“Damage. Evidence. Whether she thought litigation would fix something.”
“She sued the city.”
“Eventually.”
“Why not sue everyone?”
“Because lawsuits are tools.”
“Not religion.”
The boy frowned.
“Did she say that?”
“No.”
“But she would have.”
Probably.
The son returned to the electronics shop months later because it had the computer component he needed.
His mother asked whether he was sure.
“Yes.”
He bought it.
The guard did not recognize him.
He came home.
“That was weird.”
“What?”
“I thought I’d feel something.”
“Did you?”
“Not really.”
She smiled.
“Good.”
Ordinary again.
Meanwhile, the old retailer that had employed the boutique manager changed ownership twice.
Its reform policies survived both acquisitions because lawyers had converted them from scandal-specific promises into formal compliance standards.
The names changed.
The reasons blurred.
New executives had no personal memory of the case.
They only knew that certain practices existed because failing to follow them increased risk.
That was not poetic.
It was durable.
One new regional director tried to weaken the independent video-retention requirement because cloud storage was expensive.
A compliance analyst objected.
“Why do we keep this?”
The director asked.
The analyst pulled historical incident report.
Not the full boutique video.
A summary.
Fabricated theft reports.
Manual deletion.
Civil settlements.
Criminal convictions.
The director read.
Then approved budget.
No speech.
The policy survived another year.
Twenty years after the architect’s death, the retailer developed an artificial-intelligence loss-prevention system.
The vendor promised predictive detection.
Executives loved the presentation.
Cameras could identify “behavioral anomalies.”
Movement patterns.
Repeated visits.
Shelf interaction.
Bag placement.
Compliance staff immediately raised concern.
One analyst pulled old profiling case from archive.
The system’s training data had learned patterns from historical human suspicion.
That included biased reports.
If old managers disproportionately flagged Black shoppers, the algorithm could learn that those behaviors were more “suspicious.”
The company paused rollout.
External researchers audited.
They found disparities.
The vendor insisted model was race-blind.
The auditors answered that variables need not include race to reproduce racial patterns.
The company retrained.
Changed thresholds.
Required human corroboration.
Published limitations.
The old architect never saw any of this.
But the logic of her case remained.
Data could look objective while inheriting human choices.
A machine did not eliminate bias simply by moving it behind mathematics.
The family learned about the new audit through a newspaper article.
The granddaughter sent it to her mother.
Her mother replied:
Your grandmother would have hated the phrase behavioral anomaly.