ethical risk assessment

Current AI ethics assessments are fragmented, focusing predominantly on fairness, transparency, privacy, and trust at the model or output level while neglecting inter-component system interactions, real-world harm contexts, and causal harm propagation pathways—resulting in evaluations disconnected from actual risk scenarios and lacking actionable thresholds. Method: Through a scoping review synthesizing nearly 800 ethics metrics, this study constructs the first four-dimensional relational framework—“System Components–Attributes–Risks–Harms”—to systematically map ethical assessment dimensions. Contribution/Results: The framework uncovers three critical gaps: insufficient system integration, weak contextual embedding, and poor actionability. It advances AI ethics evaluation from isolated metric measurement toward a systemic, traceable, and intervention-oriented paradigm—enhancing regulatory alignment and practical deployment efficacy in industry settings.

Market fundamentalism and demographic imbalances among software practitioners jointly impede the integration of ethics in software development. Method: A mixed-methods study—comprising a survey of 217 practitioners across roles, industries, and countries, supplemented by qualitative analysis—provides the first empirical evidence that marginalized groups (women, BIPOC, and persons with disabilities) exhibit significantly higher ethical sensitivity, frequency of ethical issue reporting, and willingness to intervene than their majority-group counterparts—challenging the “neutral developer” assumption and revealing structural demographic bias in ethical advocacy. The study identifies two primary barriers: market-driven organizational cultures that suppress ethical deliberation, and widespread institutional deficits—including absent ethical processes, insufficient authority delegation, and inadequate ethics training. Contribution/Results: It establishes demographic background as a critical analytical dimension for understanding variation in ethical practice and provides empirically grounded foundations for designing inclusive, equity-oriented ethics governance mechanisms in software engineering.

This study investigates cross-role (e.g., engineers, product managers, ethics specialists) and cross-national (43 countries, N=414) variations in AI ethics awareness, policy comprehension, and risk mitigation practices within AI development teams. Employing a mixed-methods design, it integrates large-scale surveys with in-depth interviews, combining quantitative statistical analysis and qualitative thematic coding. Results reveal a pronounced role-based ethical responsibility gap and a non-uniform global distribution of regulatory sensitivity and implementation capacity. Building on these findings, the study proposes a “collaborative, role-sensitive ethics governance framework” that mandates multi-stakeholder engagement across the AI lifecycle and incorporates localization mechanisms for contextual adaptation. This framework advances AI ethics practice from prescriptive, one-size-fits-all guidelines toward inclusive, situated governance—offering an actionable, differentiated pathway for global AI policy implementation and responsible innovation.

This work addresses the current lack of a systematic framework for evaluating ethical risks in data collection practices for large language models (LLMs). It proposes the first quantifiable assessment framework that integrates multiple prominent ethical theories, structuring evaluation around core ethical principles through a set of targeted questions and establishing a scoring system to measure ethical risk. This approach enables systematic, quantitative ethical auditing of LLM data curation processes. By offering a practical tool for assessing ethical compliance in AI development, the framework fills a critical gap in existing research—particularly in the integration of diverse ethical theories and the empirical evaluation of real-world data practices—thereby advancing the responsible development of artificial intelligence.

This study systematically examines the multidimensional ethical challenges and cross-domain governance dilemmas arising from real-world deployments of generative AI—particularly large language models (LLMs). Method: Through a systematic literature review (SLR) and thematic coding, we structurally map 39 empirical studies using an original five-dimensional ethical framework. Contribution/Results: Our analysis uncovers a fundamental tension between the dynamic evolution of ethical risks and the persistent lag in governance responses—a finding not previously documented. We demonstrate that existing mitigation strategies exhibit severe adaptive deficits in high-stakes domains such as healthcare and public administration, stemming from misalignment among technological development, ethical reasoning, and institutional evolution. To address this, we propose a tripartite co-evolutionary pathway integrating ethics, technology, and institutions, offering both theoretical grounding and an actionable framework for resilient governance of generative AI.

This study addresses the frequent yet unsubstantiated claims of “AI for good” by proposing Impact-AI, a novel method that operationalizes public benefit and sustainability as dual pillars of AI impact assessment. Through qualitative interviews with diverse stakeholders, the approach systematically audits an AI project’s governance structure, theory of change, data characteristics, and its social, environmental, and economic consequences. Impact-AI delivers a structured, reusable, and civil society–oriented evaluation framework accompanied by standardized public reporting formats. By rendering the evaluation process transparent and participatory, the method significantly enhances the credibility of “AI for good” assertions and enables broader public deliberation on the societal implications of artificial intelligence.

This study addresses the absence of unified and transparent research ethics guidelines in top-tier security and privacy conferences, which has led to ambiguous review criteria and inconsistent enforcement, thereby hindering the community’s ethical awareness. Through a systematic analysis of ethics policies across four leading conferences over multiple years and semi-structured interviews with 20 researchers, this work presents the first comprehensive account of the evolution of ethical practices in the field, identifying a critical gap in ethics education as the primary bottleneck. Drawing on qualitative findings and principles of community-based participatory design, the paper proposes an innovative framework featuring an inter-conference coordination mechanism and an open Ethics Wiki. It delineates current progress and key barriers to consensus-building and has already launched the Ethics Wiki as an initial step toward collaborative governance.

This study addresses the frequent neglect of environmental impacts in computationally intensive research—such as artificial intelligence—due to ambiguous ethical review policies. It presents the first systematic framework integrating environmental sustainability into the ethical oversight of computational research. By delineating clear review boundaries, establishing evidentiary standards, and developing researcher self-assessment tools, the framework enables institutional ethics committees to effectively evaluate the environmental costs of proposed projects. This approach provides actionable guidance for ethical review processes and encourages researchers to proactively consider the ecological footprint of their work during early design stages, thereby addressing a critical gap in current research ethics frameworks concerning sustainability.

Current generative AI systems lack systematic approaches to detect latent ethical harms—such as intellectual property violations or harmful content—in their outputs. This work proposes a novel paradigm termed “ethical testing,” introducing the first ethics-oriented testing framework specifically designed for generative AI. Distinct from conventional fairness testing, this framework focuses on identifying software harms stemming from unethical behaviors. The approach integrates large language models, automated test case generation, and ethical risk modeling to establish a scalable testing methodology. Through five case studies, the framework demonstrates its effectiveness in uncovering real-world ethical risks across diverse generative AI systems.

Ethical rule conflicts in AI systems often exhibit fine-grained, context-dependent incompatibilities that are difficult to resolve systematically. Method: This paper proposes a three-stage dynamic trade-off decision framework: (1) proactive identification of contextualized ethical conflicts; (2) priority ranking via multi-dimensional weight modeling; and (3) generation of traceable, auditable, structured decision rationales. Contribution/Results: It introduces the first systematic taxonomy of five ethical trade-off pathways, integrating ethical impact assessment, regulatory compliance alignment, and documentation traceability—thereby jointly ensuring contextual adaptability, explanatory transparency, and regulatory compatibility. Empirical evaluation demonstrates significant improvements in ethical decision transparency and auditability. The framework provides organizations with a scalable, scenario-adaptive pathway for deploying responsible AI systems.