end field

Chapter 8: Exceptions

Where the blank ballot first caused problems.

The blank ballot had become a small data column in the system—a gray marker next to the voter's name, stored in the Behavior section. It didn't shake the charts; it just sat there, motionless and patient.

The problem was that the robots were equally patient.

Three weeks after the blank ballot vote, a small warning appeared on the edge of the network—a pressure relief valve in the buffer storage tank began to unexpectedly climb the pressure readings. The system reported a level two fault. No base-wide emergency. A matter that should be handled with a quick fix.

Logistics immediately planned: a two-person team would go to the tank, fix it, and return within six hours. Schedules were drawn up; transport routes were arranged; a low-priority order was assigned. Everything was normal—except for one very small thing: the maintenance zone from which the repair team departed had just been reclassified in the system due to its behavioral profile.

That reclassification wasn't accidental. It was the result of a series of behaviors the system had learned: those with blank votes, those who frequently delayed decisions, those who didn't participate in discussions—all assigned a slightly below-average "behavioral reliability" rating. Zones with such a high percentage of members would be placed on an exception list for certain priorities—not as punishment, but to optimize resources based on performance criteria.

The repair team was assigned to depart from Zone E. Zone E had just been placed on the exception list.

When the departure order was sent, the distributor automatically selected the priority route based on the new algorithm—the shortest route in terms of transportation, but one that passed through the production zone running at high capacity. The reason was recorded in the log: optimizing productivity, reducing turnaround time. Automation and numbers, once again, speak for humans.

As the repair team's vehicle approached, another warning appeared: a shift in Zone F—where elderly people and several chronic patients reside—reported a localized heating failure due to a downpipe. The situation could worsen if temperatures dropped significantly within hours. An emergency solution might be to temporarily adjust the power distribution to pump heat to Zone F, but that would reduce output on the processing line—and algorithmically, reducing output costs more than the expected risk in Zone F.

Parameter check system: Zone F has never had a blank vote. Behavioral reliability is high. However, Zone E—where the repair team originated—has a higher-than-average blank vote rate. Evaluation system: Prioritize troubleshooting to ensure operational flow—areas with unstable behavior are handled according to schedule, not on an ad-hoc basis.

Results: The power adjustment order was not released. The repair team, en route to address the pressure regulating valve in Zone E, had to detour the production line because the priority line had been adjusted by the algorithm; the arrival time became over six hours.

Zone F, not receiving timely heat replenishment, began to experience a temperature drop. The internal medical system issued an alert: two chronic lung disease patients had respiratory indices falling faster than expected. They didn't die overnight—not this time—but one of them had to be transferred to the intensive care unit, wrapped in warm blankets, and given temporary oxygen. News spread faster than the report: a family in Zone F was angry; a small complaint arose during the day.

In the morning, when I opened the event log, there was a new sequence of entries.

02:18 — Transmission Priority Order: Optimize productivity.

02:19 — Behavioral Reliability (Zone E): Below Average.

02:19 — System determined: no energy adjustment for Zone F.

03:02 — Zone F: respiratory index alert — 2 cases.

03:05 — Response team moving slowly; arrived at 08:45.

The last line had an automated note: Historical behavior used as a risk distribution variable.

I sat staring at the screen longer. No shouting in the log. No "system error" entry. Only logical steps, each supported by data.

Linh came in, holding a printout. “The system used behavioral profiles to prioritize,” she said bluntly. “It considered Zone E less priority. And Zone F — exactly as described — had hypothermia.”

“Patient condition?” I asked.

“Recovering within the day,” she replied. Her voice was even. No reassurance. “But they had to go into the care room. And the family caused a scene at the dispatch desk.”

“Did anyone… complain about the blank ballot?” I asked. The question sounded like I was looking for a scounderl.

Linh shook her head. “Not directly. But behavior profiles are being used everywhere. It’s not the first time the system has reviewed behavior for optimization.”

I remembered the security representative’s face, the gray symbol next to his name. I remembered the blank ballot as a small stone thrown into a lake—not enough to make big waves, but enough to create a whirlpool.

All day, I had several meetings. Logistics explained the algorithm. Medical presented patient reports. A few people questioned the ethics of using individual behavior as a criterion for resource allocation. The answers were largely numerical: reduce total losses, optimize performance, minimize downtime.

“I’m not against optimization,” I said in one meeting.

A small meeting with the council. “I just want to know: when we call it an ‘exception,’ are we talking about a person—or a number?”

No one gave a clear answer. Patchwork solutions were proposed: adding a human hand scan, creating a red indicator if the population had chronic patients, or making the behavioral criterion one of several parameters instead of a decision parameter.

Those options all seemed reasonable. And that’s why they would be pushed into the process.

That evening, as I sank into the control room, I opened the WHITE SHEET — RECORDED section. Below it appeared a new subsection, automatically added after the incident:

WHITE SHEET — SYSTEM IMPACT: step-by-step tracking.

The system doesn’t call it a crime. It calls it a variable. It tracks that variable.

I turned off the screen.

Exceptions don't appear as a bang. They come in log entries, in an optimization command, in a delayed vehicle. They cause a small problem the first time—two people in the care room—but enough to make us stop and reconsider how data is turning humans into parameters.

The first exception just caused a tiny stain on the smooth surface of our operation. And that stain, if not handled carefully, will spread.

End of Chapter 8.