"Anthropomorphization of AI" refers to the widespread practice of describing artificial intelligence systems using language and concepts associated with human minds, agency, and experience, beyond their technical context. This could include but is not limited to:
- describing AI outputs using cognitive or experiential terms, such as saying a model "thinks," "reasons," "understands," "hallucinates," or "lies"
- referring to AI systems as "agents" or "assistants" with goals, intentions, or personalities, including giving them names or personas
- framing interactions with AI in relational terms, such as saying an AI is "helpful," "creative," "friendly," or that it "wants" or "refuses" to do something