The ethical implementation gap in AI-driven disaster risk reduction: empirical evidence from Australia’s multi-sector analysis

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Context While artificial intelligence offers transformative potential for disaster risk reduction, its ethical deployment raises critical concerns when life-critical decisions depend on algorithmic systems with varying levels of accessibility across responsible organizations. This study quantifies AI adoption disparities across Australian DRR stakeholders and identifies institutional factors driving the adoption gap between innovation and operational deployment. A mixed-methods approach combined systematic review of 217 global AI-DRR publications (1999–2024) with qualitative analysis of 14 Australian organizations across government, academic, and NGO/private sectors (February-March 2025). Semi-structured interviews and document analysis examined AI technique adoption, data integration, and downscaling applications, with implementation levels systematically coded on a 1–5 scale following an inter-coder validated protocol. Government agencies demonstrate lower AI adoption (mean score 2.1/5) compared to academic institutions (4.0/5) and NGO/private organizations (3.4/5), representing a 47.5% adoption gap between government and academic sectors (Cohen’s d ≈ 2.0, indicating large practical difference despite small sample size). While academic institutions deploy advanced techniques including GANs and hybrid AI-physics models achieving 45% accuracy improvements, government agencies remain anchored to traditional systems like ACCESS with minimal AI integration. Critical barriers include procurement constraints (18–24-month cycles), computational limitations (10x below requirements), and regulatory mandates for explainable decisions. Academic innovations in AI-based downscaling, improving spatial resolution from 60 km to 10 km, remain inaccessible to operational agencies due to absence of knowledge transfer mechanisms. The study reveals an ethical adoption disparity where organizations with greatest responsibility for community safety show lowest AI adoption, creating accountability voids and equity concerns in disaster protection. This misalignment raises fundamental ethical questions about algorithmic justice, distributive justice, and procedural fairness in public safety contexts. Bridging this gap requires ethical governance frameworks addressing transparency, accountability, and equitable access to life-saving technologies, alongside institutional reforms enabling cross-sector knowledge transfer.