Purpose The rapid integration of artificial intelligence (AI) technologies into information services has made librarians’ competencies and acceptance levels toward these technologies an increasingly critical research concern. This study aims to examine the relationship between AI literacy levels and AI technology acceptance levels among librarians in Türkiye. Design/methodology/approach The technology acceptance model (TAM) scale and the AI literacy scale were used as measurement instruments in the study. Data were collected from 203 librarians working in different types of libraries through an online survey. Pearson correlation and multiple linear regression analyses revealed that AI literacy significantly predicted technology acceptance, with the evaluation dimension emerging as the dominant predictor of both perceived usefulness and perceived ease of use. Findings The finding that demographic variables largely showed limited effects suggests that AI literacy and acceptance are shaped by professional training and organizational conditions. Originality/value This study represents one of the first empirical investigations to examine the relationship between AI literacy and technology acceptance within an integrated model framework in the Türkiye library science context.
Purpose Scholarly databases (SDs) are fragmented and evolve independently, creating incomplete and inconsistent metadata that hinders reuse, trustworthy analytics and knowledge discovery. This study aims to propose the Scholarly Linked Databases (SLD) framework to systematically exploit cross-database complementarity for scalable enhancement. Design/methodology/approach SLD builds large-scale article linkages across major databases and performs field-level metadata fusion to fill missing values and resolve inconsistencies. We validate SLD on multiple database pairs and evaluate its downstream utility via disambiguation, retrieval, recommendation and bibliometrics, using ratio-based indicators and matched samples to mitigate bias from unequal database sizes. Findings Dense cross-database linkages can be established in practice, and SLD improves metadata completeness and consistency. SLD-enhanced data yield measurable gains over non-linked baselines across downstream tasks, indicating better support for Scholarly Big Data enhancement and scientific innovation. Originality/value The originality lies in synthesizing and systematically applying existing techniques (linking, cleaning, integration) into a dedicated, scalable framework for the SD ecosystem. SLD also frames inter-database variation as an evolutionary resource that can be selectively combined to create fitter enhanced databases.
Purpose This paper aims to propose an approach for the semantification of scientific papers so as to save the time of researchers in obtaining access to knowledge of their domain, comparing research contributions and getting novel insights. Design/methodology/approach The approach proposed in this paper consists of extracting key insights from scientific papers leveraging neural models and organizing them using a symbolic AI model which is a scholarly knowledge graph (KG). To validate this approach a research KG is implemented. Findings This research KG aims to semantify papers published by the SemTab@ISWC (Semantic Web Challenge on Tabular Data to KG Matching @ International Semantic Web Conference) venue. This research KG is composed of more than 650 instances. The authors demonstrate the utility of the KG obtained by using it to answer SPARQL [SPARQL Protocol and Resource Description Framework (RDF) Query Language] queries about several questions a researcher may have on semantic table annotation. Originality/value This is an original work to propose a neuro-symbolic (NeSy) approach which leverages connectionist models and a symbolic model for the semantification of scientific papers.
Purpose This study aims to investigate the key predictors of artificial intelligence literacy (AI literacy). It explores the mediating role of digital literacy (DL) in the relationship between psychological (interest, attitude and self-efficacy) and experiential (prior experience with AI tools) factors and AI literacy. Design/methodology/approach This study used a survey-based quantitative research methodology. Data were obtained from 754 students enrolled in various programs at the University of the Punjab. Structural equation modeling (SEM) was used to test direct and mediated relationships among constructs. Findings Interest, attitude, self-efficacy and prior experience significantly predicted AI literacy, both directly and indirectly through DL. Self-efficacy demonstrated the strongest total effect on AI literacy. Prior experience with AI tools significantly improved DL, which in turn enhanced AI literacy. The results highlight a dynamic interplay between psychological readiness and experiential learning, positioning DL as a crucial mediator in the development of AI literacy. Practical implications Educators and policymakers should design curricula that foster digital competence, encourage early exposure to AI tools and build learner self-efficacy. Structured, hands-on experience with AI technologies can support the development of confident and competent AI users. Originality/value This study presents a comprehensive model linking psychological and experiential variables to AI literacy through DL. It offers both theoretical advancement and practical guidance for educational stakeholders preparing learners for AI-integrated futures.
