With a sufficiently strong LLM, I think you could still elicit reports of inner dialogs if you prompt lightly, such as “put yourself into the shoes of...”. That’s because inner monologs are implied in many reasoning processes, even if not explicitly mentioned so.
With a sufficiently strong LLM, I think you could still elicit reports of inner dialogs if you prompt lightly, such as “put yourself into the shoes of...”. That’s because inner monologs are implied in many reasoning processes, even if not explicitly mentioned so.