> they have unnecessarily caused panic among 180,000 people Hindsight is 20/20. If the dam had failed, people would be saying "they unnecessarily killed 180,00

> they have unnecessarily caused panic among 180,000 people Hindsight is 20/20. If the dam had failed, people would be saying "they unnecessarily killed 180,000 people because they were too lazy and politically timid to order an evacuation." It's like tuning sensitivity vs. specificity in medical tests. In this case the cost of a false negative (thousands of deaths) is a lot higher than the cost of a false positive (a few days of disruption). So the threshold is set low so as to minimize the total expected cost, summing probability of outcome * cost to society. Perhaps counterintuitively, the goal is not to minimize the odds of being wrong. We should expect to see more false positives than false negatives. If it were any other way, the system would be working sub-optimally.