Definition
Defect detection rate of AI-assisted remote surveys compared to physical surveyor attendance.
The Measurement Gap
What This Metric Measures
RSDDE measures the ratio of defect detection rate during AI-assisted remote surveys to defect detection rate during physical surveyor attendance surveys, compared on equivalent equipment and structural scopes. It quantifies whether the industry's move toward remote surveys with AI-assisted defect highlighting produces equivalent safety outcomes to traditional physical attendance.
The Decision It Supports
Is the AI-assisted remote survey finding as many defects as a physical survey, and is it finding the same types of defects, or is it systematically missing categories that require physical inspection? The decision to accept remote survey as equivalent to physical attendance is a risk decision with class and flag state implications. RSDDE provides the evidence base for that decision.
Why Legacy Metrics Cannot See This
Legacy survey metrics track deficiencies found per survey, survey completion rates, and survey timing compliance. When a remote survey finds 8 defects and a physical survey on a comparable vessel finds 12, the remote survey reports its 8 defects as the survey result. Nobody compares the two. The assumption that remote plus AI equals equivalent detection is embedded in the decision to allow remote surveys, but the equivalence is never validated. RSDDE provides the validation by systematically comparing detection rates across matched remote and physical surveys, segmented by defect type so the organization knows exactly which categories remote surveys capture well and which they miss.
Formula
Primary Formula
RSDDE = (Defects_remote / Scope_remote) / (Defects_physical / Scope_physical)
Variable Definitions
- Defects_remote = defects found during AI-assisted remote surveys (count, weighted by severity)
- Scope_remote = survey scope units covered in remote surveys (number of inspection points or area surveyed)
- Defects_physical = defects found during physical surveyor attendance surveys (count, same severity weighting)
- Scope_physical = survey scope units covered in physical surveys
Units: Ratio. A value of 1.0 indicates equivalent detection effectiveness. Values below 1.0 mean remote surveys find fewer defects per unit of scope. Values above 1.0 mean remote surveys find more, which is possible if AI detects subtle defects the human eye misses in certain conditions.
Edge Cases
Direct comparison requires matched survey scopes: compare remote and physical surveys on the same vessel type for the same survey type to control for vessel condition and survey scope variables. Some defects are only detectable physically: loose fittings, thin plate identified by touch, weld quality assessment requiring close proximity. Segment by defect type to identify which categories remote surveys handle equivalently and which they cannot. Minimum sample of 10 matched remote and physical survey pairs per measurement period.
Proposed formula, pending validation pilots.