Role-Play with Large Language Models

On we develop effective ways to describe their behaviour the one hand, it's natural to use the same in high-level terms without falling into folk-psychological language to describe dialogue the trap of anthropomorphism. In this paper, we agents that we use to describe human behaviour, foreground the concept of role-play. Casting dialogue to freely deploy words like "knows", "understands", agent behaviour in terms of role-play allows and "thinks". Attempting to avoid us to draw on familiar folk psychological terms, such phrases by using more scientifically precise without ascribing human characteristics to language substitutes often results in prose that is clumsy models they in fact lack. Two important and hard to follow. On the other hand, taken cases of dialogue agent behaviour are addressed too literally, such language promotes anthropomorphism, this way, namely (apparent) deception and (apparent) exaggerating the similarities between self-awareness.