Aviation AI Conscience Governance 360 A Design-Science Framework for Accountable, Safety-Critical and Data-Governed Airline AI

Résumé fourni par la source

Artificial intelligence (AI) is moving from experimental analytics into operational airline decision pathways, including operations control, predictive maintenance, disruption recovery, crew planning, passenger communication, revenue management, safety analytics, procurement and maintenance, repair and overhaul (MRO) logistics. The governance problem is therefore no longer whether AI can support airline performance, but whether AI-enabled recommendations remain safe, lawful, explainable, data-qualified, ethically defensible, economically justified and human-accountable. This article develops Aviation AI Conscience Governance 360 (AICG-360) as a design-science framework for accountable airline AI. The framework treats conscience not as machine morality, but as an executive governance architecture linking evidence, data quality, decision authority, risk thresholds, legal obligations, ethical constraints, economic value and post-decision learning. Drawing on aviation AI guidance, risk-management standards, data-quality models, design-science research and KPI-driven airline governance literature, the paper proposes eight governance dimensions: safety conscience, data conscience, model conscience, human-authority conscience, legal-compliance conscience, ethical conscience, economic-value conscience and institutional-learning conscience. It further specifies decision gates, escalation triggers, KPI domains, accountability roles and dashboard logic for airline boards, accountable managers, safety leaders, CIOs, CDOs, operations executives and compliance officers. The article contributes by reframing airline AI governance as auditable decision execution rather than isolated compliance, model validation or productivity automation; by operationalizing data quality as the epistemic foundation of responsible AI; and by offering a publishable artifact for future Delphi validation, case-study testing and dashboard prototyping.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.