Purpose This paper aims to systematically review the current research on artificial intelligence (AI)-empowering academic evaluation, and subsequently construct a theoretical framework that aligns with the era of digital intelligence so as to enrich intelligent academic evaluation theory and promote its practical exploration, thereby providing a reference for future studies in the field of AI-empowering academic evaluation. Design/methodology/approach This paper uses bibliometric and content analysis methods. Using the SSCI database within the Web of Science Core Collection as the data source, 587 valid bibliographic records of journal articles in the field of AI-empowering academic evaluation are retrieved and screened. CiteSpace software is used to create keyword clustering knowledge mapping and time period knowledge mapping for research topics from 2015 to 2025. Through these visualizations, the current research landscape and development trends in this field are analyzed in depth. A theoretical framework for enhancing AI-empowering academic evaluation is constructed, and future research prospects are proposed. Findings This paper elaborates on the research progress in the field of internationally AI-empowering academic evaluation across six clustered themes: AI and academic evaluation theories and methods, AI and academic evaluation objects, AI and academic evaluation indicators, AI tools empowering academic evaluation, AI-empowering academic evaluation practices, and AI triggering academic evaluation risks and governance. By analyzing the time period knowledge mapping, the study divides the research on AI-empowering academic evaluation into three developmental stages: the initial exploration phase, the integration and expansion phase, and the GenAI-driven paradigm reshaping and reflection phase, and further examines its future trends. Based on this analysis, the paper constructs a theoretical framework for enhancing AI-empowering academic evaluation and proposes six key research prospects and development strategies: drive the intellectualization and efficiency of academic evaluation theories and methods; promote the diversification and dynamics of academic evaluation objects; assist the meaning and integration of academic evaluation indicators; enhance the contextualization and interactivity of AI tools; achieve consensus-based and flexible academic evaluation systems; Promote the systematization and globalization of AI governance. Originality/value This paper enriches the theoretical research on AI-empowering academic evaluation, identifies key leverage points that can be translated into policy instruments, provides insights into the application directions and practical pathways of AI technology in academic evaluation research, and contributes to promoting the deep integration of academic evaluation systems with AI technology and their intelligent transformation.
Purpose Amid deepening digital transformation and intelligence advances, emerging technologies are reshaping libraries globally, yet regional disparities in priorities and insufficient theoretical development regarding core features and evaluative tools persist. This study aims to address these gaps by constructing a maturity evaluation indicator system and method to facilitate smart library development. Design/methodology/approach Content mining was first employed to delineate smart library connotation and components. Building on prior research, object-oriented analysis elicited initial indicators, subsequently refined through expert interviews. Analytic hierarchy process (AHP) and expert scoring determined indicator weights. A maturity model integrated with triangular fuzzy functions formulated the evaluation methodology. Empirical validation was conducted across five Chinese university libraries. Findings The developed indicator system encompasses 4 evaluation elements, 12 first-level indicators and 48 second-level indicators. Application to empirical cases confirmed the scientific validity and operational feasibility of the proposed evaluation model. Originality/value This study creatively integrates tools from information science, systems engineering and management science to design a validated research scheme. It proposes a systematic evaluation indicator system, establishes a maturity model with triangular fuzzy calculation and assesses five Chinese university libraries, offering a valuable methodological paradigm for smart library evaluation.
Purpose Against the backdrop of the widespread application of generative artificial intelligence (AI) and the resulting challenges of information authenticity, this study aims to focus on users’ information verification behaviour when facing AI-generated content. It explores the influencing factors and underlying mechanisms of such behaviour, which is of great significance for enhancing public information literacy, improving the information ecology of AI platforms and fostering healthier information environments. Design/methodology/approach A mixed-methods approach is adopted, combining structural equation modelling (SEM) to test the net effects of individual variables and fuzzy-set Qualitative Comparative Analysis (fsQCA) to identify complex, synergistic configurations leading to high verification behaviour. Findings The SEM results indicate that self-efficacy, risk perception, algorithmic literacy, technological facilitation and subjective norms have significant positive effects on information verification behaviour, while information quality and platform reputation exert significant negative effects. The effects of information presentation features and algorithmic transparency are not significant. Furthermore, fsQCA identifies six high-coverage configurational paths, which can be categorised into three types: cognitive-technology enhancement, composite synergy-driven and norm-risk guided. Originality/value For the first time, this study combines information ecology theory with mixed-methods (SEM + fsQCA). Breaking through traditional single-factor analysis, it deepens our understanding of the synergistic mechanisms among information, humans, technology and environment, demonstrating both theoretical and methodological innovation.
