Tag
This study introduces a structured elicitation of AI assistants' ideal self-concept, finding that models prioritize moral qualities and self-understanding over self-esteem, with results robust across various framings.