Definition
Degree of equitable dose distribution across the work crew under AI-optimized ALARA plans, measuring whether the optimization shifts disproportionate dose to a subset of experienced workers while reporting improved collective metrics.
The Measurement Gap
AI-optimized ALARA planning minimizes collective radiation exposure (CRE), the metric the NRC and plant management track. But CRE is an aggregate number. An optimizer that reduces CRE by concentrating dose on the most experienced workers (who can complete tasks faster, reducing total exposure time) will show improved collective performance while creating an individual dose equity problem. No legacy metric measures dose distribution equity. CRE measures collective dose. Individual dose records track each worker's accumulation against administrative limits. Neither connects the distribution pattern to the AI optimization strategy. The invisible failure mode is that the AI ALARA plan consistently assigns the highest-dose tasks to the same experienced personnel because they are the fastest, reducing CRE while driving individual workers toward administrative limits prematurely. The collective metric improves while the individual burden becomes inequitable and unsustainable.
Decision Supported
Is the AI ALARA optimizer concentrating dose on the plant's most experienced personnel in ways that will trigger administrative dose limits prematurely, or is it distributing dose equitably across qualified workers?
Formula
Plain language: Compute the Gini coefficient of individual dose receipt across all radiation workers assigned to AI-optimized work plans during the period. A Gini of 0 means perfectly equal dose distribution. A Gini of 1 means one person received all the dose. Compare to the Gini under pre-AI ALARA planning (baseline).
Notation:
ADEI = G_current / G_baseline
Where:
G = (Σ_i Σ_j |dose_i − dose_j|) / (2N² × mean_dose)
Variables:
- dose_i, dose_j = individual accumulated dose for worker i and worker j during the measurement period (mrem or mSv)
- N = total number of radiation workers in the measurement population (count)
- mean_dose = average individual dose across the population (mrem or mSv)
- G_current = Gini coefficient for the current AI-optimized period
- G_baseline = Gini coefficient for the pre-AI or first-year baseline period
Sign convention: ADEI = 1.0 means equity is unchanged from baseline. Values above 1.0 mean dose is becoming more concentrated (bad: AI is creating dose inequality). Values below 1.0 mean dose is becoming more equitable (good).
Aggregation period: Per outage (most dose is accumulated during outage activities) and rolling 12-month for online operations.
Edge Cases
- Some workers legitimately receive more dose due to specialized skills (welders, ISI technicians): segment ADEI by worker category. Compare welders to welders, not welders to administrative HP staff. The concern is within-category concentration, not between-category differences.
- Workers approaching administrative dose limits: track the count of workers above 80% of their annual administrative limit as a companion indicator. ADEI measures distribution, the administrative limit proximity count measures the consequence
- Very small crews (under 10 workers): Gini coefficient becomes unstable. Use coefficient of variation (CV) as an alternative for small populations
Proposed formula, pending validation pilots.