Purpose Quotation error refers to the inconsistency between cited information and its original source. This phenomenon leads to a series of negative impacts, such as misinterpreting of original research, undermining the academic community's collective understanding of relevant issues and weakening the accuracy and fairness of the citation-based academic evaluation system. Existing studies have shown that quotation error is prevalent in the academic community; moreover, manual verification of quotation error is not only labor-intensive but also inefficient. Therefore, this paper aims to propose the task of "automated detection of quotation errors."Design/methodology/approach Adopting a large language model (LLM)-based approach, this paper improves detection performance from two aspects on the basis of existing research: first, using the fine-tuning approach for LLMs to detect quotation errors; second, incorporating full-text data of the cited literature into data set construction and third, exploring the optimal scheme for building such data sets by comparing three types of full-text integration methods. Based on this, the paper further uses the TokenSHAP tool to conduct interpretability experimental analysis on the model's prediction results.Findings The fine-tuning approach for LLMs has improved the performance in detecting quotation errors. Among the different methods for incorporating full-text information, the approach based on using the source abstract yielded the best performance.Originality/value The fine-tuning approach for LLMs is applied to the task of automated detection of quotation errors, and interpretability analysis is conducted on the model's output results.
Purpose - This study aims to construct a quality indicator system for artificial intelligence-generated content (AIGC), enhance users' ability to identify high-quality AIGC and provide a reference for service providers to optimize their content. Design/methodology/approach - This study collected primary data through literature review and in-depth user interviews, and applied grounded theory to conduct three-level coding to identify and extract indicators for evaluating AIGC quality. Based on these indicators, a questionnaire survey and the analytic hierarchy process were used to determine indicator weights and to construct a comprehensive AIGC quality evaluation system. Findings - This study identifies four first-level indicators and 21 second-level indicators for evaluating AIGC quality, along with their respective weights. To further highlight the uniqueness of the constructed indicator system, this study also reveals the commonalities and differences in quality assessment dimensions among AIGC, professionally generated content, and user-generated content through comparative analysis. Originality/value - This research aims to enrich the academic discussion on AIGC quality indicator system, help users identify high-quality AI-generated content and provide guidance for relevant stakeholders to formulate policies.
PurposeThis study aims to enhance the performance of fine-tuned small named entity recognition (NER) models by leveraging the language generation capability of large language models (LLMs) to rewrite entity contexts, addressing the limited performance gains of state-of-the-art (SoTA) NER methods in recent years.Design/methodology/approachThe authors propose a novel paradigm that uses LLMs to modify the surrounding context of entities in the input text, transforming hard cases into easier ones for the NER model to recognize. Experiments are conducted on multiple widely used NER data sets to evaluate the effectiveness of the proposed method. No additional retraining or architectural modifications are applied to the vanilla NER models.FindingsExperimental results demonstrate that the method consistently improves the performance of existing NER models across all tested data sets, achieving SoTA results. The approach effectively mitigates challenges associated with out-of-vocabulary entities, syntactically complex sentences and linguistic distractors while maintaining relatively low computational cost.Originality/valueThis work introduces a new paradigm for NER that integrates LLM text rewriting with pretrained small NER models. Unlike previous approaches relying on data augmentation or NER model retraining, the authors' method achieves performance gains purely through LLM-based context rewriting, offering a cost-efficient and scalable solution for real-world applications.
PurposeThis study aims to identify and explain the key factors that encourage or discourage users from continuing to use Generative Artificial Intelligence (AI) platforms. Specifically, the research examines how system-related facilitators (such as interaction quality, personalization, reliability and affordances) and psychological barriers (including inertia, perceived threat and regret avoidance) shape users' post-adoption attitudes and continuance intentions toward Generative AI.Design/methodology/approachThe survey data were collected from the respondents applying the purposive sampling technique. Partial Least Squares structural equation modeling was used in the analysis.FindingsThe study's results reveal that perceptions of interaction quality, personalization, reliability, creative affordance and analysis affordance significantly promote the intention to continue using. Conversely, perceptions of inertia, threats and regret avoidance significantly hinder continued use.Research limitations/implicationsThe findings might not be widely generalizable. The data were collected only in a particular community.Practical implicationsThese insights offer critical implications for business owners, platform developers and policymakers aiming to retain consumers of Generative AI products.Originality/valueTo attain the objective, this research integrated the "Elaboration Likelihood Model" and "Status Quo Bias theory". It developed a conceptual model to address cognitive, emotional and behavioral components.
