Given TFA, it seems that having human determinations involved might run the risk of prejudicing the AI.
alt Hacker News