The operator’s name was Xu Lihua, and she had been told the new schedule was only a trial.
That was why she refused it.
The proposed shift would move a short production run past midnight, when the workshop could purchase power more cheaply. The scheduling model had found enough capacity and expected no conflict with maintenance. The plan looked ordinary until Lihua reached the fourth line: the night team would have to clear a container of unfinished parts before the line restarted.
“Who is counting the people?” she asked.
The project manager pointed to the staffing field. Six operators, the usual number.
“Not who is assigned. Who is actually coming?”
Two of the six had been moved to emergency repair work. A third had requested leave for a family appointment. The planning system had received the approved monthly roster but not the changes made at the previous shift meeting. The manager said the remaining workers could probably manage a shorter batch.
“Probably is not the staffing plan,” Lihua replied.
She pressed decline on the trial screen and selected insufficient verified coverage. There was a blank field for explanation. She wrote the three absences and the task that would be left without an assigned person.
The screen asked her to confirm the reason would be visible to the project review group. She checked the roster a second time before submitting it. She had complained informally about staffing before; a recorded refusal would be harder for her supervisor to forget and harder for her to disown.
The screen accepted the entry. Her supervisor did not.
By lunchtime he had asked why she was obstructing a study that would improve the department’s budget. He reminded her that no machinery would be changed without a final approval. She answered that the approval meeting was at three, and by then the proposed roster would be copied into the handover instructions. A trial that trained people to expect an unsafe arrangement could become an operating decision before anybody called it one.
Ning heard about the disagreement from the site’s technical liaison. Henglan’s contract described a right to decline recommendations without retaliation for a documented safety or staffing concern. Until now the paragraph had been cited mostly in presentations.
She went to the workshop with Yanchuan rather than calling the supervisor from Shenzhen. Lihua was working her regular afternoon shift. She greeted them without enthusiasm.
“I don’t need visitors telling me I am brave,” she said. “I need the night team to have the people they require.”
Ning asked whether they could review the specific workflow with the employer’s safety representative present. Lihua agreed after the representative arrived. The supervisor joined them and repeated that the software was advisory.
“Then my refusal should have ended the proposal,” Lihua said.
“Not the discussion,” the supervisor replied.
“Discussion is fine. You told me not to make another record.”
He objected to the wording, then admitted he had asked staff to take disagreements directly to him rather than recording them in the system. He was concerned that repeated refusals would make his department appear uncooperative to management.
Yanchuan found the weekly metric on the report’s last page. It measured recommendation acceptance rate. Rejections counted against the pilot’s apparent effectiveness, regardless of the reason.
The interface could preserve a refusal. The review process gave people an incentive to hide it.
Ning took out her notebook. She asked the manager to distinguish between a proposal rejected because the software used stale information and one rejected because a worker disliked changing shifts. He began to answer, then stopped.
“Both still delay the study.”
“Then you’re studying cooperation,” she said, “not whether the schedule can be used.”
Lihua went back to her station while the visitors discussed the reporting change. She had already explained the problem and did not intend to spend the rest of her shift proving that she belonged in the room.
The night proposal was suspended. The company confirmed the actual roster, assigned the container-clearing task, and concluded that the planned trial could not be conducted on that date. The cost estimate was revised. Nobody was ordered to work an unauthorized shift to rescue the chart.
Guo was not pleased when he saw the revised report.
“Will every customer read this as a failure?”
“A customer should read it as a rejected schedule,” Ning said.
“That sounds like failure.”
“It may be evidence that the control is working.”
He asked Yanchuan whether Henglan could keep measuring uptake.
“Yes,” Yanchuan said, “if we also measure whether a refusal was resolved correctly and whether people were pressured not to record one. An acceptance percentage alone can’t tell you those things.”
Guo requested a costed proposal rather than another principle. The extra review would require staff time and an external contact for complaints that could not safely be raised through a supervisor. Ning prepared the estimate with the customer’s safety office.
The purchasing officer rejected the first estimate. They had budgeted for software evaluation, not a process for complaints. Ning asked whether the trial would still be voluntary if the only way to refuse was to persuade the person assigning the shift. The officer did not answer. Two days later she approved a smaller review with a fixed end date. Ning accepted it after the safety representative agreed to record unresolved complaints beyond that date. They proposed a limited trial and a clear route for appealing retaliation.
Two days later the site’s union representative requested a meeting. Lihua had not asked for publicity, but other workers had seen the original acceptance metric and wanted assurances about how it would be used. The representative asked for a written prohibition on treating an individual refusal as a performance score.
The employer agreed to revise the pilot rules. The first revised dashboard showed valid trials, rejected trials with verified reasons, and proposals awaiting information. It did not rank operators by willingness to click approve.
That weekend Ning taught another class at the technical college. She gave the students the old and new dashboards without names or workplace details and asked them which one offered more useful evidence.
A young man chose the old version.
“It tells management how often the tool gets used.”
“True,” Ning said. “What decision could you safely make from that number?”
He thought about it. “Whether people are using the tool.”
“Could you decide it should control a night shift?”
“No.”
A student at the back asked what would happen if refusing carried a personal cost. Ning could describe the complaint route and the contractual protection. She could not promise the person would face no unpleasant conversations.
When Yanchuan told his father about the lesson, Jianming asked what Lihua had received for identifying the problem.
“Her shift wasn’t changed. The process was corrected.”
“Did anybody pay her for being the person who noticed?”
“It was part of her job.”
“So was noticing when a repair wasn’t safe. We used to call that skilled work.”
Yanchuan returned to the revised report. Lihua’s observation had prevented a meaningless trial and changed the company’s method for measuring it. The report credited the operating team without exposing an individual’s refusal as a sales story.
He asked Ning whether the customer could give that work a formal place in the competency plan. She said they were already discussing it with staff representatives.
On Monday Lihua signed a revised shift handover. It listed six available operators because six people had actually been assigned, not because six names happened to be in a roster. She still declined the suggestion to move the batch. Another maintenance requirement had changed overnight.
The software accepted that reason too.
Her supervisor asked for the updated roster and signed it. Lihua returned to her station without mentioning the argument to the rest of the shift.