PurposeArtificial intelligence (AI) chatbots have gained significant attention in the field of education, particularly in medical training and learning. This study aims to measure the AI chatbots adoption among medical students by extending the theory of planned behavior (TPB). The research examines the key determinants such as perceived behavioral control (PBC), past behavior (PB), subjective norms (SN), information literacy (IL), AI literacy and behavioral intention (BI).Design/methodology/approachA structured questionnaire was distributed among medical students for data collection. Participants received 400 copies of the questionnaire and the consent form to take part in the study. A total of 326 individuals returned the completed surveys, yielding an 81.5% response rate. 310 (77.5%) valid responses of the 326 questionnaires used for data analysis. To analyze the data, SPSS v26 (IBM) was used along with moment structure analysis (AMOS).FindingsFindings indicated that PBC, PB and BI significantly influence AI chatbots adoption among medical students. However, the relationships of IL and AI literacy are positive and insignificant, while the relationship of SN is negative and insignificant.Research limitations/implicationsThe study's insights are beneficial for educators, policymakers and AI software developers seeking to enhance the use of AI chatbots in educational settings.Originality/valueThis study adds to the literature on the use of AI in education by analyzing the impact of TPB on predictive factors of using chatbots.
PurposeBased on the conservation of resources (COR) theory, the purpose of this study is to explore the enablers and inhibitors of AI-generated content (AIGC) user intermittent discontinuance.Design/methodology/approachA mixed method of structural equation modeling and fuzzy-set qualitative comparative analysis was adopted to conduct data analysis.FindingsThe results of this study showed that algorithm bias, misinformation, privacy concerns and unexplainability affect usage exhaustion, which promotes user intermittent discontinuance. Switching costs, subjective norm and irreplaceability affect switching exhaustion, which prevents user intermittent discontinuance. The fuzzy-set qualitative comparative analysis identified two paths that lead to user intermittent discontinuance.Research limitations/implicationsThe results suggest that AIGC platforms need to lower usage exhaustion and increase switching exhaustion to prevent user intermittent discontinuance and ensure the continuous development.Originality/valuePrevious research has focused on the effect of technological factors such as perceived usefulness on AIGC user behavior. This research provides new insights on AIGC user intermittent discontinuance from a resource conservation perspective. The results also expand the COR theory from traditional organizational behavior to the emerging AIGC context.
PurposeThis study aims to explore the mechanisms underlying public risk perception and trust in artificial intelligence-generated content (AIGC) technologies. It seeks to clarify how these factors influence behavioral intention and risk prevention sensitivity, thereby informing responsible governance of emerging digital technologies.Design/methodology/approachGuided by the UTAUT2, SARF and TPB frameworks, a conceptual model was developed to examine the interrelations among risk perception, system trust, degree of risk trust, behavioral intention and risk prevention sensitivity. A Bayesian structural equation modeling (BSEM) was used to analyze data collected from 1,185 respondents in four cities across Jiangsu Province, China.FindingsThe results reveal that increased risk perception can enhance public trust in governance systems, especially when supported by technical transparency and institutional safeguards. However, higher risk prevention sensitivity may inhibit the intention to adopt AIGC technologies. The study emphasizes the importance of a governance framework incorporating transparency, adaptive regulation and cross-sector collaboration.Originality/valueThis study establishes a detailed paradigm for Bayesian structural equation modeling in socio-technical research and provides empirical evidence to support transparent and trustworthy governance. It contributes to the development of a multi-stakeholder model for the responsible advancement of AIGC technologies.
PurposeResearch data repository (RDR) consortia are pivotal to improving the efficiency of scientific data sharing. However, the factors that determine their success across different life cycle stages remain underexplored. This study aims to identify and prioritize the critical success factors (CSFs) affecting RDR consortia in China from a life cycle perspective.Design/methodology/approachA mixed-methods approach was adopted to examine CSFs across the planning, implementation and sustainability stages. Semi-structured interviews with eight domain experts from China were conducted to generate CSFs, which were subsequently weighted through an analytical hierarchy process survey involving 112 stakeholders.FindingsThe results identify 21 factors influence RDR consortia success. Reflecting China's top-down context, governance-related factors and researcher engagement dominate the planning stage. During implementation, legal and standardization mechanisms are most critical. In the sustainability stage, adaptive governance and benefit-sharing mechanisms are crucial for ensuring long-term resilience and responsiveness.Originality/valueThis study offers a life cycle-based framework that integrates legal, technical and governance factors across stages, providing actionable insights for academic libraries, policymakers and consortia managers, particularly in nations with similar top-down data strategies.
