Definition
Quality of worker behavioral response to AI-generated FME monitoring alerts, measured as the percentage of alerts receiving a response that meets the plant's FME procedure requirements within the required response time, tracking whether false positive volume has conditioned workers to under-respond.
The Measurement Gap
Foreign material exclusion (FME) monitoring systems using computer vision generate alerts when potential foreign material is detected in controlled areas. The procedurally required response to an FME alert is immediate: stop work, inspect the area, identify the source, notify the FME coordinator. But when the monitoring system has a high false positive rate, workers learn to perform cursory checks or simply acknowledge the alert without investigation. No legacy metric measures the quality of worker response to FME alerts. FME event rate tracks confirmed FME incidents. FME zone compliance tracks procedural adherence at entry and exit points. Neither measures whether workers respond to monitoring alerts with the thoroughness the procedure requires. The invisible failure mode is that alert fatigue from false positives degrades the behavioral response to genuine alerts, and the next real FME event in containment or the reactor cavity is inadequately investigated because the worker has been conditioned by hundreds of false positives to perform a cursory glance rather than the procedurally required investigation.
Decision Supported
Are FME monitoring alerts receiving the procedurally required response, or has the false positive rate conditioned workers to perform cursory checks or ignore alerts?
Formula
Plain language: For each AI-generated FME alert, evaluate whether the worker response met three criteria: (1) response within the required timeframe, (2) response scope met procedural requirements (area inspection, not just acknowledgment), (3) appropriate escalation to FME coordinator if warranted. Score each response as Full (all three criteria met), Partial (some criteria met), or Inadequate (acknowledgment only or no response).
Notation:
FARQI = (Full + 0.5 × Partial) / N_alerts
Variables:
- Full = FME alerts receiving a response that met all three criteria (count)
- Partial = FME alerts receiving a response that met some but not all criteria (count)
- Inadequate = FME alerts acknowledged without required response scope, or not responded to within the required timeframe (count)
- N_alerts = total FME monitoring alerts in the measurement period (count)
Sign convention: FARQI = 1.0 means every alert received a full procedural response. Lower is worse.
Aggregation period: Per outage (most FME-critical work occurs during outage), with rolling quarterly for online operations.
Edge Cases
- Alert determined to be a definite false positive (camera artifact, lighting change, etc.): still evaluate the response quality. The worker doesn't know it's a false positive at the time of the alert, and the procedurally correct response is to investigate regardless. The alert quality problem (false positive rate) is measured by a companion metric; FARQI measures response behavior regardless of alert accuracy.
- FME-critical area sealed (containment, reactor cavity): alerts in these highest-consequence areas should be weighted separately from general FME areas
- Alert during time-critical activities (critical path outage work): track response quality separately for time-pressured vs. normal-pace activities: the concern is whether time pressure degrades FME response
- Minimum sample: all FME alerts in FME-critical areas (census)
Proposed formula, pending validation pilots.