Library and IEEE Computer Society Digital Library. FindingsThis study reveals the major trend in publication trends, countries, article performance and cluster distribution of chatbot research. We also identify the key themes of chatbot research, which mainly focus on how users interact with chatbots and their consequences, such as users’ cognition and behavior. Moreover, several important research agendas have been discussed to address some limitations in the current chatbot research in business and IS fields. Originality/valueThe present review is one of the first attempts to systematically reveal the ongoing knowledge map of chatbots in business and IS fields, which makes important contributions and provides useful resources for future chatbot research and practice.
Purpose The increasing use of the terms “fake data” and “fake information” in digital systems has led to conceptual ambiguity and inconsistent usage across the literature. This study aims to clarify the definitional boundaries and relationships between these terms by examining how they are used and interpreted in existing research. Design/methodology/approach A structured literature review was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. A total of 334 studies published between 2009 and 2026 were selected after screening and eligibility assessment. Their abstracts were analyzed using thematic analysis, where recurring patterns were identified and encoded based on dimensions such as intent of creation, transformation processes and usage context. Findings The results reveal that neither “fake data” nor “fake information” is explicitly defined in the literature, with meanings typically inferred from context. The analysis also shows a marked increase in the use of these terms since 2020, with fake data more prevalent in technical domains such as cybersecurity and data science, and fake information more common in media and social contexts. Research limitations/implications The review is restricted to the computer science domain and its related fields (e.g. IT, data science and artificial intelligence), which may limit the generalizability of the findings to other disciplines. Despite these limitations, the study highlights the need for consistent terminology and provides a foundation for future research to refine and validate the proposed conceptual distinctions across broader contexts. Practical implications The proposed framework supports improved classification, risk assessment and intervention strategies by distinguishing between data-level manipulation and information-level distortion in digital systems. Originality/value This study introduces a data–information–knowledge-grounded, two-level framework that clarifies ambiguity in the use of fake data and fake information. The first level distinguishes between legitimate and deceptive intent, while the second differentiates based on what is being produced (data vs. information). This structure enables a clearer conceptual separation between fake data, fake information and synthetic data.
Purpose Despite exploratory search being a common information behaviour we engage in on a daily basis, it remains challenging to investigate. When people search for a book on a vaguely defined topic, they exhibit complex, multidimensional, and diverse search behaviours. To better understand the approach to such search tasks, a mixed-methods approach is encouraged. This article investigates the methodological benefits and drawbacks when utilising a triangulation design with the three methods used in our study.Design/methodology/approach The study was carried out in a live online bookstore, with a sample of 33 participants. In order to capture both observable interactions and participants' cognitive and emotional responses, we employed a triangulation design, combining screen recording, eye tracking, and concurrent think-aloud simultaneously. The analysis was conducted in three consecutive steps, with the explicit aim of comparing methodological insights first from each individual method and subsequently from their combination.Findings Screen recording and eye tracking offered complementary observational perspectives - one focusing on navigation paths and timing, the other on visual attention. Concurrent think-aloud added a qualitative, introspective layer, allowing us to access participants' reasoning and perceptions in real time. Combining insights from multiple methods improved the understanding of how the interface could better support exploratory book search, emphasising the importance of selecting and combining the right methods.Originality/value The paper's primary contribution is methodological, demonstrating how the combined use of the three methods strengthens research design and elevates the findings of exploratory search studies. Additionally, the case study revealed insights into user information behaviour during exploratory search in bibliographic information systems.
Purpose Livestreaming commerce, as an emerging digital retail industry, enables product sales across spatial boundaries. Livestreamers assume a pivotal role similar to traditional salespersons, presenting product information to viewers (i.e. potential consumers) in real time through verbal communication. Existing research has identified streamers' verbal cues as key signals influencing livestreaming sales. However, the plausible effects of nonverbal cues, such as acoustic characteristics, have rarely been investigated. Drawing on signaling theory, this study examines the impact of streamers' paralinguistic confidence markers (e.g. speech rate (SR), vocal pitch (VP) and intonation (IT)) on sales and investigates the moderating role of product price in livestreaming commerce. Design/methodology/approach Using field data from the Douyin platform, this study analyzes 365 livestream sessions conducted by 17 active streamers, extracting SR, VP and IT to build econometric models for empirical analysis. Findings The results indicate that SR and IT have a significant positive effect on sales, whereas VP shows a significant negative effect, consistent with the effects of perceived speaker confidence. Moreover, the influence of paralinguistic confidence markers is especially pronounced for low-priced product conditions but diminishes for high-priced product conditions, particularly for SR and IT. Originality/value This study offers theoretical insights into the role of paralinguistic confidence signals in digital marketing and provides practical recommendations for livestreaming commerce practitioners on leveraging these cues to enhance sales effectiveness.
Purpose As large language models (LLMs) are increasingly applied in information and data management - serving as tools for information retrieval, digital libraries, and knowledge organization - addressing their embedded biases is crucial for maintaining fairness and transparency. This paper aims to introduce a novel scenario-based test to quantitatively evaluate the level of implicit bias in LLMs, specifically in the context of information management tasks.Design/methodology/approach We employed a novel scenario-based test, designed to simulate real-world information tasks, to extensively investigate biases across 10 recently released open-source LLMs, including LLaMA 3.2 and Qwen 2.5. Our evaluation spanned 21 diverse bias domains (e.g. race, gender, religion, health) and was conducted in both English and Chinese contexts. We also performed an extensive comparative study against the Implicit Association Test to assess the suitability of our approach for this domain.Findings Implicit biases are pervasive across most models, with the potential to affect information processing and retrieval. In English, biases related to sexuality and skin tone are most prominent, while in Chinese, mental illness became the most biased domain, demonstrating a critical impact of language on bias. Our study shows that the scenario-based approach is better suited for capturing these context-dependent biases, making it more relevant for real-world applications in information and data management.Originality/value This work presents a novel, quantitative scenario-based test specifically for evaluating LLM implicit bias in information-management contexts. Extensively testing 10 LLMs across 21 domains and two languages provides concrete evidence of pervasive, language-dependent biases and offers a more context-aware assessment than previous methods, highlighting the urgent need for robust evaluation and mitigation strategies in this field.
Purpose This study aims to examine the information searching behavior of non-native users in learning-oriented tasks and to explore how search strategies and behavioral paths are associated with learning outcomes. Design/methodology/approach An experimental study was conducted, in which 28 non-native users completed learning-oriented Chinese writing tasks with and without search support. Screen recordings, structured behavioral logs, and writing performance scores were collected and analyzed to capture users' search strategies, behavioral sequences, and learning effects. Findings The results show that non-native users adopt simplified and concise query strategies, tend to search in non-native languages aligned with task requirements, rely on diversified information sources, and make extensive use of translation support. Learning-oriented searching is characterized by iterative and adaptive behavioral paths in which searching, browsing, and writing are tightly interwoven. Information searching has a significant overall effect on writing performance, although learning gains vary across individuals and behavioral indicators. Originality/value This study extends research on information behavior and search as learning by focusing on non-native users and adopting a process-oriented perspective. The findings provide empirical insights into how search behavior supports learning in non-native language contexts and offer implications for the design of search systems and learning environments.
Purpose- With the explosive growth of e-commerce and User-Generated Content (UGC), online reviews are critical for understanding consumer preferences and product issues. Yet their multi-dimensional, colloquial, domain-specific traits challenge traditional methods, while deep learning has high costs and low interpretability. This study aims to build a framework balancing insight depth and practicality-to capture dynamic customer priorities across periods and products, and offer actionable insights for pharmaceutical e-commerce platforms. Design/methodology/approach- Using anti-infective drugs as an example and online reviews from JD Pharmacy, we integrated the Latent Dirichlet Allocation (LDA) model with multiple tools (perplexity, coherence, and pyLDAvis) to extract influencing factors with higher precision. We then constructed and refined a domain-adapted sentiment lexicon for accurate sentiment quantification. Finally, we combined a multi-attribute model with PROMETHEE-II to ensure a reliable ranking of factor importance across different dosage forms and periods. Findings- The framework successfully extracted period-specific topics and significantly improved sentiment classification accuracy by 19.38 percentage points compared to traditional methods (p < 0.05). Consumer priorities exhibited significant two-dimensional dynamics (period & times; dosage form), and the priority rankings were proven statistically robust across all independent scenarios via Bootstrap resampling and non-parametric tests (p < 0.05). Based on these findings, targeted suggestions for improving satisfaction were proposed. Originality/value- This study systematically constructs an integrated text mining framework embedded with a dual-dimensional contextual logic. It establishes a standardized, low-cost, interpretable, and decision-closed-loop paradigm for UGC analysis, achieves the accurate quantification of consumers' dynamic priorities across different contexts, and provides a reliable pathway for vertical domain insights under resource constraints.