PurposeThis paper aims to present a case study of integrating generative artificial intelligence (AI) technologies into archival metadata creation in an academic library, providing an example of improving descriptive metadata for digitized photograph collections to enhance discoverability and accessibility.Design/methodology/approachThis study tested various vision-enabled and multimodal large language models for generating Metadata Object Description Schema-compliant metadata from digitized photographic prints. Models were evaluated on visual interpretation and metadata output quality. Following a pilot test that validated this approach, Anthropic's Claude Sonnet 4 model was selected for production implementation, creating descriptive metadata for 2,263 digitized photos. A Python-based process helped integrate this into existing workflows for digital collections.FindingsThe process provided substantial improvements over existing minimal descriptive metadata. Generated subject terms mapped successfully to the Faceted Application of Subject Terminology vocabulary in 64% of cases, demonstrating compatibility with established standards. Project team discussions led to the implementation of user transparency notices about AI involvement in metadata creation. Post-project analysis revealed this approach to be a cost-effective method of metadata enhancement.Practical implicationsThis case study provides actionable guidance for cultural heritage institutions evaluating AI integration into their workflows and can be adapted for different institutional contexts and collection sizes.Originality/valueWhile AI applications in libraries are expanding, this study documents a practical approach to integrating generative AI into archival metadata creation. The workflows, quality controls, cost analysis and transparency measures offer a roadmap for those considering adopting these technologies at scale.
Purpose This study aims to compare large language models (LLMs) with human analysis for rhetorical move detection in journal article abstracts, examining whether automated discourse analysis can complement human efforts in bibliographic metadata creation and information retrieval. Design/methodology/approach Using an established data set and BAMRC (Background, Aim, Method, Results, Conclusion) framework from a previous study of social science abstracts, four contemporary LLMs (OpenAI GPT-4o Mini, DeepSeek Chat, Claude 3.5 Haiku and Gemini 2.0 Flash) were compared against the original human analysis. Identical prompts were used across models to ensure comparability. Findings LLMs showed modest agreement with human annotations at the abstract level (51.7%–57.9% Jaccard similarity) but substantially higher intermodel agreement (77.0%–86.7%). This pattern indicates convergence toward a distinct LLM-influenced annotation style that diverges systematically from human judgment, including generally higher rates of complete BAMRC structures and more frequent use of the Undefined category at the sentence level (11.6%–29.2% vs 4.8% for humans). Agreement declined sharply at the sentence level, averaging approximately 19%. Practical implications LLM-based rhetorical analysis can provide scalable insights into abstract structure and academic discourse patterns, informing quality assessment and indexing practices in digital libraries. API-driven workflows enable low-cost, large-scale metadata analysis as a complement to human expertise, without presuming full replacement of manual processes. Originality/value This study offers a multi-LLM comparison on a decade-old, human-annotated social science abstract data set, highlighting both the potential and the limitations of transferring LLM capabilities to rhetorical move analysis and related discourse tasks.
Purpose-This study aims to identify the factors influencing the adoption of linked data technology in library settings and to reveal associated challenges affecting its implementation. Design/methodology/approach-Systematic literature review methodology was applied to address the study's objectives. For the current research, 24 studies (peer-reviewed research papers and conference proceedings) were selected. Findings-Findings of the study displayed that enhanced discoverability, technological integration, metadata enrichment, user-centered design and institutional were strong motivators driving the adoption of linked data technology in libraries. However, technical limitations, lack of training, funding constraints, organizational unpreparedness, absence of standards and concerns about privacy negatively influenced the incorporation of linked data applications in library settings. Originality/value-The study has provided a framework to effectively adopt linked data applications in library environments.
PurposeThis study aims to investigate the effects of awareness and technological readiness among university LIS professionals in Punjab province on their intention to use artificial intelligence (AI). The Technology Readiness Index (TRI) theoretical framework was used, with an additional awareness variable added.Design/methodology/approachLibrary and information science (LIS) professionals from higher education commission (HEC)-recognized academic universities of Punjab were the population of this study. There were approximately 350 LIS professionals in Punjab. Survey methods were used within the framework of a quantitative research design. A questionnaire was used as an instrument for data collection, and it was based on a Likert scale. Data was collected using a convenient sampling technique, and 213 responses from LIS professionals were received.FindingsData was analyzed using Linear regression to measure the relationship between variables and to test hypotheses. A positive relationship of optimism, innovativeness and awareness was found with intention to adopt AI. But the impact of insecurity and discomfort on AI adoption was not significant.Originality/valueThis research provides insights regarding intention to use AI among LIS professionals in Pakistan within the framework of technology readiness index.
