If we program human emotions into a big dataset and keep crunching the algorithums the result should be something that mimics humans emotions so well you can't tell the difference.

We can argue if it really FEELS or not, but from our perspective it should be able to easily mimic all human behavior convincingly. Humans are not THAT complex, rather we tend to all act very similar, so we won't be that hard to mimic.