trust modeling

This study addresses the lack of a unified semantic foundation for the concept of trust, which hinders the development of human-machine co-trust systems. To bridge this gap, the authors propose ONTrust, a reference ontology that systematically formalizes the notion of trust—including its types, influencing factors, and risk relationships—grounded in the Unified Foundational Ontology (UFO) and rigorously specified using OntoUML. ONTrust provides a shared semantic basis for domains such as trustworthy AI, enterprise architecture, and requirements engineering. Its validity and generalizability are demonstrated through multiple real-world case studies, showcasing its effectiveness in conceptual modeling, language evaluation, trust management, and human-machine collaboration tasks.

This study addresses the limitations of prevailing AI trust frameworks, which often overlook the subjective, culturally embedded, and relational nature of trust, thereby hindering the development of genuinely inclusive and equitable systems. Moving beyond conventional techno-compliance paradigms, this work introduces African communitarian ethics to reconceptualize trust as a dynamic, temporally situated moral relationship. By integrating relational ethics theory, participatory design, and cross-cultural philosophical perspectives, the research operationalizes a set of trust principles centered on sustained community engagement, cultural sensitivity, and mutual respect through transparency. These principles were empirically validated in healthcare and education contexts, demonstrating enhanced community participation and improved fairness, acceptability, and contextual adaptability of AI systems.

Existing static “human-in-the-loop” classification frameworks fail to capture the dynamic evolution of human-AI trust as AI transitions from tool to collaborative partner—particularly amid foundation model emergent capabilities and multi-agent autonomous goal-setting, where decision authority, autonomy, and accountability exhibit continuous, non-binary evolution. Method: We propose the Human-AI Trust Governance (HAIG) framework, introducing the first dual-objective governance paradigm balancing trust and utility. It establishes a three-dimensional continuous spectrum—decision authority allocation, process autonomy, and accountability configuration—anchored by empirically grounded critical thresholds. HAIG integrates cross-level conceptual modeling, continuity-aware quantification, and threshold-triggered adaptive responses. Results: Evaluated in healthcare and EU regulatory contexts, HAIG demonstrates superior complementarity and dynamic adaptability over traditional frameworks, enabling proactive governance alignment with emergent AI capabilities.

This work addresses the lack of a unified trust mechanism across data, service, and knowledge layers in industrial intelligence, which undermines reliability, accountability, and regulatory compliance. To bridge this gap, the authors propose Trisk—a novel framework that establishes the first holistic view of cross-layer trustworthy governance in industrial settings. Trisk introduces a five-dimensional trust model integrating quality, safety, privacy, fairness, and explainability, supported by a complementary three-dimensional classification scheme. By synergistically combining enabling technologies—including knowledge graphs, zero-trust architectures, causal reasoning, and federated/edge computing—the framework systematically addresses critical gaps in semantic interoperability and policy enforcement. Grounded in a comprehensive review and maturity assessment of over 120 studies, this research provides a systematic roadmap for advancing both theoretical foundations and practical deployment of trustworthy AI in Industry 5.0.

This paper addresses the trust deficit in generative Geospatial Artificial Intelligence (GeoAI) by proposing a three-dimensional trust framework grounded in geographic context: cognitive trust (assessing regional representativeness and cultural adaptability of training data), operational trust (evaluating spatial interpretability and robustness of model functionalities), and interpersonal trust (clarifying developer accountability and multi-stakeholder governance mechanisms). Drawing on geoinformation science theory, explainable AI (XAI) techniques, and ethical governance frameworks, the study conducts conceptual analysis and interdisciplinary inquiry, emphasizing spatial heterogeneity, alignment with regional policy contexts, and dynamic bias mitigation. Its key contribution is the first systematic articulation of a geography-sensitive trust paradigm for GeoAI—explicitly positioning geoinformation scientists as pivotal actors in AI governance—and delivering a theoretically rigorous yet practice-oriented roadmap for trust assessment and cultivation, tailored for researchers, practitioners, and policymakers.

This study addresses the limitations of existing research, which predominantly relies on trust metrics to assess human reliance on AI and fails to capture “appropriate reliance”—the extent to which users judiciously adopt AI advice in suitable contexts. Through a systematic literature review and construct analysis, this work is the first to integrate and distinguish three measurement perspectives of appropriate reliance: traditional, appropriateness-based, and dominance-oriented, thereby clarifying its theoretical boundaries relative to trust and mere dependence. The analysis reveals a fragmented landscape of current measurement approaches and proposes a unified, objective evaluation framework. This framework establishes a reproducible and comparable empirical foundation for future research on human–AI collaborative systems.

This study addresses the governance and accountability challenges arising from third-party deep-embedded digital services, which erode customer visibility and control over data security. Drawing on organizational trust theory and agency theory, the authors analyze the OpenAI-Mixpanel security incident to introduce the concept of “transitive trust” and propose a “fortress-gatekeeper” analytical framework that delineates cybersecurity governance boundaries centered on trust and data flows. Through qualitative theoretical construction via document analysis, the research formulates four propositions concerning vendor integration, metadata exposure, vendor assurances, and data diffusion. These insights advance novel governance strategies—including vendor tiering, data classification, and continuous assurance—thereby offering both theoretical grounding and practical guidance for managing third-party cybersecurity risks.

This study addresses the challenge enterprises face in effectively governing and verifying compliance of emergent AI systems, which creates a trust gap between regulatory expectations and operational capabilities. To bridge this gap, the work proposes a continuous, autonomous AI governance architecture that translates compliance requirements into real-time telemetry mechanisms at the operating system layer. By leveraging a zero-trust telemetry boundary, ephemeral read-only probes, and AI observability agents—integrating LangSmith and Datadog LLM telemetry—the approach automatically discovers AI systems, collects control assertions, and continuously generates verifiable evidence without accessing source code or sensitive payloads. Validated against major regulatory frameworks including ISO/IEC 42001, the EU AI Act, SOC 2, GDPR, and HIPAA, this method enables a fundamental shift from document-based policy trust to empirically grounded architectural trust.

This study addresses the dynamic interplay of mutual trust and distrust between humans and artificial intelligence in AI governance, challenging conventional unidirectional models that treat AI solely as an object of trust. It proposes that AI can also function as an agentic subject capable of exercising trust or distrust. By integrating perspectives from philosophy, governance theory, human-computer interaction, and sociotechnical systems, the work develops an analytical framework that uncovers the core tensions and structural dilemmas this bidirectional mechanism generates within AI regulation. The research provides a theoretical foundation for understanding the emerging politics of trust in AI governance and identifies critical challenges that future institutional designs must confront.

This study addresses the trust imbalance that arises when users interact with increasingly prevalent yet opaque autonomous systems, often due to an inadequate understanding of their capabilities and limitations. Building upon the Human-Computer Trust Scale (HCTS), this work proposes the first practice-oriented, context-sensitive explanatory framework that enables reflective interpretation of trust dispositions. Through empirical validation, the research not only confirms the efficacy of HCTS as an initial trust assessment instrument but also introduces context-aware calibration guidelines for aligning user trust with system performance. The resulting framework provides both theoretical grounding and practical support for dynamically regulating trust in human–computer interaction.