An Ethical Framework and Explainable AI in Geospatial Intelligence

De-identified experimental data from a between-subjects study evaluating the CLEAR (Contextualized, Legible, Equitable, Auditable, Responsible) ethical framework for AI-enabled geospatial intelligence (GEOINT). Fifty GEOINT analysts completed four imagery analysis scenarios using either a baseline OversightML workflow or a CLEAR-enhanced workflow integrating spatially aware explainability overlays, bias indicators, and provenance tracking. The dataset includes participant demographics, primary outcome measures (task completion time, detection accuracy, contextual relevance, subjective confidence, NASA-TLX cognitive load), NASA-TLX subscale scores, per-scenario results, sensitivity and specificity calculations, qualitative coding frequency tables with inter-rater reliability metrics, ground truth validation metrics, system performance benchmarks, and the complete semi-structured interview protocol.