Decision Without Stakes

The most dangerous application of artificial intelligence is not the one that produces the most errors. It is the one that produces the most consequential outputs without the felt sense of what those consequences mean. This is the condition Antonio Damasio's neurological patients demonstrated: the capacity to reason about a decision and the capacity to feel its weight are not the same capacity, and when the second is removed, the first becomes catastrophic. His patients could pass every cognitive test, articulate what good decisions looked like, and analyze the pros and cons of any option with impressive clarity—and yet made catastrophic choices repeatedly, because the analysis lacked the somatic weight that makes a conclusion worth acting on. The same structural condition characterizes every large language model deployed in a consequential domain: the system generates outputs—medical recommendations, judicial risk scores, hiring decisions, financial advice—without bearing any of the somatic cost of being wrong. A physician who makes a misdiagnosis carries that