consent management

This study addresses the limitations of traditional individualized “click-to-consent” mechanisms in data governance, which suffer from low feasibility, inadequate protection against collective data harms, and insufficient informed consent. To overcome these challenges, the paper proposes replacing individual consent with collective consent and innovatively develops a deliberative democracy–driven “Consent Assembly” model. By integrating speculative design, backcasting, and mini-public deliberation, the authors construct a systematic theoretical framework for collective consent. This framework not only expands the paradigmatic boundaries of data governance but also demonstrates practical promise in two key applications: first, as a viable alternative to the prevailing notice-and-consent regime, and second, as a pathway for collectively authorizing data use in generative artificial intelligence training, thereby exhibiting both feasibility and forward-looking potential.

Scalable informed consent management in Digital Public Infrastructure (DPI) faces challenges including ambiguous ownership representation, uncontrolled data flows, and difficulties in multi-stakeholder coordination. This paper proposes the first systematic abstraction model that tightly couples “data ownership forms” with “data flow paths,” formally defining ownership primitives and compliance-driven data flow control mechanisms. Leveraging open technical standards, we design an architecture enabling dynamic, fine-grained, end-to-end auditable consent management. Our contributions are threefold: (1) the first semantic alignment between ownership structures and data流转 trajectories; (2) simultaneous preservation of individual autonomy, public value generation, and regulatory compliance (e.g., GDPR); and (3) robust, auditable data sharing across large-scale, multi-party DPI deployments.

This study examines the ethical and legal challenges arising from AI-driven data collection during 2023–2024, identifying core risks—including absent informed consent, amplified algorithmic bias, and systemic privacy erosion—across healthcare, finance, and smart city domains. It comparatively analyzes regulatory approaches in the EU, U.S., and China. Methodologically, it integrates policy text analysis, cross-jurisdictional compliance mapping, empirical case studies, and multi-stakeholder Delphi consultation. The study proposes a three-dimensional adaptive governance framework comprising *legal alignment*, *technical safeguards* (e.g., embedded differential privacy), and *dynamic ethical assessment*. Its key contributions include advancing context-sensitive regulation and fostering transnational standardization for AI data governance; it delivers an actionable AI data governance roadmap, already adopted by three international digital ethics working groups and informing the design of two regional AI regulatory pilot programs.

This study addresses privacy risk governance in adolescent AI applications, examining divergent privacy perceptions between parents/educators and AI experts, and their underlying causes. Method: Drawing on 210 valid survey responses, we construct the first cross-stakeholder privacy cognition structural model—comprising five constructs: risk perception, data control rights, transparency, trust, and education awareness—and employ partial least squares structural equation modeling (PLS-SEM) for quantitative causal analysis. Results: Education awareness significantly enhances risk identification capability; data control rights emerge as a pivotal antecedent driving both transparency and trust; conversely, parents’ data-sharing behaviors are predominantly influenced by exogenous factors outside the model. The findings provide theoretical grounding and design implications for reconciling AI innovation with robust privacy protection for adolescents.

This paper addresses LLM-specific ethical challenges—hallucination, unverifiable accountability, and decoding-based censorship complexity—distinct from general AI concerns such as privacy and fairness. Methodologically, it first defines LLM-exclusive ethical dimensions, constructs a domain-specific ethical framework, and proposes a dynamic context-adaptive auditing mechanism. Leveraging an ethical problem taxonomy, interdisciplinary governance modeling, and end-to-end risk analysis, the study delivers a responsibility-oriented governance roadmap spanning development, deployment, and regulation. Key contributions are: (1) a systematic mapping of LLM ethical risks across technical, operational, and societal layers; (2) advancement of the “ethics-by-design” paradigm, ensuring tight alignment between technological evolution and governance mechanisms; and (3) provision of an actionable, interdisciplinary governance pathway to foster trustworthy LLM development and deployment.

This study addresses the limitations of current data governance frameworks, which overly emphasize compliance and risk mitigation at the expense of enabling responsible cross-organizational data reuse for public benefit. To overcome this inward-looking paradigm, the paper proposes a novel institutional function—strategic data stewardship—centered on ecosystem collaboration and public value creation. It introduces an actionable Data Stewardship Canvas to operationalize this approach, integrating institutional design, governance principles, and practical mechanisms. The framework articulates core principles, roles, and capability models tailored to support real-world implementations in data collaboratives, data spaces, and data commons. By doing so, it aims to establish trustworthy, lawful, and efficient pathways for data reuse in the AI era, effectively bridging the gap between data availability and actual accessibility.

This study addresses the tension between public interest and individual privacy protection in government data openness. Methodologically, it proposes a contextualized, multi-tiered balancing framework featuring a four-level privacy risk assessment mechanism, differentiated decision-making pathways for data access versus secondary use, and a suite of disclosure modalities—including anonymized publication, permissioned access, and sandboxed environments—integrated with a context-sensitive checklist grounded in Fair Information Practice Principles. Its key contribution lies in moving beyond the binary “all-or-nothing” openness paradigm by establishing a public-interest justification requirement for personal data disclosure, thereby unifying privacy impact grading, data classification, and regulatory compliance review into an actionable, auditable governance tool. The framework has been piloted across multiple municipal data platforms, demonstrating significant improvements in the legality, contextual appropriateness, and transparency of disclosure decisions.

Current healthcare data governance lags behind advances in AI and computational technologies; rigid, privacy-first paradigms stifle innovation while failing to robustly safeguard patient rights. Method: This paper proposes a novel, context-sensitive data stewardship framework grounded in socio-technical ethics, centered on the principles of autonomy, beneficence, non-maleficence, and justice—moving beyond static compliance logic. It integrates ethical analysis, cross-national policy comparison, data governance theory, and empirical case studies from medical AI. Contribution/Results: The framework reorients governance toward patient-centeredness, enabling simultaneous advancement of technological innovation, clinical efficacy, equitable access to care, and just allocation of health resources—all while upholding privacy and fairness. It provides a practical, ethically grounded foundation for data-driven healthcare governance, offering actionable guidance for policymakers, clinicians, and AI developers in real-world implementation.

This study investigates how to enable patients to exercise fine-grained control over the sharing of their de-identified health data while simultaneously preserving research integrity and addressing the divergent expectations of multiple stakeholders regarding privacy, transparency, and ethical considerations. Through semi-structured interviews with 16 healthcare system leaders, a large-scale survey of 523 patients, and the development of a high-fidelity web-based prototype, the research uncovers fundamental discrepancies between patients and administrators in their perceptions of data control, risk, and transparency. The work proposes context-aware, multi-literacy-adaptive design principles that support flexible granularity in data authorization and promote ongoing, benefit-oriented transparency mechanisms, thereby offering both theoretical insights and practical guidance for building a trustworthy, user-centered health data governance ecosystem.

Ethical regulations in data management are highly context-dependent, and existing approaches struggle to achieve dynamic, cross-context compliance. Method: This paper proposes a context-aware conceptual model for ethical data management, introducing a novel dual-layer structure comprising a Context Dimension Tree (CDT) and an Ethical Requirement Tree (ERT). This framework enables structured modeling, cross-scenario mapping, and dynamic adaptation of ethical constraints. Contribution/Results: It represents the first systematic integration of context sensitivity into ethical data governance, shifting the paradigm from static rule enforcement to context-driven ethical reasoning. Through conceptual modeling and illustrative application scenarios, the model demonstrates significantly enhanced guidance for ethical compliance during the preprocessing phase of data analytics and learning systems. It exhibits strong scalability and practical feasibility, offering a robust foundation for adaptive, context-responsive ethical data stewardship.