EAIFCH-Ethical-Framework-for-Cultural-Heritage

Overview This repository contains the complete source code, validation scripts, and supplementary materials for the paper "Green AI for Cultural Heritage: An Ethical Framework with Minimal Environmental Footprint" (submitted to Digital Scholarship in the Humanities, Oxford University Press). EAIFCH v1.1 is the first framework integrating systematic ethical assessment with green computing for cultural heritage digitisation. It combines: (1) A hierarchical cultural sensitivity classification system grounded in CARE Principles and UNESCO standards (7 categories, 4 sensitivity levels) (2) A governance escalation mechanism blocking export for HIGH/CRITICAL sensitivity items when community authority is unclear (GovernanceEscalationError — architectural enforcement of ethics) (3) Three differentiated access tiers (Public / Institutional / Community) with three GPS obfuscation modes (≥10km radius offset, boundary rounding, full removal) (4) Energy-efficient algorithms achieving 78% CO₂ reduction compared to conventional ML approaches (~300,000× less CO₂ than CNN-based pipelines) (5) A Bayesian statistical validation protocol (BF > 10, 10,000-iteration permutation testing) with 91% pattern rejection rate Feasibility demonstrated across three case studies: - Case 1: Medieval Torah Fragment (sensitivity 40.0/100, Medium) - Case 2: Wiradjuri Ceremonial Site (sensitivity 65.0/100, High) - Case 3: Mandinka Griot Recording (sensitivity 52.5/100, High — export blocked) Performance: sub-0.15ms assessment times, 23MB memory footprint, 0.000023–0.000049g CO₂ per assessment. License: GPL v3.0 Python: 3.9+ Tests: 20 passing, 94% coverage