Reasoning about nuclear armageddon is hard because it's so singular. Traditional risk/benefit assessments don't really work when the outcome is so catastrophic

Reasoning about nuclear armageddon is hard because it's so singular. Traditional risk/benefit assessments don't really work when the outcome is so catastrophic that its utility goes to (negative) infinity. But I'd argue that this makes armageddon less plausible, because our psychological judgement of MAD is fundamentally different from normal risk assessment, and probably more cautious. A useful analogy is individual risk. Consider a doctor giving a new treatment to a sick patient. It has some uncertainty, some risk including death and some potential gain including being healed. But it's possible to reason about the tradeoff and learn to better estimate this over time. However, to a single patient, while there are both risk and potential benefit, the realized outcome is absolute: Either he lives or he dies. The mental model is very different from the doctors' because both scenarios are extremely charged. In nuclear war, you are always the patient who either lives completely or dies completely. I guess that makes everyone humble. Or at least, it should.