This is an interesting thought. Though I wonder what happens, when the A.I. comes to the conclusion, that of all of the avaiable solutions lead to the same 'amount' of regret (assuming regret can be quantified in complex situations).
Who makes the call then? Ultimately it's up to the human agent again in such a scenario.
Also, regret ultimately seems to me a very subjective thing, so I'm interested how a computer would process it. But I've read the wiki-link you added, and found some answers, thanks.
Have a nice day!