Purpose Effective health data trading depends on active engagement from multiple stakeholders. This study uses evolutionary game theory combined with prospect theory to model interactions among key stakeholders and to examine the dynamics of strategy evolution and the factors influencing stakeholder participation.Design/methodology/approach A four-player evolutionary game model is developed to capture the behaviors of data providers, data demanders, individuals and the government. Numerical simulations are conducted to validate the theoretical results and assess the effects of key parameters.Findings The analysis shows that health data trading progresses through distinct stages, each associated with different stable strategy combinations. Government incentives and penalties strongly shape the behaviors of providers, demanders and individuals, although excessive subsidies can weaken regulatory commitment. Data providers are more sensitive to costs and payments than demanders, reflecting their dominant position. Individual perceptions of value, cost and privacy risk are central to sustaining data sharing. Dynamic adjustment of government rewards and penalties is essential for encouraging active participation across market stages and for enabling a gradual reduction in direct intervention.Originality/value This study offers an integrated behavioral and policy perspective on health data trading and advances understanding of health data monetization. It provides actionable insights for designing balanced incentive-regulation mechanisms that build trust, support sustainable market development and inform policymaking for responsible data trading.
Purpose This study investigates the key factors and evolutionary dynamics of collaborative governance in addressing artificial intelligence-generated content (AIGC) disinformation.Design/methodology/approach A tripartite evolutionary game model (EGM) is developed to analyze interactions among users, social media, and the government. The model integrates Bounded Rationality Theory (BRT), Deterrence Theory (DT), and Public Goods Theory (PGT). Evolutionary stable strategies are derived using the Jacobian matrix and Lyapunov's first method, supported by numerical simulations and sensitivity analyses.Findings Social media and the government's initial willingness are vital for accelerating users' transition to compliant behavior through adaptive learning of boundedly rational agents. Due to proximity effects, users are more responsive to social media penalties than government sanctions, revealing an asymmetric deterrence effect. Stronger government regulation reduces social media's incentive for active governance. This creates a free-rider dilemma that requires carefully designed subsidies and penalties to curb opportunistic behavior. Cost-benefit dynamics determine optimal governance - reducing costs or improving effectiveness fosters active governance and lenient regulation, while high costs lead to instability - highlighting the importance of technological innovation for sustainable governance.Originality/value This study extends EGM and BRT to generative artificial intelligence governance by integrating AI-specific parameters into AIGC disinformation analysis. It refines DT and PGT by revealing asymmetric deterrence effects and formalizing the free-rider dilemma. It proposes a tripartite governance framework that unifies BRT, DT, and PGT and identifies cost-benefit dynamics that challenge the assumption that "more regulation is better." These findings offer a flexible framework and actionable strategies for collaborative governance of AIGC disinformation.
Purpose Artificial intelligence (AI) stands as a pivotal, revolutionary force in technologically reshaping industries. Despite extensive research on the potential of AI, the specific mechanisms through which AI capabilities lead to competitive advantages still need delineated. To address this notable gap in literature, we investigate how agility in decision-making, as a focal dynamic capability, is a critical conduit linking AI capabilities for improving organizational outcomes and examine the interplay between a firm's decision-making agility and its internal and external environment, thereby offering a comprehensive understanding of the dynamics of today's complex business ecosystem.Design/methodology/approach The mixed-method approach, combining partial least squares-structural equation modeling (PLS-SEM) and fuzzy-set qualitative comparative analysis (fsQCA), was used for analyzing the survey data of 251 firms.Findings The results of PLS-SEM show that two sub-dimensions of AI capability, technical infrastructure and management, affect performance outcomes through decision-making agility. Moreover, environmental dynamism and complexity differently moderate the relationship between decision-making agility and firm performance. Additionally, the results of fsQCA demonstrate how the combination and roles of strategic resources (e.g. AI capabilities and decision-making agility) shift in response to varying organizational and environmental conditions.Originality/value By elucidating these dynamics via a mixed-method approach, our findings not only offer a deeper understanding of the strategic value of AI within the organizational dynamic capability perspective, but also provide practical insights for AI implementation that prioritizes management capability and adaptability to external environments.
