Screening Throughput-Sensitivity Trade-off Index (STSTI)

GMI-AVI-022AviationInfrastructure-Dependent

Definition

Correlation between AI-driven screening lane throughput and detection sensitivity by lane type: measuring whether faster AI-optimized lane configurations have lower threat detection rates than slower configurations. A negative correlation indicates the AI is trading security effectiveness for processing speed.

The Measurement Gap

The measurement gap STSTI addresses: aviation security checkpoint management tracks throughput (passengers per hour, wait times, lane efficiency scores) and security performance (detection rates from covert testing), but always as separate reporting streams with no systematic analysis of the relationship between them. An airport can report both high throughput and high detection rates without either metric revealing whether the high-throughput lanes are the same lanes producing the high detection rates.

AI-driven lane optimization applies throughput pressure to all lanes simultaneously. When the optimization works, it finds configurations that improve speed without sacrificing detection. When it trades one for the other, that trade-off appears in neither the throughput report nor the detection rate report: only in the correlation between the two. STSTI is the metric that closes this gap.

The decision STSTI supports: should minimum detection sensitivity standards be established per lane configuration, constraining what the throughput optimizer can change? Should certain lane configurations be excluded from throughput optimization because they serve security functions where speed should not be a consideration? Should the AI system be required to demonstrate detection-neutral throughput improvements before deployment? These policy decisions require evidence that a trade-off exists, and STSTI is the metric that produces that evidence.

Formula

Plain language: For each screening lane configuration type (automated vs. manual, CT vs. X-ray, AI-assisted vs. standard, specific staffing combinations), measure both throughput and detection sensitivity from the same measurement period. Compute the correlation across lane configurations. A negative correlation indicates higher-throughput configurations have lower detection sensitivity: the trade-off is real.

Notation:
STSTI = ρ(Throughput_lane, Detection_sensitivity_lane) across lane configurations

Variables:

  • Throughput_lane = passengers processed per hour by lane configuration type (count/hour)
  • Detection_sensitivity_lane = threat detection rate by lane configuration type from covert testing (percentage)
  • ρ = Pearson or Spearman correlation coefficient (use Spearman if the relationship is non-linear or if sample sizes are small)

Units: Correlation coefficient, -1 to +1. Negative = higher throughput correlates with lower sensitivity (trade-off exists). Zero = no relationship. Positive = higher throughput configurations also have better detection (optimization is working on both dimensions).

Lane configuration definition precision: "AI-assisted CT lane" is an insufficient configuration definition. Specify equipment type, AI system version, staffing level, and TSO experience category. "AI-assisted CT lane with experienced TSO" and "AI-assisted CT lane with probationary TSO" are different configurations that should be analyzed separately.

Confounding variable controls: lane staffing quality, time of day, peak vs. off-peak passenger volume, and passenger mix (frequent vs. infrequent travelers) all affect both throughput and detection independently. Include these as covariates in the analysis. When sample sizes are insufficient for full regression, stratify by time-of-day and report STSTI separately for peak and off-peak periods.

Proposed formula, pending validation pilots.

At a Glance

Category
Security and Compliance AI Calibration
← Fuel Tail-Risk Calibration Index (FTRCI)

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment