I don't think the that the big trick is utilizing FEA for this; rather the big trick is trying to quantify 'feel' (i.e. what parameters describe good 'feel' vs. bad 'feel' -- acceleration?, jerk?, others?).
If one knows how to quantify this, then numerical simluation can be done. I know people in automotive who have quantified 'good' door handle opening behavior and then successfully simulated it (similarly I am familiar with people quantifying 'good' door slam based on door excitations).
If the reality can be quantified, then the numerical simulation can likely be tied to it. Unfortunately, objectively quantifying these things is very difficult.
Brad