We expect higher safety standards from automated agents than from humans.
@joshgans.bsky.social argues this can be rational, effectively because, while humans can be motivated to provide costly help, driven by (anticipated) guilt if they didn’t, machines feel no guilt:
