> LLM decisionmaking cannot possibly be like human decisionmaking

I mean how can it possibly be like human decisionmaking? It's not like it's trained on human data