Because that is what people do in roleplaying situations if the option is there.

That anyone would ask language models to analyze circumstances, perform logic and reason or conjure an application of knowledge and skill is kind of their own fault.

It is a language model, it excels at rephrasing given ideas.

If you put nuke buttons under a flock of pigeons or toddlers just to see what happens, they might launch. It’s not much of a study.

i’m so sick of media pretending that LLMs are like a sentient person making decisions.

Well that’s a good way to win so yeah