Generative artificial intelligence (GenAI) is rapidly transforming the landscape of mental health and well-being by enabling scalable, personalized, and interactive interventions. Despite growing research interest, the field remains fragmented across psychological, technological, and ethical dimensions. This study maps the scientific landscape of GenAI in mental health through a bibliometric and science mapping analysis.A dataset of 475 peer-reviewed publications (2007–2026) was retrieved from Scopus and Web of Science and analyzed using the Bibliometrix R package. The analysis integrates performance indicators, collaboration networks, conceptual structure mapping, and temporal evolution modeling.The results indicate a rapidly expanding field characterized by a high annual growth rate and a compressed innovation cycle. However, these trends should be interpreted with caution, as growth projections are based on model estimates. Intellectual leadership is concentrated in a limited number of countries and authors, while collaboration networks remain fragmented and weakly interconnected. Conceptually, GenAI is positioned as a bridging construct linking information systems, psychological mechanisms, and behavioral outcomes, rather than a fully consolidated research paradigm. Thematic evolution suggests a shift from traditional psychosocial and clinical frameworks toward more technology-mediated mental health research.The findings highlight critical gaps in theoretical integration, long-term validation, and ethical operationalization. This study contributes by providing an integrated mapping of the field, introducing a conceptual framework positioning GenAI as a bridging construct, and identifying directions for future interdisciplinary research, Information Systems integration, and responsible innovation in mental health care.
Health literacy is increasingly recognized as a critical determinant of health equity, yet dominant frameworks continue to conceptualize it primarily as an individual cognitive capability. Such perspectives inadequately explain how health knowledge is constructed and used in marginalized settings where information circulates through social relationships and informal knowledge infrastructures. This study develops a relational model of health literacy that conceptualizes health knowledge as a socially embedded information practice.Using constructivist grounded theory, qualitative data were collected from a Chenchu tribal community in Andhra Pradesh, India, through in-depth interviews and two focus group discussions involving 37 participants. Iterative coding and constant comparison were used to develop a relational health literacy ecosystem model structured around contextual, causal, and intervening preconditions.Findings show that health literacy emerges through distributed processes of meaning-making across causal preconditions. Health knowledge was constructed through embodied experience, intergenerational dialogue, and community validation rather than individual information processing alone. Trusted intermediaries, particularly ASHA workers, functioned as relational nodes within local knowledge infrastructures, translating and legitimizing health information. Intervening preconditions (including dialogic exchange, feedback loops, and cultural translation) mediated the movement of knowledge across relational contexts.This study contributes to information science by conceptualizing health literacy as a relational knowledge infrastructure and by extending distributed health literacy theory to marginalized settings. The relational health literacy ecosystem provides a theoretically grounded framework for designing culturally responsive and equity-oriented health information systems that move beyond linear models of information delivery.
In the digital age, information has become a key determinant of national strength, economic development, and institutional confidence. However, traditional legal doctrines of sovereignty are becoming increasingly inadequate for regulating data mobility that is not confined to any particular geographical location. This paper proposes a risk-based, responsive framework in which data sovereignty is redefined as a quantifiable, persistently evolving governance potential. The framework integrates benchmarking (SDI), dynamic assessment (DSAM), and institutional coordination. Their integration creates a feedback-driven system, which connects national preparedness (macro-level) to (micro-level) dataset risk management. Institutionalization of the framework entails establishing a proposed National Data Sovereignty Authority (NDSA) to coordinate, certify, and adjust policies.A comparative study of the EU, US, BRICS, OECD, and GCC settings reveals the model's flexibility and global applicability. Building on this insight, the model defines sovereignty as dynamic rather than static and, by providing evidence-based guidance, gives governments and industries a systematic way to secure interests, build interoperability, and establish trust in cross-border data flows.
Understanding how firms allocate attention to industrial policies is crucial for advancing sustainable industrial transformation. However, existing approaches to identifying corporate policy attention are constrained by the lack of domain-specific semantic alignment in general-purpose large language models (LLMs). To address this limitation, this study develops a domain-adaptive large language model framework for enhancing policy-attention theme identification in the new energy vehicle (NEV) sector, a representative policy-driven emerging technological sector. The framework involves constructing an instruction-based dataset reflecting the nuances of corporate policy attention in the NEV domain, applying Low-Rank Adaptation fine-tuning to a general large language model to obtain the domain-specific ChatGLM-LoRA-NEVPR model, and conducting comprehensive objective, human, and intelligent evaluations. Experimental results show that the proposed model improves policy-attention recognition accuracy by approximately 12% compared with the baseline, while demonstrating higher semantic coherence and authenticity across multiple dimensions. Overall, this research provides a robust methodological foundation for applying adaptive large language model techniques to analyze corporate policy attention in the NEV industry and other strategically significant, policy-supported industrial contexts.
