Anthropomorphism in synthetic agents: A relational view on AI ethics