It's frustrating to make mistakes, but they are our mistakes to make. To play without risk of an error is a less fulfilling experience.
Similarly you might ask why not allow a certain number of take backs during a game?
Actually... I think a certain amount of take backs would be a good idea as well. There is more to chess than calculating tactics and yet the higher thinking in chess has to give way to a hugeeee tree and to extensive blunder-checking for every move (especially in long games). I also think both players should be able to beat the computer algorithm easily based on their own positional skill, so the computer is just a light check on clear blunders that can ruin games. It's not like you won't have to be looking ahead. There are moves that are "right" or "wrong" positionally, including positional calculating blunders, but at least they don't cost the game.
I'm not saying it should overtake chess as the main game, but it could be some variant of it.
Every chess player at one time or another has felt the pain of a beautiful, well-orchestrated build up of pressure suddenly in ruins because of a pawn fork or some other terrible oversight. The only solution is to go through every variation many times and make sure you are not overlooking anything. At some point in tactical positions you just have to go through and test many different variations in your head, and sometimes a fork move in 10 moves time can only be described as luck.
So why not Advanced chess, where both sides have a sort of chess tactic "calculator"?