Positive reinforcement which is perceived as insincere will at best have no effect and at worst be seriously insulting. I can't imagine how an automated system which says vaguely positive things about code that it cannot understand would be seen as anything but insincere.

If you build a system which successfully encourages actual humans to say nice things about one another's code _and mean it_ while incorporating constructive feedback, it would work beautifully, but that is a markedly more challenging problem...