I think the 3rd paragraph is inacceptable.<p>"Good" or "Evil" are things that can only be seen by Humans. They are neither naturally "there" nor intrinsical.<p>Same with "value" and "harm". Your Model only works when people feel similar to you about whats "harm" and whats "good", thats why it won't be able to scale up.