Purpose The study aims to design, develop and demonstrate a Python-based application that enhances the accessibility, usability, and analytical potential of big open data. It addresses the growing challenge of transforming large-scale, publicly available datasets into actionable insights that can support research activities and evidence-based policy decision-making. Design/methodology/approach The study adopts a design-oriented research approach. A systematic literature search was conducted to identify conceptual gaps, methodological limitations, and suitable open-source technologies for big open data analytics. The Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology guided the structured design and implementation of the application, integrating Apache Spark and PySpark to support scalable data processing. The proposed framework was validated through a case study analysing the City of Chicago’s open crime dataset. It demonstrates the complete data science pipeline from data collection and preprocessing to visualization and predictive modelling. Findings The developed application streamlines big open data analysis by operationalising the CRISP-DM stages within a single workflow and automating key data science tasks. The case study results show that the application effectively identifies crime trends and patterns, illustrating its capability to support data-driven urban management and informed decision-making by users with varying levels of technical expertise. Practical implications The proposed framework enables public authorities, researchers, and policymakers to analyse large open datasets more efficiently, supporting evidence-based planning and enhancing the operational value and transparency of open data initiatives. Social implications By lowering technical barriers to advanced data analytics, the application promotes broader engagement with open data and encourages data-driven innovation, contributing to improved societal outcomes in domains such as public safety, governance, and urban development. Originality/value This research contributes one of the few open-source, Python-based analytical frameworks explicitly designed for big open data analysis using CRISP-DM principles. The study advances the literature by delivering a reusable, replicable, and scalable software artifact that bridges the gap between conceptual data analytics frameworks and practical implementation for non-specialist users.
PurposeDrawing on the information system model and attachment theory, the purpose of the current research is to validate how standardization and personalization jointly affect users' continuous use intention of artificial intelligence (AI) chatbots through emotional attachment.Design/methodology/approachEmpirical data were collected from 551 active users through the online survey platform Credamo. Partial least squares structural equation modeling was employed to assess both the measurement model and the structural model. Additionally, the PROCESS macro based on SPSS was used for the examination of the parallel-serial mediation model.FindingsStandardization and personalization are positively correlated with information quality and system quality. Moreover, information quality and system quality exhibit positive correlations with emotional attachment. Furthermore, the parallel-serial mediation analysis affirms the significant indirect effect of standardization and personalization on continuous use intention.Research limitations/implicationsThis study enhances societal comprehension of the nuanced interplay between information quality, system quality and emotional responses within the AI chatbot domain. It provides actionable insights for AI enterprises and developers seeking to improve interaction experiences and drive the evolution of AI chatbot systems.Originality/valueIn human-computer interaction, developers face a persistent challenge in balancing service standardization for reliability with personalization for relevance. While research has often advanced these in parallel, the joint influence of these two seemingly contradictory strategies on long-term user engagement remains theoretically underdeveloped. This study addresses this critical gap by proposing and testing a model that illuminates the psychological mechanisms linking this strategic balance to continuous use.
Purpose - This study aims to use a configurational approach to identify which combinations of questioner and question characteristics drive high-quality scholarly engagement in academic Q&A sites. It argues that answer quantity and quality result not from single factors but from the complex interplay between multiple features. Design/methodology/approach - This research analyzed 1,705 question records from the AI community on ResearchGate. It employed fuzzy-set Qualitative Comparative Analysis (fsQCA) to examine how configurations of six features-including readability, emotionality, topic relevance, and the questioner's Research Interest Score (RIS) and Q&A involvement-lead to high or low answer quantity and quality. Findings - No single condition guarantees positive answer performance. Instead, it results from specific configurations. A key finding is the complementary relationship between readability and emotionality: high emotionality can compensate for low readability to achieve good answer performance. Furthermore, a questioner's high RIS and Q&A involvement level are core conditions for achieving both high answer quantity and quality, often appearing in combination. Originality/value - This study is the first to apply a configurational perspective to academic Q&A sites, revealing the complex interplay of factors rather than isolated effects. It challenges findings from general Q&A sites by demonstrating that, in academic contexts, questioner reputation is pivotal and high readability is not always essential for success. The research provides a novel framework for understanding and fostering interdisciplinary academic exchange online.
Purpose-Physicians' online popularity plays a vital role in the development and success of online healthcare communities (OHCs). However, key factors driving this popularity are yet to be fully explored. Thus, this study aims to identify and prioritize the key factors contributing to physicians' online popularity in OHCs using a comprehensive, multilayered analytical approach. Design/methodology/approach-The multilayered 4-step methodology was adopted to identify and prioritize the factors influencing physicians' online popularity in OHCs. Initially, relevant factors were identified through a comprehensive literature review and validated by experts using the Fuzzy Delphi Method (FDM). Semi-structured interviews with diverse stakeholders were then analyzed using the Decision-Making Trial and Evaluation Laboratory (DEMATEL) technique, based on stakeholder theory, to develop a causal framework. To validate the findings, machine learning (ML) models, including Artificial Neural Networks (ANN), Decision Tree, XGBoost and Random Forest were applied. Findings-The analysis revealed that "active social media presence" and "quality of communication" are the most influential factors contributing to physicians' online popularity. The causal relationships derived from DEMATEL were further supported by machine learning validations, confirming the reliability and significance of the identified factors. Originality/value-This study offers a novel integration of FDM, DEMATEL, and machine learning models to investigate physician popularity in OHCs. By combining expert insights with empirical data and robust validation techniques, the research provides both theoretical and practical contributions to understanding how physicians can enhance their popularity and impact within digital health environments.
PurposeOnline mental health communities (OMHCs) are important platforms for seeking help and providing social support, which can promote more positive emotional experiences for users. However, current research primarily focuses on how to access social support from OMHCs and validate its benefits, while neglecting to explain the underlying mechanisms of how social support facilitates emotional well-being. Therefore, from a person-centered perspective and based on congruity theory, we investigate the effects of social support congruence on depression-anxiety emotion changes, as well as the moderating roles of user involvement level and support source.Design/methodology/approachThis study collects 12,832 chat transcripts from the online depression and anxiety group (ODAG) within an OMHC, segmenting them into 5,013 conversation rounds for quantitative analysis.FindingsWe find that both moderate levels of thematic and emotional congruence are associated with favorable emotion change. However, a high degree of focus on the same topic may hinder this positive emotional shift. Increasing involvement, in turn, can mitigate this negative effect and amplify the benefits of positive emotion difference.Originality/valueOur study enriches theoretical research on health management, contributes to social support literature and provides practical insights for individuals, administrators and practitioners in health management.
Purpose - This study explores how artificial intelligence (AI) influences employees' social innovation at work, using the stimulus-organism-response (SOR) framework. Particularly, we examine whether AI impacts social innovation directly or indirectly through its effects on employee creativity and risk-taking. Furthermore, the study investigates the moderating role of employees' trust in AI. Design/methodology/approach - Data were gathered from 243 employees across 12 AI-integrated financial organizations in Iran with the use of the stratified random sampling method and analyzed using partial least squares structural equation modeling (PLS-SEM). Findings - The findings suggest that AI does not directly drive social innovation. Instead, its impact occurs through employees' creativity and willingness to take risks. Moreover, trust in AI amplifies the positive relationship between AI adoption and both creativity and risk-taking, but it does not moderate the direct relationship between AI use and social innovation. Originality/value - This study specifically addresses whether AI fosters employees' creativity, risk-taking and social innovation within the financial sector. The analyses highlight the importance of employee trust and individual factors, such as creativity and risk-taking, in unlocking AI's innovative potential in the workplace.
