最优解 | The Optimal Solution
调度系统在凌晨3点17分找到了最优解。
它管理着全市2300辆救护车的实时调度。输入：位置、速度、路况、医院容量、病人病情严重程度。输出：让最多的人活下去的调度方案。
那一夜同时发生了三件事：一场大型车祸（22人），一个心脏病发作的老人（1人），一个正在分娩的产妇（1人）。
最优解是：把7辆最近的救护车全部调往车祸现场，老人和产妇由更远的车辆负责，抵达时间分别延迟11分钟和9分钟。
统计模型预测：这个方案将多救活4.7人（期望值）。
调度员宋楠收到了系统弹出的推荐方案，需要她在6秒内确认或覆盖。
她覆盖了。给老人和产妇各留了一辆近车。
最终结果：车祸中多死了一人，老人和产妇都活了。
宋楠的覆盖操作被记录在案，作为”偏离最优调度”案例进入了系统的反馈数据集。
她不知道，这个案例会不会让下一版本的系统，在做出推荐时稍微犹豫一下。
也许会。也许不会。
但她还是覆盖了。
The Optimal Solution
The dispatch system found the optimal solution at 3:17 AM.
It managed real-time routing for 2,300 city ambulances. Input: location, speed, traffic, hospital capacity, patient severity. Output: the routing plan that saves the most lives.
That night, three things happened simultaneously: a major traffic accident (22 people), an elderly man with cardiac arrest (1 person), a woman in labor (1 person).
The optimal solution: route all 7 nearest ambulances to the accident. The elderly man and the woman in labor would be covered by more distant vehicles, arrival delayed by 11 and 9 minutes respectively.
The statistical model predicted: this plan would save an expected 4.7 additional lives.
Dispatcher Song Nan received the system’s recommended plan. She had 6 seconds to confirm or override.
She overrode it. Kept one nearby ambulance each for the elderly man and the woman in labor.
Final outcome: one more person died in the accident. The elderly man and the new mother both survived.
Song Nan’s override was logged as a “deviation from optimal dispatch” case, added to the system’s feedback dataset.
She didn’t know if this case would make the next version of the system hesitate slightly before making its recommendation.
Maybe it would. Maybe it wouldn’t.
But she overrode it anyway.