Overview
5–6 August 2025 — AARO sponsored a two-day workshop with Associated Universities, Inc. About 40 participants from government, academia, and independent research organisations examined how witness narratives are collected, standardised, compared, and analysed.
Key developments
- Selection effects and inconsistent terminology
- Memory error and duplicated cases
- Data infrastructure, interoperability, and reproducibility
Evidence-base findings
The workshop identified inconsistent terminology and reporting formats; missing dates, coordinates, sensor parameters, and environmental conditions; weak retention of original records; uncertain chains of custody; classification barriers; translation and transcription errors; and fabricated, altered, or AI-generated imagery. It recommended common metadata templates, links to weather, air-traffic, astronomical, and other contextual databases, preservation of original files, and human oversight of AI-assisted classification.
Scientific response
Comparable variables and known selection effects are prerequisites for meaningful statistical pattern detection. A report cluster cannot be treated as a physical pattern when reporting rates, witness populations, and data quality vary unpredictably. Narrative evidence can identify dates, locations, and possible corroborating records, but cannot independently establish speed, distance, mass, or acceleration. For example, high angular motion cannot be converted into high physical velocity without a measured range.
Assessment
Methodologically strong and scientifically constructive. Its importance will depend on whether AARO implements common schemas, releases machine-readable datasets, and records the provenance of every image, video, and testimony. Better data architecture is essential, although narrative reports alone cannot establish extraordinary performance or origin.
Research context
This page distinguishes the fact of an official action or public release from claims about what the underlying material proves. Unresolved observations require original files, metadata, provenance, and reproducible analysis before firm conclusions can be drawn.