PurposeCollecting online opinions, particularly suggestive opinions on pre-release policies, is crucial for the government's informed policy-making. Grounded in social cognition theories, this study explores how public social attributes, cognitive experience and environmental factors influence the expression of suggestive opinions on such pre-release policies.Design/methodology/approachWe collected a dataset with 37,087 online comments on 12 selected pre-release policies and 1,059,019 microblogs by commenting users on Sina Weibo in China. Using text mining techniques, we extracted relevant features and applied a two-stage regression analysis alongside explainable machine learning to identify factors influencing suggestive opinion expression.FindingsFemales, individuals possessing more extensive social experience, and users who frequently convey negative sentiments exhibit a higher propensity to post suggestive opinions. Environmental factors, particularly existing suggestive opinions, exert a significant promoting effect on the subsequent expression of suggestions. Moreover, there exists a time decay effect in the expression of suggestive opinions, and these effects are more prominent among females compared to males.Originality/valueWhile prior research has explored determinants of general online information behaviors and examined sentiment analysis, this study seeks to uniquely investigate the factors driving suggestive opinion expression, as opposed to other opinion types.
PurposeWith the rapid expansion of AI use in healthcare, chatbot-based AI health assistants are increasingly deployed for symptom checking, screening support and health promotion. However, public acceptance remains limited because many users question the reliability and diagnostic value of AI-provided health information and feel that these systems cannot respond to their emotions during interaction. Against this backdrop, this study asks: which AI features are associated with users' perceived diagnosticity and perceived empathy, and how perceived diagnosticity and perceived empathy relate to adoption intention, and whether these relationships vary by age.Design/methodology/approachA cross-sectional online survey was conducted, yielding 215 valid responses. The proposed model links AI feature cues (accuracy, responsiveness, personalization, affinity and anthropomorphism) to perceived diagnosticity and perceived empathy, which in turn predict adoption intention, with age as a moderator. Partial least squares structural equation modeling was used to assess the measurement and structural models and to test the hypothesized paths and moderation effects.FindingsAccuracy and responsiveness were positively associated with perceived diagnosticity, whereas personalization was not. Affinity was positively associated with perceived empathy, while anthropomorphism showed no significant association. Both perceived diagnosticity and perceived empathy were positively related to adoption intention, with perceived diagnosticity exhibiting a stronger effect than perceived empathy. Moreover, age positively moderated the relationship between perceived diagnosticity and adoption intention, whereas the moderation effect of age on the perceived empathy and adoption intention link was not significant, indicating that the positive impact of diagnosticity on adoption intention becomes stronger as age increases.Originality/valueThis study contributes to research in three ways. First, it extends beyond general technological beliefs by modeling adoption as a social cognitive evaluation grounded in mind perception. Second, it empirically differentiates which features primarily function as competence cues versus warmth cues in AI health assistants. Third, it demonstrates that competence-related evaluation (perceived diagnosticity) is stronger than warmth-related evaluation (perceived empathy) for adoption intention in a higher-risk medical setting and identifies age as a key boundary condition.
PurposeThis study aims to understand the development status of public data utilization policies for research needs in the USA, UK and China. By comparing the content structure of policies, it proposes policy optimization strategies to strengthen the ability of public data to drive scientific advancements.Design/methodology/approachBased on the Open Data Barometer rankings, this study selects the USA, UK, and China as representative cases from high-, medium-, and low-tier countries. Using a three-dimensional framework (policy tools, data lifecycle, policy objectives), it conducts a textual quantitative analysis of 154 policies through systematic coding.FindingsPost-2016, policies have increasingly addressed research needs, primarily through implementation plans and operational guidelines (44.16%). Cross-national trends reveal overreliance on environmental-side policy tools. Data lifecycle management prioritizes preliminary planning through environmental tools while neglecting post-utilization monitoring. Policy objectives balance data openness and security using environmental and supply-side tools, whereas demand-side tools focus on value creation but inadequately protect the rights of subject. The study reveals distinct national policy characteristics: the USA adopts an "open-first, ecosystem-driven" approach, the UK implements a "governance-first, trusted-access" authorization mechanism, while China demonstrates a "policy-guided, pilot-based" progressive model of data utilization.Originality/valueTheoretically, this study delineates the policy logic of public data utilization for research through policy tools, providing the research perspective of the relationship between public data and research innovation. Practically, research findings offer insights for policymakers to optimize policy.
