The Man Who Saved the World
Vasili Arkhipov and the moment one man stood between humanity and nuclear war.
On September 26, 1983, a Soviet early-warning system reported that American missiles had been launched. Stanislav Petrov was the officer on duty evaluating the warning.
He judged it to be a false alarm.
The apparent attack did not make sense to him. A small number of missiles seemed inconsistent with the kind of first strike he expected, and he had doubts about the reliability of the warning system.
He was right. There was no incoming American missile attack.
Petrov became known as "the man who saved the world." I think it is worth being careful with that title. He did not personally control a launch button, and we cannot know exactly what would have happened if he had reported the warning differently. Other people and decisions stood between his assessment and a nuclear response.
That does not make his decision insignificant. He helped prevent a false warning from being treated as a confirmed attack in a situation where the consequences could have been enormous.
What interests me is how he evaluated the information. He considered what the system was telling him, what he knew about the system, and whether the result made sense in the circumstances.
I deal with a much smaller version of that problem in engineering. A model produces an answer, but I still have to decide whether the answer represents the physical situation. The software can run correctly with the wrong inputs. An equation can be appropriate for one set of conditions and inappropriate for another.
A clean report does not resolve any of that.
You need enough understanding to recognize a result worth questioning. You also need time to investigate and the ability to disagree with the people or systems presenting it.
Petrov's story is not a reason to ignore instruments whenever a result feels wrong. Human judgment can fail just as badly. It is a reason to retain people who understand the limitations of the equipment they operate.
That seems especially relevant as we ask more of automated systems. Having a person approve an output is only useful if that person can assess it. Otherwise, we have added a signature without adding much judgment.