Brianna felt something cold move through her.
Years earlier, Naomi Carter had said employees should never have to depend on whether one supervisor believed them.
Now there was no supervisor to argue with.
Just a yellow box.
Brianna asked:
“So a computer cut my hours?”
“No.”
“The scheduling system prioritizes workers.”
“That sounds like yes.”
The coordinator lowered her voice.
“I didn’t build it.”
Brianna left angry.
But unlike the temporary workers before Naomi’s intervention decades earlier, Brianna had a direct appeal channel.
She opened the employee kiosk.
Selected:
Scheduling / Opportunity Concern
Then typed:
My shifts were reduced because of a reliability score. I want to know what information produced that score and how I can challenge it.
The system routed her complaint automatically to corporate workforce compliance.
That mechanism existed because of reforms originally created after Dayton.
The people who designed it had no idea it would one day be used against an algorithm.
Within forty-eight hours, the complaint landed on the desk of Carter Manufacturing’s current Chief People Officer, Elena Ruiz.
Elena had never met Naomi Carter.
She had joined the company twelve years after Naomi retired.
She had watched the famous break-room clip only once during executive onboarding and found it uncomfortable.
What mattered more to her was the policy framework built afterward.
Complaint visibility.
Payroll transparency.
Reviewable discretion.
No local suppression.
So when Brianna’s complaint arrived, Elena did not ask:
“Is this discrimination?”
Not yet.
She asked:
“What exactly happened?”
Her team pulled the scheduling data.
Then they found Brianna was not alone.
Several Dayton temporary workers had received reduced shifts after being downgraded by the Reliability Index.
That was not automatically wrong.
Predictive scheduling existed to reduce no-shows.
The plant needed workers who would arrive.
The question was how the prediction worked.
Elena requested model documentation.
The vendor sent seventy-two pages.
Machine-learning features.
Historical shift completion.
Late arrivals.
Accepted overtime.
Transportation-distance estimate.
Prior short-notice cancellations.
Staffing-agency tenure.
Availability consistency.
Then one variable caught Elena’s attention.
Transit Dependency Probability.
She read it twice.
The model did not know whether a worker personally relied on buses.
Instead, it inferred the likelihood from home address, vehicle registration information where available through staffing records, and local commuting patterns.
Workers with higher predicted transit dependency received slightly lower reliability scores.
The vendor explanation was simple:
Public-transit-dependent workers historically had a higher rate of weather-related lateness.
Statistically, that was true in the data used to train the model.
Operationally, the algorithm believed it was being rational.
But Elena kept reading.
Transit dependency correlated strongly with income.
Neighborhood.
And in several factory regions, race.
Not because the system knew anyone’s race.
Because geography and transportation carried history inside them.
Elena called the company’s general counsel.
Then internal audit.
Then an outside data-science team.
By Friday, the issue was on the board risk committee’s agenda.
The current CEO was Thomas Avery, a sixty-one-year-old engineer who had spent most of his career inside manufacturing.
He had no Carter surname.
No family connection.
That mattered.
Thomas entered the meeting carrying the report.
“What’s the actual business impact?”
Elena answered.
“We don’t know yet.”
“Then why are we discussing shutdown?”
“Because we are using an inferred transit-status variable to decide who gets work.”
Thomas looked at the model documentation.
“Does it predict attendance?”
“Yes.”