@BalrogDeMorgoth Every singe person's empathy is different from each other. Of course it would come out as different from yours or some one else's but if it was programmed for such important thing, it would be as near objective morality as you could possibly get cause it's morality would be based on people consensus.
I just don't understand what you have problem with. You can have same AI without any empathy. That AI would be least judgmental object that can make any decisions.
You do understand that by just adding labels of accuser or accused nothing really happens. If such AI was programmed, it would hold some specific values over evidence presented, and if enough value was presented then punishment would be given based on committed crime.
No one would punish based on accusation. It isn't human that could be bribed or pushed around. How you programmed, that how it will act.
It seems that we will go back and forth with same arguments. So instead of that, could you answer few of my questions. Why do you think AI would go from initial labels to instant judgement when it was programmed to act based on evidence (being impartial). And why does empathy in an AI even matters if we only need cold logical judgement